apps/worker/video_processing/dedup.py

@@ -1,8 +1,12 @@
-"""Video deduplication module - compute fingerprints and detect duplicates."""
+"""Video deduplication module - compute fingerprints and detect duplicates.
+
+Dynamic keyframe detection + sliding window temporal matching (Issue #1659).
+"""
 
 import hashlib
 import logging
 import os
+import statistics
 import tempfile
 from dataclasses import dataclass, field
 from typing import Optional
@@ -21,10 +25,28 @@ from packages.shared.storage import get_storage_service
 
 logger = logging.getLogger(__name__)
 
-# 分片策略常量
-SHORT_VIDEO_CHUNK_SEC = 2  # ≤60秒视频，每 2 秒一个分片
-LONG_VIDEO_CHUNK_SEC = 5  # >60秒视频，每 5 秒一个分片
-SHORT_VIDEO_THRESHOLD_SEC = 60
+# ── 关键帧检测常量 ──────────────────────────────────────────────
+SCENE_CHANGE_THRESHOLD = 30  # 灰度差异阈值
+MIN_KEYFRAME_INTERVAL_SEC = 1.0  # 最小关键帧间隔（秒）
+MAX_KEYFRAMES = 30  # 最大关键帧数
+MIN_KEYFRAMES = 5  # 最小关键帧数
+LONG_VIDEO_SEGMENT_SEC = 30  # 长视频每段秒数
+LONG_VIDEO_DURATION_THRESHOLD_SEC = 180  # 3 分钟阈值
+MIN_FRAMES_PER_SEGMENT = 2  # 长视频每段最少帧数
+
+# ── 滑动窗口匹配常量 ────────────────────────────────────────────
+SEGMENT_MATCH_THRESHOLD = 8  # 帧匹配汉明距离阈值
+MIN_CONSECUTIVE_MATCHES = 5  # 最少连续匹配帧数
+MAX_GAP = 2  # 允许的最大间隙帧数
+
+# ── 融合判定常量 ────────────────────────────────────────────────
+PHASH_WEIGHT = 0.7  # pHash 权重
+HISTOGRAM_WEIGHT = 0.3  # 直方图权重
+MATCH_RATIO_THRESHOLD = 0.7  # 至少 70% 帧匹配
+DUPLICATE_THRESHOLD = 0.70  # 融合后相似度阈值
+
+
+# ── 感知哈希 & 颜色直方图工具函数 ────────────────────────────────
 
 
 def compute_phash(image: np.ndarray, hash_size: int = 8) -> str:
@@ -87,15 +109,107 @@ def compute_color_histogram(image: np.ndarray, bins: int = 32) -> list[float]:
     return hist
 
 
-def compute_chunk_interval(duration: float) -> float:
-    """根据视频时长返回分片间隔（秒）。
+# ── 关键帧检测 ──────────────────────────────────────────────────
 
-    短视频（≤60秒）：每 2 秒一个分片
-    长视频（>60秒）：每 5 秒一个分片
+
+def detect_keyframe_timestamps(
+    video_path: str,
+    *,
+    min_interval_sec: float = MIN_KEYFRAME_INTERVAL_SEC,
+    max_frames: int = MAX_KEYFRAMES,
+    min_frames: int = MIN_KEYFRAMES,
+) -> list[float]:
+    """检测视频中的场景切换点，返回关键帧时间戳列表（秒）。
+
+    算法：
+    1. 降采样到 320x240，逐帧转灰度
+    2. 计算相邻帧灰度差异（像素均值差）
+    3. 差异 > SCENE_CHANGE_THRESHOLD(30) 标记为候选关键帧
+    4. 相邻关键帧间隔 < min_interval_sec 的，保留差异更大的那个
+    5. 数量裁剪到 [min_frames, max_frames]
+
+    对于长视频（>3分钟）：
+    - 每 30 秒一个分段
+    - 每个分段至少选 2 个关键帧（如果分段内无场景切换，均匀取 2 帧）
     """
-    if duration <= SHORT_VIDEO_THRESHOLD_SEC:
-        return SHORT_VIDEO_CHUNK_SEC
-    return LONG_VIDEO_CHUNK_SEC
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        raise RuntimeError(f"Cannot open video: {video_path}")
+
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    duration = frame_count / fps if fps > 0 else 0
+
+    if duration <= 0:
+        cap.release()
+        return []
+
+    # 逐帧检测场景切换
+    candidates: list[tuple[float, float]] = []  # (timestamp_sec, diff_score)
+    prev_gray = None
+
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+
+        # 降采样 + 灰度
+        small = cv2.resize(frame, (320, 240))
+        gray = cv2.cvtColor(small, cv2.COLOR_BGR2GRAY).astype(np.float32)
+
+        if prev_gray is not None:
+            diff = float(np.mean(np.abs(gray - prev_gray)))
+            if diff > SCENE_CHANGE_THRESHOLD:
+                pos_ms = cap.get(cv2.CAP_PROP_POS_MSEC)
+                candidates.append((pos_ms / 1000.0, diff))
+
+        prev_gray = gray
+
+    cap.release()
+
+    # 按最小间隔过滤（保留差异更大的）
+    filtered: list[tuple[float, float]] = []
+    for ts, diff in sorted(candidates):
+        if filtered and (ts - filtered[-1][0]) < min_interval_sec:
+            if diff > filtered[-1][1]:
+                filtered[-1] = (ts, diff)
+        else:
+            filtered.append((ts, diff))
+
+    keyframe_times = [ts for ts, _ in filtered]
+
+    # 数量不足 min_frames 时，在时间轴上均匀补充
+    if len(keyframe_times) < min_frames:
+        uniform = [duration * (i + 0.5) / min_frames for i in range(min_frames)]
+        keyframe_times = sorted(set(uniform) | set(keyframe_times))
+        # 如果合并后还不足 min_frames，直接用均匀分布
+        if len(keyframe_times) < min_frames:
+            keyframe_times = uniform
+
+    # 数量超过 max_frames 时，均匀采样
+    if len(keyframe_times) > max_frames:
+        step = len(keyframe_times) / max_frames
+        keyframe_times = [keyframe_times[int(i * step)] for i in range(max_frames)]
+
+    # 长视频分段保底（>3分钟）
+    if duration > LONG_VIDEO_DURATION_THRESHOLD_SEC:
+        segment_count = int(duration / LONG_VIDEO_SEGMENT_SEC)
+        for seg_idx in range(segment_count):
+            seg_start = seg_idx * LONG_VIDEO_SEGMENT_SEC
+            seg_end = min((seg_idx + 1) * LONG_VIDEO_SEGMENT_SEC, duration)
+            seg_frames = [t for t in keyframe_times if seg_start <= t < seg_end]
+            if len(seg_frames) < MIN_FRAMES_PER_SEGMENT:
+                # 均匀补齐
+                for i in range(MIN_FRAMES_PER_SEGMENT):
+                    t = seg_start + LONG_VIDEO_SEGMENT_SEC * (i + 0.5) / MIN_FRAMES_PER_SEGMENT
+                    if t not in keyframe_times and seg_start <= t < seg_end:
+                        keyframe_times.append(t)
+        keyframe_times.sort()
+
+    return keyframe_times
+
+
+# ── 数据类 ──────────────────────────────────────────────────────
 
 
 @dataclass
@@ -109,6 +223,17 @@ class FingerprintChunk:
     frame_count: int = 1
 
 
+@dataclass
+class DuplicateSegment:
+    """一段重复片段的描述。"""
+
+    query_start_ms: int
+    query_end_ms: int
+    target_start_ms: int
+    target_end_ms: int
+    avg_distance: float  # 该段内帧的平均汉明距离
+
+
 @dataclass
 class VideoFingerprint:
     """Video fingerprint containing multiple similarity metrics."""
@@ -163,6 +288,137 @@ class VideoFingerprint:
         return models
 
