Python 实现 68 点人脸关键点检测实战

面部结构可视化增强

单纯画点可能不够直观，我们可以根据关键点定义绘制面部区域的轮廓，这样能更清晰地看到五官结构。

def draw_facial_structures(image, landmarks):
    """绘制面部结构轮廓"""
    regions = {
        "jaw": (0, 17),
        "right_eyebrow": (17, 22),
        "left_eyebrow": (22, 27),
        "nose": (27, 36),
        "right_eye": (36, 42),
        "left_eye": (42, 48),
        "mouth_outer": (48, 60),
        "mouth_inner": (60, 68)
    }

    for region, (start, end) in regions.items():
        pts = landmarks[start:end]
        if region in ["right_eye", "left_eye", "mouth_outer", "mouth_inner"]:
            hull = cv2.convexHull(pts)
            cv2.drawContours(image, [hull], -1, (0, 255, 0), 1)
        else:
            for i in range(1, len(pts)):
                cv2.line(image, tuple(pts[i-1]), tuple(pts[i]), (0, 255, 0), 1)

实时视频流处理

静态图片检测完成后，实际应用中往往需要处理视频流。这里需要注意循环中的帧读取逻辑以及退出机制。

def real_time_landmark_detection():
    """实时视频流处理"""
    detector = dlib.get_frontal_face_detector()
    predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
    cap = cv2.VideoCapture(0)

    while True:
        ret, frame = cap.read()
        if not ret:
            break

        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        faces = detector(gray, 0)

        for face in faces:
            shape = predictor(gray, face)
            landmarks = np.array([[p.x, p.y] for p in shape.parts()])

            # 绘制关键点和轮廓
            for point in landmarks:
                cv2.circle(frame, tuple(point), 2, (0, 255, 0), -1)
            draw_facial_structures(frame, landmarks)

        cv2.imshow('Real-time Face Landmarks', frame)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break

    cap.release()
    cv2.destroyAllWindows()

性能优化建议

在实际部署中，直接运行上述代码可能会遇到性能瓶颈。为了提升效果，建议采用以下策略：

• 图像预处理：调整过大图像尺寸，避免内存溢出。
• 智能检测：交替使用全检测和跟踪算法，减少重复计算。
• 区域聚焦：基于历史位置缩小检测范围。
• 并行处理：分离图像采集与分析线程。

下面是一个经过优化的检测函数示例，加入了图像缩放和直方图均衡化处理：

def efficient_detection(image, detector, predictor):
    """高效检测实现"""
    height, width = image.shape[:2]
    
    # 图像尺寸优化
    if width > 800:
        scale = 800 / width
        image = cv2.resize(image, (800, int(height * scale)))

    # 图像增强处理
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    gray = cv2.equalizeHist(gray)

    