 
+# ── 滑动窗口时序匹配 ────────────────────────────────────────────
+
+
+def find_duplicate_segments(
+    query_chunks: list,
+    target_chunks: list,
+    *,
+    match_threshold: int = SEGMENT_MATCH_THRESHOLD,
+    min_consecutive: int = MIN_CONSECUTIVE_MATCHES,
+    max_gap: int = MAX_GAP,
+) -> list[DuplicateSegment]:
+    """滑动窗口时序匹配：找出两组分片之间的重复片段。
+
+    算法：
+    1. 对每个 query chunk，找到 target 中汉明距离最小的 chunk
+    2. 距离 <= match_threshold 视为匹配
+    3. 找连续匹配的 run（允许 max_gap 帧间隙）
+    4. 连续匹配数 >= min_consecutive 的 run 报告为重复片段
+
+    Args:
+        query_chunks: 查询视频的分片列表（FingerprintChunk 或 dict）
+        target_chunks: 目标视频的分片列表
+        match_threshold: 汉明距离匹配阈值
+        min_consecutive: 最少连续匹配帧数
+        max_gap: 允许的最大间隙帧数
+
+    Returns:
+        DuplicateSegment 列表
+    """
+    if not query_chunks or not target_chunks:
+        return []
+
+    def _get_phash(chunk) -> str:
+        if isinstance(chunk, dict):
+            return chunk["phash_binary"]
+        return chunk.phash_binary
+
+    def _get_start(chunk) -> int:
+        if isinstance(chunk, dict):
+            return chunk["start_time_ms"]
+        return chunk.start_time_ms
+
+    def _get_end(chunk) -> int:
+        if isinstance(chunk, dict):
+            return chunk["end_time_ms"]
+        return chunk.end_time_ms
+
+    # Step 1: 逐帧匹配
+    frame_matches: list[tuple[bool, int, int]] = []  # (is_match, min_dist, best_target_idx)
+    for qc in query_chunks:
+        qc_phash = _get_phash(qc)
+        best_dist = 64
+        best_idx = 0
+        for j, tc in enumerate(target_chunks):
+            d = hamming_distance(qc_phash, _get_phash(tc))
+            if d < best_dist:
+                best_dist = d
+                best_idx = j
+        frame_matches.append((best_dist <= match_threshold, best_dist, best_idx))
+
+    # Step 2: 找连续匹配的 runs
+    runs: list[tuple[int, int]] = []  # list of (start_idx, end_idx)
+    run_start = None
+    gap_count = 0
+
+    for i, (is_match, _dist, _idx) in enumerate(frame_matches):
+        if is_match:
+            if run_start is None:
+                run_start = i
+            gap_count = 0  # 重置间隙
+        else:
+            if run_start is not None:
+                gap_count += 1
+                if gap_count > max_gap:
+                    # 中断当前 run
+                    run_end = i - gap_count  # 最后一个匹配帧的索引
+                    # 计算 run 内的实际匹配帧数（总跨度 - 间隙数）
+                    total_gaps = sum(1 for k in range(run_start, run_end + 1) if not frame_matches[k][0])
+                    matching_count = (run_end - run_start + 1) - total_gaps
+                    if matching_count >= min_consecutive:
+                        runs.append((run_start, run_end))
+                    run_start = None
+                    gap_count = 0
+
+    # 处理末尾 run
+    if run_start is not None:
+        last_idx = len(frame_matches) - 1
+        # 回退找到最后一个匹配帧的位置（跳过尾部非匹配帧）
+        while last_idx >= run_start and not frame_matches[last_idx][0]:
+            last_idx -= 1
+        if last_idx >= run_start:
+            # 计算 run 内的总间隙数
+            total_gaps = sum(1 for k in range(run_start, last_idx + 1) if not frame_matches[k][0])
+            matching_count = (last_idx - run_start + 1) - total_gaps
+            if matching_count >= min_consecutive:
+                runs.append((run_start, last_idx))
+
+    # Step 3: 构建 DuplicateSegment
+    segments: list[DuplicateSegment] = []
+    for start, end in runs:
+        query_start = _get_start(query_chunks[start])
+        query_end = _get_end(query_chunks[end])
+
+        # 取目标范围（按最佳匹配的目标 chunk 时间范围）
+        target_indices = [frame_matches[k][2] for k in range(start, end + 1) if frame_matches[k][0]]
+        if target_indices:
+            t_min = min(target_indices)
+            t_max = max(target_indices)
+            target_start = _get_start(target_chunks[t_min])
+            target_end = _get_end(target_chunks[t_max])
+        else:
+            target_start = _get_start(target_chunks[0])
+            target_end = _get_end(target_chunks[-1])
+
+        avg_dist = sum(frame_matches[k][1] for k in range(start, end + 1)) / (end - start + 1)
+        segments.append(
+            DuplicateSegment(
+                query_start_ms=query_start,
+                query_end_ms=query_end,
+                target_start_ms=target_start,
+                target_end_ms=target_end,
+                avg_distance=avg_dist,
+            )
+        )
+
+    return segments
+
+
+# ── VideoDeduplicator ───────────────────────────────────────────
+
+
 class VideoDeduplicator:
     """Video deduplication using multiple fingerprint methods."""
 
@@ -170,11 +426,11 @@ class VideoDeduplicator:
     HISTOGRAM_THRESHOLD = 0.85
 
     def compute_fingerprint(self, video_path: str) -> VideoFingerprint:
-        """Compute video fingerprint using MD5, pHash, and color histogram.
+        """Compute video fingerprint using dynamic keyframe detection.
 
-        按时间分片抽帧：短视频（≤60s）每 2s 一片，长视频每 5s 一片。
-        每片取 1 帧计算 pHash + color_histogram。
-        同时保留 keyframe_phashes/color_histograms 聚合字段（向后兼容）。
+        使用 detect_keyframe_timestamps() 检测内容感知关键帧，
+        在每个关键帧处取帧计算 pHash + color_histogram。
+        同时保留 MD5 计算和分片数据结构。
         """
         cap = cv2.VideoCapture(video_path)
         if not cap.isOpened():
@@ -186,41 +442,55 @@ class VideoDeduplicator:
         width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
         height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
 
+        cap.release()
+
+        # 1. 检测关键帧时间戳
+        keyframe_times = detect_keyframe_timestamps(video_path)
+
+        if not keyframe_times:
+            return VideoFingerprint(
+                md5="",
+                keyframe_phashes=[],
+                color_histograms=[],
+                duration=duration,
+                resolution=(width, height),
+                chunks=[],
+            )
+
+        # 2. 打开视频，逐个关键帧取帧
+        cap = cv2.VideoCapture(video_path)
         md5_hash = hashlib.md5(usedforsecurity=False)
         chunks: list[FingerprintChunk] = []
 
-        # 分片间隔（秒）
-        chunk_interval_sec = compute_chunk_interval(duration)
-        chunk_interval_ms = int(chunk_interval_sec * 1000)
-        duration_ms = int(duration * 1000)
-
-        # 遍历每个分片时间窗口，取 1 帧
-        start_ms = 0
-        while start_ms < duration_ms:
-            end_ms = min(start_ms + chunk_interval_ms, duration_ms)
-            # 定位到分片中点
-            seek_ms = (start_ms + end_ms) / 2
+        for i, t_sec in enumerate(keyframe_times):
+            seek_ms = t_sec * 1000
             cap.set(cv2.CAP_PROP_POS_MSEC, seek_ms)
             ret, frame = cap.read()
-            if ret:
-                # MD5 计算
-                _, buffer = cv2.imencode(".jpg", frame)
-                md5_hash.update(buffer)
+            if not ret:
+                continue
 
-                phash = compute_phash(frame)
-                hist = compute_color_histogram(frame)
+            # MD5 计算
+            _, buffer = cv2.imencode(".jpg", frame)
+            md5_hash.update(buffer)
 
-                chunks.append(
-                    FingerprintChunk(
-                        start_time_ms=start_ms,
-                        end_time_ms=end_ms,
-                        phash_binary=phash,
-                        color_histogram=hist,
-                        frame_count=1,
-                    )
+            phash = compute_phash(frame)
+            hist = compute_color_histogram(frame)
+
+            # 计算分片时间范围（从前一个关键帧到下一个关键帧的中点）
+            prev_boundary = keyframe_times[i - 1] * 1000 if i > 0 else 0
+            next_boundary = keyframe_times[i + 1] * 1000 if i < len(keyframe_times) - 1 else duration * 1000
+            start_ms = int((prev_boundary + seek_ms) / 2)
+            end_ms = int((seek_ms + next_boundary) / 2)
+
+            chunks.append(
+                FingerprintChunk(
+                    start_time_ms=start_ms,
+                    end_time_ms=end_ms,
+                    phash_binary=phash,
+                    color_histogram=hist,
+                    frame_count=1,
                 )
-
-            start_ms = end_ms
+            )
 
         cap.release()
 
@@ -255,14 +525,39 @@ class VideoDeduplicator:
             for r in rows
         ]
 
+    @staticmethod
+    def _bhattacharyya_coefficient(hist_a: list[float], hist_b: list[float]) -> float:
+        """Bhattacharyya 系数：Σ √(a[i] * b[i])，范围 [0, 1]，1=完全相同。"""
+        min_len = min(len(hist_a), len(hist_b))
+        a = hist_a[:min_len]
+        b = hist_b[:min_len]
+        return float(sum(np.sqrt(ai * bi) for ai, bi in zip(a, b, strict=False)))
+
+    @staticmethod
+    def _compute_histogram_similarity(
+        histograms_a: list[list[float]],
+        histograms_b: list[list[float]],
+    ) -> float:
+        """对每组直方图，找到最佳匹配的 Bhattacharyya 系数，取平均。"""
+        if not histograms_a or not histograms_b:
+            return 0.0
+        similarities = []
+        for ha in histograms_a:
+            best = 0.0
+            for hb in histograms_b:
+                bc = VideoDeduplicator._bhattacharyya_coefficient(ha, hb)
+                best = max(best, bc)
+            similarities.append(best)
+        return sum(similarities) / len(similarities) if similarities else 0.0
+
     def check_duplicate(self, fingerprint: VideoFingerprint, project_id: str, session: Session) -> Optional[dict]:
         """检查视频是否与项目中已有视频重复。
 
         查重逻辑：
         1. MD5 精确匹配 → similarity=1.0
-        2. pHash 相似度（优先从分片表读取，回退到 JSON 字段）
+        2. pHash 中位数距离 + 帧匹配比例 + 直方图融合判定
 
-        判定阈值：avg_distance < PHASH_THRESHOLD(10)
+        判定为重复后，调用 find_duplicate_segments() 获取具体重复片段。
 
         Args:
             fingerprint: 待检测视频的指纹
@@ -270,7 +565,7 @@ class VideoDeduplicator:
             session: 数据库会话
 
         Returns:
-            重复信息字典（含 duplicate, duplicate_of, reason, similarity），
+            重复信息字典（含 duplicate, duplicate_of, reason, similarity, duplicate_segments），
             或 None 表示未找到重复。
         """
         video_repo = SQLAlchemyGeneratedVideoRepository(session)
@@ -298,23 +593,65 @@ class VideoDeduplicator:
             if not existing_phashes:
                 continue
 
-            # 计算每个新关键帧到已有关键帧的最小汉明距离，取平均
+            # 计算每个新关键帧到已有关键帧的最小汉明距离
             min_distances = []
             for phash in fingerprint.keyframe_phashes:
                 distances = [hamming_distance(phash, ep) for ep in existing_phashes]
                 min_distances.append(min(distances))
-            avg_distance = sum(min_distances) / len(min_distances) if min_distances else 100
 
-            if avg_distance >= self.PHASH_THRESHOLD:
+            # 帧匹配比例检查
+            matching_frames = sum(1 for d in min_distances if d < self.PHASH_THRESHOLD)
+            match_ratio = matching_frames / len(min_distances) if min_distances else 0
+            if match_ratio < 0.7:
                 continue
 
-            phash_similarity = 1.0 - (avg_distance / 64)
+            # 中位数距离
+            median_distance = statistics.median(min_distances) if min_distances else 64
+            if median_distance >= self.PHASH_THRESHOLD:
+                continue
+
+            # 直方图融合
+            existing_histograms = []
+            if chunk_data:
+                existing_histograms = [c["color_histogram"] for c in chunk_data if c.get("color_histogram")]
+            else:
+                existing_histograms = ef.get("color_histograms", [])
+
+            phash_similarity = 1.0 - (median_distance / 64)
+            hist_similarity = (
+                self._compute_histogram_similarity(fingerprint.color_histograms, existing_histograms)
+                if existing_histograms
+                else 0.5
+            )
+            combined_score = 0.7 * phash_similarity + 0.3 * hist_similarity
+
+            # DUPLICATE_THRESHOLD from module level
+            if combined_score < DUPLICATE_THRESHOLD:
+                continue
+
+            # 滑动窗口时序匹配：获取具体重复片段
+            existing_chunk_objects = (
+                chunk_data
+                if chunk_data
+                else [{"phash_binary": p, "start_time_ms": 0, "end_time_ms": 0} for p in existing_phashes]
+            )
+            segments = find_duplicate_segments(fingerprint.chunks, existing_chunk_objects)
 
             return {
                 "duplicate": True,
                 "duplicate_of": existing.id,
-                "reason": "phash_similar",
-                "similarity": phash_similarity,
+                "reason": "phash_histogram_fusion",
+                "similarity": combined_score,
+                "duplicate_segments": [
+                    {
+                        "query_start_ms": s.query_start_ms,
+                        "query_end_ms": s.query_end_ms,
+                        "target_start_ms": s.target_start_ms,
+                        "target_end_ms": s.target_end_ms,
+                        "avg_distance": round(s.avg_distance, 2),
+                    }
+                    for s in segments
+                ],
             }
 
         return None
@@ -328,7 +665,8 @@ class VideoDeduplicator:
     ) -> Optional[dict]:
         """检查视频是否与同批次内其他视频重复。
 
-        逻辑与 check_duplicate 一致（MD5 + pHash），但搜索范围限定为同 batch_id 的视频。
+        逻辑与 check_duplicate 一致（MD5 + pHash + 直方图融合 + 时序匹配），
+        但搜索范围限定为同 batch_id 的视频。
 
         Args:
             fingerprint: 待检测视频的指纹
@@ -373,59 +711,63 @@ class VideoDeduplicator:
             for phash in fingerprint.keyframe_phashes:
                 distances = [hamming_distance(phash, ep) for ep in existing_phashes]
                 min_distances.append(min(distances))
-            avg_distance = sum(min_distances) / len(min_distances) if min_distances else 100
 
-            if avg_distance >= self.PHASH_THRESHOLD:
+            # 帧匹配比例检查
+            matching_frames = sum(1 for d in min_distances if d < self.PHASH_THRESHOLD)
+            match_ratio = matching_frames / len(min_distances) if min_distances else 0
+            if match_ratio < 0.7:
                 continue
 
-            phash_similarity = 1.0 - (avg_distance / 64)
+            median_distance = statistics.median(min_distances) if min_distances else 64
+            if median_distance >= self.PHASH_THRESHOLD:
+                continue
+
+            # 直方图融合
+            existing_histograms = []
+            if chunk_data:
+                existing_histograms = [c["color_histogram"] for c in chunk_data if c.get("color_histogram")]
+            else:
+                existing_histograms = ef.get("color_histograms", [])
+
+            phash_similarity = 1.0 - (median_distance / 64)
+            hist_similarity = (
+                self._compute_histogram_similarity(fingerprint.color_histograms, existing_histograms)
+                if existing_histograms
+                else 0.5
+            )
+            combined_score = 0.7 * phash_similarity + 0.3 * hist_similarity
+
+            # DUPLICATE_THRESHOLD from module level
+            if combined_score < DUPLICATE_THRESHOLD:
+                continue
+
+            # 滑动窗口时序匹配
+            existing_chunk_objects = (
+                chunk_data
+                if chunk_data
+                else [{"phash_binary": p, "start_time_ms": 0, "end_time_ms": 0} for p in existing_phashes]
+            )
+            segments = find_duplicate_segments(fingerprint.chunks, existing_chunk_objects)
+
             return {
                 "duplicate": True,
                 "duplicate_of": existing.id,
-                "reason": "batch_phash_similar",
-                "similarity": phash_similarity,
+                "reason": "batch_phash_histogram_fusion",
+                "similarity": combined_score,
+                "duplicate_segments": [
+                    {
+                        "query_start_ms": s.query_start_ms,
+                        "query_end_ms": s.query_end_ms,
+                        "target_start_ms": s.target_start_ms,
+                        "target_end_ms": s.target_end_ms,
+                        "avg_distance": round(s.avg_distance, 2),
+                    }
+                    for s in segments
+                ],
             }
 
         return None
 
-    @staticmethod
-    def _average_histogram_similarity(histograms_a: list[list[float]], histograms_b: list[list[float]]) -> float:
-        """
-        计算两组颜色直方图之间的平均余弦相似度。
-
-        对每组直方图对取最小长度对齐，计算余弦相似度后取平均。
-
-        Args:
-            histograms_a: 第一组直方图（每帧一个 list）
-            histograms_b: 第二组直方图
-
-        Returns:
-            平均余弦相似度，范围 [0, 1]
-        """
-        if not histograms_a or not histograms_b:
-            return 0.0
-
-        similarities = []
-        for ha in histograms_a:
-            best = 0.0
-            vec_a = np.array(ha, dtype=np.float64)
-            norm_a = np.linalg.norm(vec_a)
-            if norm_a == 0:
-                continue
-            for hb in histograms_b:
-                vec_b = np.array(hb, dtype=np.float64)
-                # 对齐长度
-                min_len = min(len(vec_a), len(vec_b))
-                va, vb = vec_a[:min_len], vec_b[:min_len]
-                norm_b = np.linalg.norm(vb)
-                if norm_b == 0:
-                    continue
-                sim = float(np.dot(va, vb) / (norm_a * norm_b))
-                best = max(best, sim)
-            similarities.append(best)
-
-        return sum(similarities) / len(similarities) if similarities else 0.0
-
     def compute_duplicate_rate(
         self,
         fingerprint: VideoFingerprint,
@@ -438,9 +780,9 @@ class VideoDeduplicator:
         """计算当前视频与用户库内已有视频的最高相似度百分比。
 
         优先按 user_id 全局比较（跨项目），user_id 为空时回退到项目级比较。
-        遍历最近 200 个其他有指纹的视频，对每个计算相似度：
+        遍历最近 200 个其他有指纹的视频，对每个计算融合相似度：
         - MD5 精确匹配 → 100%
-        - pHash 相似度 → (1.0 - avg_distance / 64) * 100
+        - pHash + 直方图融合 → 0.7 * phash_sim + 0.3 * hist_sim
         取最高值作为 duplicate_rate（0~100）。
         如果没有其他视频可比较，返回 0.0。
 
@@ -454,7 +796,6 @@ class VideoDeduplicator:
         Returns:
             duplicate_rate: 0~100 的浮点数
         """
-        # 限制查询最近 200 个视频，避免大库内存溢出
         from packages.adapters.sqlalchemy_impl.models import GeneratedVideoModel
 
         # 优先按 user_id 全局比较（跨项目），否则回退到项目级
@@ -469,7 +810,7 @@ class VideoDeduplicator:
             )
             logger.debug("compute_duplicate_rate: project-level fallback project_id=%s", project_id)
 
-        # 排除当前视频自身（记录可能已写入 DB，必须在查询层排除）
+        # 排除当前视频自身
         if current_video_id:
             query = query.filter(GeneratedVideoModel.id != current_video_id)
 
@@ -505,9 +846,30 @@ class VideoDeduplicator:
             for phash in fingerprint.keyframe_phashes:
                 distances = [hamming_distance(phash, ep) for ep in existing_phashes]
                 min_distances.append(min(distances))
-            avg_distance = sum(min_distances) / len(min_distances) if min_distances else 64
-            similarity = (1.0 - avg_distance / 64) * 100
-            max_similarity = max(max_similarity, similarity)
+
+            # 帧匹配比例检查
+            matching_frames = sum(1 for d in min_distances if d < self.PHASH_THRESHOLD)
+            match_ratio = matching_frames / len(min_distances) if min_distances else 0
+            if match_ratio < 0.7:
+                continue
+
+            median_distance = statistics.median(min_distances) if min_distances else 64
+
+            # 直方图融合
+            existing_histograms = []
+            if chunk_data:
+                existing_histograms = [c["color_histogram"] for c in chunk_data if c.get("color_histogram")]
+            else:
+                existing_histograms = ef.get("color_histograms", [])
+
+            phash_similarity = (1.0 - median_distance / 64) * 100
+            hist_similarity = (
+                self._compute_histogram_similarity(fingerprint.color_histograms, existing_histograms) * 100
+                if existing_histograms
+                else 50.0
+            )
+            combined_score = 0.7 * phash_similarity + 0.3 * hist_similarity
+            max_similarity = max(max_similarity, combined_score)
 
         return round(max(max_similarity, 0.0), 2)
 

tests/unit/test_cross_video_avoidance.py

@@ -221,7 +221,6 @@ class TestServiceLayerIntegration:
 
     def _make_service(self, clip_repo_mock, asset_repo_mock=None):
         """创建 PlanGeneratorService 并注入 mock repos."""
-        from unittest.mock import MagicMock, patch
 
         from apps.api.app.services.plan_generator_service import PlanGeneratorService
 

tests/unit/test_dedup_engine.py

@@ -285,8 +285,10 @@ class TestVideoDeduplicatorCheckDuplicate:
             result = deduplicator.check_duplicate(fingerprint, "proj-1", mock_session)
             assert result is not None
             assert result["duplicate"] is True
-            assert result["similarity"] == 1.0  # distance=0 → 1.0
-            assert result["reason"] == "phash_similar"
+            assert result["similarity"] == pytest.approx(
+                0.85, abs=0.01
+            )  # combined: 0.7*1.0 + 0.3*0.5 (no hist fallback)
+            assert result["reason"] == "phash_histogram_fusion"
         finally:
             self._restore_repo(mod, orig)
 
@@ -425,8 +427,11 @@ class TestVideoDeduplicatorCheckDuplicate:
             result = deduplicator.check_duplicate(fingerprint, "proj-1", mock_session)
             assert result is not None
             assert result["duplicate"] is True
-            # similarity = 1.0 - (1 / 64) = 0.984375
-            assert abs(result["similarity"] - (1.0 - 1.0 / 64)) < 1e-6
+            # 新算法: median_distance=1, phash_sim=1-1/64=0.984375
+            # 无直方图 → hist_sim=0.5(fallback)
+            # combined = 0.7*0.984375 + 0.3*0.5 = 0.839062
+            expected_sim = 0.7 * (1.0 - 1.0 / 64) + 0.3 * 0.5
+            assert abs(result["similarity"] - expected_sim) < 1e-6
         finally:
             self._restore_repo(mod, orig)
 
@@ -456,7 +461,9 @@ class TestVideoDeduplicatorCheckDuplicate:
             result = deduplicator.check_duplicate(fingerprint, "proj-1", mock_session)
             assert result is not None
             assert result["duplicate"] is True
-            assert result["similarity"] == 1.0  # avg_distance = 0
+            # 新算法: median_distance=0, phash_sim=1.0, hist_sim=0.5(fallback)
+            # combined = 0.7*1.0 + 0.3*0.5 = 0.85
+            assert result["similarity"] == pytest.approx(0.85, abs=0.01)
         finally:
             self._restore_repo(mod, orig)
 
@@ -539,7 +546,7 @@ class TestVideoDeduplicatorCheckBatchDuplicate:
             result = deduplicator.check_batch_duplicate(fingerprint, "batch-1", "vid-self", mock_session)
             assert result is not None
             assert result["duplicate"] is True
-            assert result["reason"] == "batch_phash_similar"
+            assert result["reason"] == "batch_phash_histogram_fusion"
         finally:
             self._restore_repo(mod, orig)
 

tests/unit/test_dedup_pure.py

@@ -181,70 +181,68 @@ class TestVideoFingerprint:
         assert d["color_histograms"] == []
 
 
-class TestAverageHistogramSimilarity:
-    """_average_histogram_similarity 直方图相似度测试."""
+class TestBhattacharyyaCoefficient:
+    """_bhattacharyya_coefficient Bhattacharyya 系数测试."""
 
     def test_identical_histograms(self):
-        """完全相同的直方图相似度为1.0."""
-        hist = [[0.5, 0.5, 0.0], [0.3, 0.4, 0.3]]
-        sim = VideoDeduplicator._average_histogram_similarity(hist, hist)
-        assert sim == pytest.approx(1.0)
+        """完全相同的直方图系数为1.0."""
+        hist = [0.5, 0.5, 0.0, 0.3]
+        bc = VideoDeduplicator._bhattacharyya_coefficient(hist, hist)
+        # Σ √(a[i]*a[i]) = Σ a[i] = 1.0 (normalized)
+        assert bc == pytest.approx(sum(h for h in hist))
 
-    def test_empty_first_list(self):
+    def test_zero_histograms(self):
+        """全零直方图系数为0."""
+        bc = VideoDeduplicator._bhattacharyya_coefficient([0.0, 0.0], [0.0, 0.0])
+        assert bc == 0.0
+
+    def test_orthogonal_histograms(self):
+        """正交直方图（无重叠）系数为0."""
+        bc = VideoDeduplicator._bhattacharyya_coefficient([1.0, 0.0], [0.0, 1.0])
+        assert bc == pytest.approx(0.0)
+
+    def test_different_lengths(self):
+        """不同长度直方图取最小长度对齐."""
+        bc = VideoDeduplicator._bhattacharyya_coefficient([1.0, 1.0, 0.0, 0.0], [1.0, 1.0])
+        # 对齐到前2维: √(1*1) + √(1*1) = 2.0
+        assert bc == pytest.approx(2.0)
+
+    def test_known_value(self):
+        """已知值验证."""
+        # [0.25, 0.25, 0.25, 0.25] vs [0.25, 0.25, 0.25, 0.25]
+        # BC = 4 * √(0.25 * 0.25) = 4 * 0.25 = 1.0
+        hist = [0.25, 0.25, 0.25, 0.25]
+        bc = VideoDeduplicator._bhattacharyya_coefficient(hist, hist)
+        assert bc == pytest.approx(1.0)
+
+
+class TestComputeHistogramSimilarity:
+    """_compute_histogram_similarity 多帧直方图相似度测试."""
+
+    def test_identical_histogram_groups(self):
+        """完全相同的两组直方图."""
+        hist = [[0.5, 0.5], [0.3, 0.4]]
+        sim = VideoDeduplicator._compute_histogram_similarity(hist, hist)
+        # Each hist finds best match = itself
+        assert sim > 0.0
+
+    def test_empty_first(self):
         """第一组为空返回0."""
-        sim = VideoDeduplicator._average_histogram_similarity([], [[0.5, 0.5]])
-        assert sim == 0.0
+        assert VideoDeduplicator._compute_histogram_similarity([], [[0.5]]) == 0.0
 
-    def test_empty_second_list(self):
+    def test_empty_second(self):
         """第二组为空返回0."""
-        sim = VideoDeduplicator._average_histogram_similarity([[0.5, 0.5]], [])
-        assert sim == 0.0
+        assert VideoDeduplicator._compute_histogram_similarity([[0.5]], []) == 0.0
 
     def test_both_empty(self):
         """两组都为空返回0."""
-        sim = VideoDeduplicator._average_histogram_similarity([], [])
-        assert sim == 0.0
+        assert VideoDeduplicator._compute_histogram_similarity([], []) == 0.0
 
-    def test_orthogonal_histograms(self):
-        """正交直方图相似度为0."""
-        # [1, 0] 和 [0, 1] 正交
-        sim = VideoDeduplicator._average_histogram_similarity([[1.0, 0.0]], [[0.0, 1.0]])
-        assert sim == pytest.approx(0.0)
-
-    def test_partial_similarity(self):
-        """部分相似."""
-        # [1, 1] 和 [1, 0] 的余弦相似度 = 1/√2 ≈ 0.707
-        sim = VideoDeduplicator._average_histogram_similarity([[1.0, 1.0]], [[1.0, 0.0]])
-        assert sim == pytest.approx(1.0 / (2**0.5), rel=0.01)
-
-    def test_multiple_frames_best_match(self):
+    def test_best_match_selection(self):
         """多帧时取最佳匹配."""
-        # 第一帧完全不同，第二帧完全相同 → 平均 best = (0 + 1) / 2 = 0.5
-        sim = VideoDeduplicator._average_histogram_similarity(
-            [[1.0, 0.0], [0.0, 1.0]],
-            [[0.0, 1.0]],  # 只有一帧，和第一帧0相似，和第二帧1相似
-        )
-        # 第一帧最佳匹配=0，第二帧最佳匹配=1，平均=0.5
-        assert sim == pytest.approx(0.5)
-
-    def test_zero_norm_histogram_skipped(self):
-        """零范数直方图被跳过."""
-        sim = VideoDeduplicator._average_histogram_similarity([[0.0, 0.0]], [[1.0, 1.0]])
-        # 第一组的零范数被跳过，similarities为空，返回0
-        assert sim == 0.0
-
-    def test_different_length_histograms(self):
-        """不同长度的直方图取最小长度对齐."""
-        sim = VideoDeduplicator._average_histogram_similarity(
-            [[1.0, 1.0, 0.0, 0.0]],  # 4维
-            [[1.0, 1.0]],  # 2维
-        )
-        # 对齐到前2维，都是[1,1]，相似度1.0
+        # ha[0] 与 hb[0] 正交，与 hb[1] 完全相同
+        a = [[1.0, 0.0]]
+        b = [[0.0, 1.0], [1.0, 0.0]]
+        sim = VideoDeduplicator._compute_histogram_similarity(a, b)
+        # Best match for [1,0]: max(BC([1,0],[0,1]), BC([1,0],[1,0])) = max(0, 1) = 1
         assert sim == pytest.approx(1.0)
-
-    def test_similarity_in_zero_one_range(self):
-        """相似度在[0, 1]范围内."""
-        hist_a = [np.random.rand(96).tolist() for _ in range(5)]
-        hist_b = [np.random.rand(96).tolist() for _ in range(5)]
-        sim = VideoDeduplicator._average_histogram_similarity(hist_a, hist_b)
-        assert 0.0 <= sim <= 1.0

tests/unit/test_dedup_v2.py

@@ -0,0 +1,532 @@
+"""Issue #1659: 动态抽帧 + 滑动窗口时序匹配 单元测试.
+
+覆盖：
+- detect_keyframe_timestamps: 关键帧检测（mock cv2）
+- find_duplicate_segments: 滑动窗口时序匹配
+- DuplicateSegment 数据类
+- _bhattacharyya_coefficient / _compute_histogram_similarity
+- 帧匹配比例条件 (match_ratio < 0.7 → 跳过)
+- 中位数 vs 均值（抵抗异常值）
+- 向后兼容（无分片数据时不崩溃）
+"""
+
+from __future__ import annotations
+
+import sys
+from unittest.mock import MagicMock, patch
+
+
+def _mock_module(**attrs):
+    """Create a mock module with __spec__ to avoid AttributeError."""
+    m = MagicMock()
+    m.__spec__ = None
+    for k, v in attrs.items():
+        setattr(m, k, v)
+    return m
+
+
+# ── Module-level setup: mock deps, import dedup, then restore sys.modules ──
+_SAVED_MODULES_KEYS = set(sys.modules.keys())
+_SAVED_MODULES_VALUES = {
+    k: sys.modules.get(k)
+    for k in [
+        "cv2",
+        "celery",
+        "sqlalchemy",
+        "sqlalchemy.orm",
+        "sqlalchemy.engine",
+        "sqlalchemy.ext",
+        "sqlalchemy.ext.declarative",
+        "worker_app.db",
+        "worker_app.celery_app",
+        "worker_app.core.config",
+        "packages.adapters.sqlalchemy_impl.session",
+        "packages.adapters.sqlalchemy_impl.generated_video_repository",
+        "packages.adapters.sqlalchemy_impl.models",
+        "packages.shared.config",
+        "packages.shared.storage",
+    ]
+}
+
+sys.modules["cv2"] = _mock_module()
+
+_mock_celery = MagicMock()
+_mock_celery.Task = MagicMock
+_mock_celery.Celery = MagicMock
+_mock_celery.__spec__ = None
+sys.modules["celery"] = _mock_celery
+
+_mock_sqla = MagicMock()
+_mock_sqla.__path__ = []
+_mock_sqla.__spec__ = None
+sys.modules["sqlalchemy"] = _mock_sqla
+
+_mock_sqla_orm = MagicMock()
+_mock_sqla_orm.__path__ = []
+_mock_sqla_orm.__spec__ = None
+_mock_sqla_orm.Session = MagicMock
+sys.modules["sqlalchemy.orm"] = _mock_sqla_orm
+sys.modules["sqlalchemy.engine"] = _mock_module()
+sys.modules["sqlalchemy.ext"] = _mock_module()
+sys.modules["sqlalchemy.ext.declarative"] = _mock_module()
+
+sys.modules["worker_app.db"] = _mock_module(SessionLocal=MagicMock())
+sys.modules["worker_app.celery_app"] = _mock_module(celery_app=MagicMock())
+sys.modules["worker_app.core.config"] = _mock_module(get_settings=MagicMock(return_value=MagicMock()))
+
+sys.modules["packages.adapters.sqlalchemy_impl.session"] = _mock_module(
+    Base=MagicMock(),
+    build_engine=MagicMock(),
+    build_session_factory=MagicMock(),
+    ensure_database_exists=MagicMock(),
+    initialize_database=MagicMock(),
+)
+sys.modules["packages.adapters.sqlalchemy_impl.generated_video_repository"] = _mock_module(
+    SQLAlchemyGeneratedVideoRepository=MagicMock
+)
+sys.modules["packages.adapters.sqlalchemy_impl.models"] = _mock_module(
+    VideoFingerprintChunkModel=MagicMock,
+    GeneratedVideoModel=MagicMock,
+)
+sys.modules["packages.shared.config"] = _mock_module(get_shared_settings=MagicMock(return_value=MagicMock()))
+sys.modules["packages.shared.storage"] = _mock_module()
+
+# Save a reference to the dedup module for use in tests (after sys.modules restore)
+import video_processing.dedup as _dedup_mod
+from video_processing.dedup import (  # noqa: E402
+    DUPLICATE_THRESHOLD,
+    HISTOGRAM_WEIGHT,
+    LONG_VIDEO_DURATION_THRESHOLD_SEC,
+    MATCH_RATIO_THRESHOLD,
+    MAX_GAP,
+    MAX_KEYFRAMES,
+    MIN_CONSECUTIVE_MATCHES,
+    MIN_KEYFRAME_INTERVAL_SEC,
+    MIN_KEYFRAMES,
+    PHASH_WEIGHT,
+    SCENE_CHANGE_THRESHOLD,
+    SEGMENT_MATCH_THRESHOLD,
+    DuplicateSegment,
+    FingerprintChunk,
+    VideoDeduplicator,
+    VideoFingerprint,
+    detect_keyframe_timestamps,
+    find_duplicate_segments,
+    hamming_distance,
+)
+
+# ── Restore sys.modules immediately after import ──
+for _key in list(sys.modules.keys()):
+    if _key not in _SAVED_MODULES_KEYS:
+        del sys.modules[_key]
+for _key, _value in _SAVED_MODULES_VALUES.items():
+    if _value is not None:
+        sys.modules[_key] = _value
+    elif _key in sys.modules:
+        del sys.modules[_key]
+del _SAVED_MODULES_KEYS, _SAVED_MODULES_VALUES, _key, _value
+
+
+# ── Helper ──────────────────────────────────────────────────────
+
+
+def _make_chunk(start_ms: int, end_ms: int, phash: str, hist: list[float] | None = None) -> FingerprintChunk:
+    """创建测试用 FingerprintChunk."""
+    return FingerprintChunk(
+        start_time_ms=start_ms,
+        end_time_ms=end_ms,
+        phash_binary=phash,
+        color_histogram=hist or [0.1] * 96,
+        frame_count=1,
+    )
+
+
+# ── TestDuplicateSegment ────────────────────────────────────────
+
+
+class TestDuplicateSegment:
+    """DuplicateSegment 数据类测试."""
+
+    def test_creation(self):
+        """正常创建."""
+        seg = DuplicateSegment(
+            query_start_ms=1000,
+            query_end_ms=5000,
+            target_start_ms=2000,
+            target_end_ms=6000,
+            avg_distance=3.5,
+        )
+        assert seg.query_start_ms == 1000
+        assert seg.avg_distance == 3.5
+
+    def test_fields(self):
+        """所有字段可访问."""
+        seg = DuplicateSegment(0, 1000, 500, 1500, 2.0)
+        assert seg.query_end_ms == 1000
+        assert seg.target_start_ms == 500
+        assert seg.target_end_ms == 1500
+
+
+# ── TestDetectKeyframeTimestamps ────────────────────────────────
+
+
+class TestDetectKeyframeTimestamps:
+    """detect_keyframe_timestamps 关键帧检测测试.
+
+    由于 cv2 在单元测试环境中是 mock，这里只测试边界条件。
+    完整的视频处理测试在集成测试中进行。
+    """
+
+    def test_cannot_open_video_raises(self):
+        """无法打开视频时抛出 RuntimeError."""
+        cv2_mock = _dedup_mod.cv2
+        mock_cap = MagicMock()
+        mock_cap.isOpened.return_value = False
+        cv2_mock.VideoCapture.return_value = mock_cap
+
+        import pytest
+
+        with pytest.raises(RuntimeError, match="Cannot open video"):
+            detect_keyframe_timestamps("/fake/path.mp4")
+
+    def test_zero_duration_returns_empty(self):
+        """视频时长为 0 时返回空列表."""
+        cv2_mock = _dedup_mod.cv2
+        mock_cap = MagicMock()
+        mock_cap.isOpened.return_value = True
+        # cv2.CAP_PROP_FPS etc. are Mock objects; configure get() to return 0 for frame_count
+        mock_cap.get.return_value = 0
+        mock_cap.read.return_value = (False, None)
+        cv2_mock.VideoCapture.return_value = mock_cap
+
+        result = detect_keyframe_timestamps("/fake/zero.mp4")
+        assert result == []
+
+    def test_function_signature(self):
+        """验证函数签名和默认参数."""
+        import inspect
+
+        sig = inspect.signature(detect_keyframe_timestamps)
+        params = sig.parameters
+        assert "video_path" in params
+        assert "min_interval_sec" in params
+        assert "max_frames" in params
+        assert "min_frames" in params
+        # 默认值
+        assert params["min_interval_sec"].default == 1.0
+        assert params["max_frames"].default == 30
+        assert params["min_frames"].default == 5
+
+
+# ── TestFindDuplicateSegments ───────────────────────────────────
+
+
+class TestFindDuplicateSegments:
+    """find_duplicate_segments 滑动窗口时序匹配测试."""
+
+    def test_identical_chunks_full_match(self):
+        """两组完全相同的 chunks → 整段匹配."""
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, "aaaaaaaaaaaaaaaa") for i in range(10)]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, "aaaaaaaaaaaaaaaa") for i in range(10)]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        assert len(segments) >= 1
+        # 应该覆盖大部分范围
+        total_query_range = segments[-1].query_end_ms - segments[0].query_start_ms
+        assert total_query_range > 5000  # 至少覆盖 5 秒
+
+    def test_completely_different_chunks(self):
+        """两组完全不同的 chunks → 空列表."""
+        # 距离都 > 阈值
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, "0000000000000000") for i in range(10)]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, "ffffffffffffffff") for i in range(10)]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        assert segments == []
+
+    def test_partial_overlap(self):
+        """部分重叠 → 只返回重叠段."""
+        # 前 5 帧相同，后 5 帧不同
+        same_hash = "aaaaaaaaaaaaaaaa"
+        diff_hash_a = "0000000000000000"
+        diff_hash_b = "ffffffffffffffff"
+
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, same_hash) for i in range(5)] + [
+            _make_chunk(i * 1000, (i + 1) * 1000, diff_hash_a) for i in range(5, 10)
+        ]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, same_hash) for i in range(5)] + [
+            _make_chunk(i * 1000, (i + 1) * 1000, diff_hash_b) for i in range(5, 10)
+        ]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        # 应该只有前 5 帧的匹配段
+        if segments:
+            assert segments[0].query_end_ms <= 5000
+
+    def test_min_consecutive_not_met(self):
+        """连续 4 帧匹配（< min_consecutive=5）→ 不报重复.
+
+        注意：使用不同的 hash 对，确保后半部分帧距离 > 阈值。
+        """
+        same_hash = "aaaaaaaaaaaaaaaa"
+        # 4 帧匹配，后面 6 帧各自不同（在 query 和 target 中使用不同 hash）
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, same_hash) for i in range(4)] + [
+            _make_chunk(i * 1000, (i + 1) * 1000, "bbbbbbbbbbbbbbbb") for i in range(4, 10)
+        ]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, same_hash) for i in range(4)] + [
+            _make_chunk(i * 1000, (i + 1) * 1000, "cccccccccccccccc") for i in range(4, 10)
+        ]
+
+        # hamming("bbbb...", "cccc...") should be > 8 (SEGMENT_MATCH_THRESHOLD)
+        # b=1011, c=1100 → 4 bits differ per hex digit × 16 digits = 64 bits total? No...
+        # Actually: hamming_distance("bbbbbbbbbbbbbbbb", "cccccccccccccccc")
+        # b=0xb=1011, c=0xc=1100 → XOR=0111=0x7 → 3 bits per digit × 16 = 48
+        # That's > 8 so won't match
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        # 只有 4 帧匹配（< min_consecutive=5），所以不报告
+        assert segments == []
+
+    def test_max_gap_behavior(self):
+        """5 帧匹配 + 1 帧间隙 + 3 帧匹配 → 验证 max_gap 行为.
+
+        关键：间隙帧必须在 query 和 target 中使用不同 hash，使其真正不匹配。
+        """
+        match_hash = "aaaaaaaaaaaaaaaa"
+        gap_hash_a = "bbbbbbbbbbbbbbbb"  # query 端
+        gap_hash_b = "cccccccccccccccc"  # target 端（与 query 端距离 > 8）
+        tail_hash_a = "dddddddddddddddd"
+        tail_hash_b = "eeeeeeeeeeeeeeee"
+
+        # 5 帧匹配, 1 帧间隙, 3 帧匹配, 5 帧不匹配
+        hashes_a = [match_hash] * 5 + [gap_hash_a] + [match_hash] * 3 + [tail_hash_a] * 5
+        hashes_b = [match_hash] * 5 + [gap_hash_b] + [match_hash] * 3 + [tail_hash_b] * 5
+
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, h) for i, h in enumerate(hashes_a)]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, h) for i, h in enumerate(hashes_b)]
+
+        # max_gap=2, 所以 1 帧间隙会被合并
+        segments = find_duplicate_segments(chunks_a, chunks_b, max_gap=2)
+        # 5 match + 1 gap + 3 match = run of 9（间隙被桥接）
+        assert len(segments) == 1
+        # run 覆盖 indices 0-8（5 match + 1 gap + 3 match），但 gap 帧不计入 match
+        # query_start = chunks_a[0].start = 0
+        # query_end = chunks_a[8].end = 9000
+        assert segments[0].query_start_ms == 0
+        assert segments[0].query_end_ms == 9000
+
+    def test_max_gap_exceeded(self):
+        """间隙超过 max_gap → 分成两段."""
+        match_hash = "aaaaaaaaaaaaaaaa"
+        gap_hash_a = "bbbbbbbbbbbbbbbb"
+        gap_hash_b = "cccccccccccccccc"
+        tail_hash_a = "dddddddddddddddd"
+        tail_hash_b = "eeeeeeeeeeeeeeee"
+
+        # 5 帧匹配, 3 帧间隙 (> max_gap=2), 5 帧匹配, 5 帧不匹配
+        hashes_a = [match_hash] * 5 + [gap_hash_a] * 3 + [match_hash] * 5 + [tail_hash_a] * 5
+        hashes_b = [match_hash] * 5 + [gap_hash_b] * 3 + [match_hash] * 5 + [tail_hash_b] * 5
+
+        chunks_a = [_make_chunk(i * 1000, (i + 1) * 1000, h) for i, h in enumerate(hashes_a)]
+        chunks_b = [_make_chunk(i * 1000, (i + 1) * 1000, h) for i, h in enumerate(hashes_b)]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b, max_gap=2)
+        # 3 帧间隙 > max_gap=2 → 分成两段（每段 5 帧匹配）
+        assert len(segments) == 2
+
+    def test_empty_chunks(self):
+        """空 chunks 返回空列表."""
+        assert find_duplicate_segments([], [_make_chunk(0, 1000, "aa")]) == []
+        assert find_duplicate_segments([_make_chunk(0, 1000, "aa")], []) == []
+        assert find_duplicate_segments([], []) == []
+
+    def test_dict_chunks_compatibility(self):
+        """dict 格式的 chunks 也能正常工作."""
+        chunks_a = [
+            {"phash_binary": "aaaaaaaaaaaaaaaa", "start_time_ms": i * 1000, "end_time_ms": (i + 1) * 1000}
+            for i in range(10)
+        ]
+        chunks_b = [
+            {"phash_binary": "aaaaaaaaaaaaaaaa", "start_time_ms": i * 1000, "end_time_ms": (i + 1) * 1000}
+            for i in range(10)
+        ]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        assert len(segments) >= 1
+
+    def test_segment_time_ranges(self):
+        """返回的 segment 时间范围正确.
+
+        每个 query chunk 匹配到 target 中对应的 chunk（相同 hash），
+        确保 target 时间范围正确映射。
+        """
+
+        # 给每个 chunk 唯一的 hash（但保证 query[i] == target[i]）
+        def _unique_hash(i: int) -> str:
+            return format(i, "016x")
+
+        chunks_a = [_make_chunk(i * 2000, (i + 1) * 2000, _unique_hash(i)) for i in range(7)]
+        chunks_b = [_make_chunk(i * 2000, (i + 1) * 2000, _unique_hash(i)) for i in range(7)]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        assert len(segments) >= 1
+        seg = segments[0]
+        assert seg.query_start_ms == 0
+        assert seg.query_end_ms == 14000
+        # target 应该映射到正确的范围
+        assert seg.target_start_ms == 0
+        assert seg.target_end_ms == 14000
+        assert seg.avg_distance == 0.0  # 完全相同
+
+
+# ── TestMedianVsMean ────────────────────────────────────────────
+
+
+class TestMedianVsMean:
+    """中位数 vs 均值：验证中位数抵抗异常值."""
+
+    def test_median_resists_outlier(self):
+        """距离 [3,3,3,3,30]：均值=8.4，中位数=3.
+        中位数 < PHASH_THRESHOLD(10)，均值也 < 10。
+        但更极端的：[3,3,3,3,60]：均值=14.4，中位数=3.
+        """
+        import statistics
+
+        distances = [3, 3, 3, 3, 60]
+        assert statistics.median(distances) == 3
+        assert sum(distances) / len(distances) == 14.4
+        # 中位数 < 10 → 通过阈值
+        assert statistics.median(distances) < 10
+
+
+# ── TestMatchRatioCondition ─────────────────────────────────────
+
+
+class TestMatchRatioCondition:
+    """帧匹配比例条件测试."""
+
+    def test_ratio_below_threshold_skips(self):
+        """10 帧中只有 5 帧距离 < 10 → match_ratio=0.5 < 0.7 → 跳过."""
+        distances = [3, 5, 7, 8, 9, 15, 20, 25, 30, 40]
+        threshold = 10
+        matching = sum(1 for d in distances if d < threshold)
+        ratio = matching / len(distances)
+        assert ratio == 0.5
+        assert ratio < 0.7  # 应该被跳过
+
+    def test_ratio_above_threshold_passes(self):
+        """10 帧中 8 帧距离 < 10 → match_ratio=0.8 >= 0.7 → 通过."""
+        distances = [3, 5, 7, 8, 9, 3, 5, 7, 20, 30]
+        threshold = 10
+        matching = sum(1 for d in distances if d < threshold)
+        ratio = matching / len(distances)
+        assert ratio == 0.8
+        assert ratio >= 0.7  # 应该通过
+
+
+# ── TestBhattacharyyaFusion ─────────────────────────────────────
+
+
+class TestBhattacharyyaFusion:
+    """直方图融合逻辑测试."""
+
+    def test_high_phash_high_hist_is_duplicate(self):
+        """pHash 高相似 + 直方图高相似 → combined_score 高."""
+        phash_similarity = 0.95  # median_distance ≈ 3
+        hist_similarity = 0.90
+        combined = 0.7 * phash_similarity + 0.3 * hist_similarity
+        assert combined > 0.70  # DUPLICATE_THRESHOLD
+
+    def test_high_phash_low_hist_maybe_not(self):
+        """pHash 高相似 + 直方图低相似 → combined_score 取决于权重."""
+        phash_similarity = 0.85  # median_distance ≈ 10
+        hist_similarity = 0.10
+        combined = 0.7 * phash_similarity + 0.3 * hist_similarity
+        # 0.7 * 0.85 + 0.3 * 0.10 = 0.595 + 0.03 = 0.625 < 0.70
+        assert combined < 0.70
+
+    def test_no_histogram_fallback(self):
+        """无直方图数据时 hist_similarity 回退到 0.5."""
+        phash_similarity = 0.90
+        hist_similarity = 0.5  # fallback
+        combined = 0.7 * phash_similarity + 0.3 * hist_similarity
+        # 0.7 * 0.90 + 0.3 * 0.5 = 0.63 + 0.15 = 0.78 > 0.70
+        assert combined > 0.70
+
+
+# ── TestBackwardCompatibility ───────────────────────────────────
+
+
+class TestBackwardCompatibility:
+    """向后兼容测试."""
+
+    def test_no_chunks_no_crash(self):
+        """已有视频无分片数据 → find_duplicate_segments 返回空列表."""
+        # 模拟：fingerprint 有 chunks，但 existing 只有 JSON phashes
+        query_chunks = [_make_chunk(i * 1000, (i + 1) * 1000, "aaaaaaaaaaaaaaaa") for i in range(10)]
+        # 没有 start_time_ms/end_time_ms 的简化 dict
+        target_as_dicts = [{"phash_binary": "aaaaaaaaaaaaaaaa"} for _ in range(10)]
+
+        # find_duplicate_segments 需要 start_time_ms/end_time_ms
+        # 在没有的情况下应该不崩溃（用默认值）
+        # 实际上我们的实现用 _get_start/_get_end 访问，缺 key 会 KeyError
+        # 所以 check_duplicate 传入时会补上默认值
+        target_with_defaults = [
+            {"phash_binary": "aaaaaaaaaaaaaaaa", "start_time_ms": 0, "end_time_ms": 0} for _ in range(10)
+        ]
+        segments = find_duplicate_segments(query_chunks, target_with_defaults)
+        # 不会崩溃
+        assert isinstance(segments, list)
+
+    def test_few_chunks_no_crash(self):
+        """少量 chunk 不崩溃."""
+        chunks_a = [_make_chunk(0, 5000, "aaaaaaaaaaaaaaaa")]
+        chunks_b = [{"phash_binary": "aaaaaaaaaaaaaaaa", "start_time_ms": 0, "end_time_ms": 5000}]
+
+        segments = find_duplicate_segments(chunks_a, chunks_b)
+        # 1 帧 < min_consecutive=5，不会报重复
+        assert segments == []
+
+
+# ── TestConstants ───────────────────────────────────────────────
+
+
+class TestConstants:
+    """常量值验证 — 使用已在模块顶部导入的常量，避免重新 import."""
+
+    def test_segment_match_threshold(self):
+        # 从已导入的 find_duplicate_segments 默认参数间接验证
+        assert SEGMENT_MATCH_THRESHOLD == 8
+
+    def test_min_consecutive_matches(self):
+        assert MIN_CONSECUTIVE_MATCHES == 5
+
+    def test_max_gap(self):
+        assert MAX_GAP == 2
+
+    def test_scene_change_threshold(self):
+        assert SCENE_CHANGE_THRESHOLD == 30
+
+    def test_min_keyframe_interval(self):
+        assert MIN_KEYFRAME_INTERVAL_SEC == 1.0
+
+    def test_max_keyframes(self):
+        assert MAX_KEYFRAMES == 30
+
+    def test_min_keyframes(self):
+        assert MIN_KEYFRAMES == 5
+
+    def test_long_video_threshold(self):
+        assert LONG_VIDEO_DURATION_THRESHOLD_SEC == 180
+
+    def test_duplicate_threshold(self):
+        assert DUPLICATE_THRESHOLD == 0.70
+
+    def test_phash_weight(self):
+        assert PHASH_WEIGHT == 0.7
+
+    def test_histogram_weight(self):
+        assert HISTOGRAM_WEIGHT == 0.3
+
+    def test_match_ratio_threshold(self):
+        assert MATCH_RATIO_THRESHOLD == 0.7

tests/unit/test_duplicate_rate.py

@@ -122,7 +122,7 @@ class TestComputeDuplicateRate:
             rate = deduplicator.compute_duplicate_rate(fingerprint, "proj1", "vid1", session)
 
         # hamming distance = 2, similarity = (1 - 2/64) * 100 = 96.875
-        assert rate == pytest.approx(96.88, abs=0.1)
+        assert rate == pytest.approx(82.81, abs=0.1)  # 新算法: 0.7*(1-2/64)*100 + 0.3*50
 
     def test_excludes_self_video(self):
         from video_processing.dedup import VideoDeduplicator
@@ -186,7 +186,7 @@ class TestComputeDuplicateRate:
             rate = deduplicator.compute_duplicate_rate(fingerprint, "proj1", "vid1", session)
 
         # max similarity: e2 distance=1, (1-1/64)*100 = 98.4375
-        assert rate == pytest.approx(98.44, abs=0.1)
+        assert rate == pytest.approx(83.91, abs=0.1)  # 新算法: 0.7*(1-1/64)*100 + 0.3*50
 
     def test_user_id_scope_cross_project(self):
         """传 user_id 时应跨项目查询，而非仅当前项目."""

tests/unit/test_fingerprint_chunks.py

@@ -110,7 +110,6 @@ from video_processing.dedup import (  # noqa: E402
     FingerprintChunk,
     VideoFingerprint,
     _save_fingerprint_chunks,
-    compute_chunk_interval,
 )
 
 # ── Restore sys.modules immediately after import ──
@@ -125,36 +124,6 @@ for _key, _value in _SAVED_MODULES_VALUES.items():
 del _SAVED_MODULES_KEYS, _SAVED_MODULES_VALUES, _key, _value
 
 
-class TestChunkInterval:
-    """测试分片间隔策略。"""
-
-    def test_short_video_interval(self):
-        """短视频（≤60秒）每 2 秒一个分片。"""
-        assert compute_chunk_interval(0) == 2
-        assert compute_chunk_interval(30) == 2
-        assert compute_chunk_interval(60) == 2
-
-    def test_long_video_interval(self):
-        """长视频（>60秒）每 5 秒一个分片。"""
-        assert compute_chunk_interval(61) == 5
-        assert compute_chunk_interval(120) == 5
-        assert compute_chunk_interval(300) == 5
-
-    def test_chunk_count_60s_video(self):
-        """60秒视频 → 30 片（60/2=30）。"""
-        duration = 60
-        interval = compute_chunk_interval(duration)
-        expected_chunks = int(duration / interval)
-        assert expected_chunks == 30
-
-    def test_chunk_count_120s_video(self):
-        """120秒视频 → 24 片（120/5=24）。"""
-        duration = 120
-        interval = compute_chunk_interval(duration)
-        expected_chunks = int(duration / interval)
-        assert expected_chunks == 24
-
-
 class TestVideoFingerprintToChunkModels:
     """测试 VideoFingerprint.to_chunk_models() 输出。"""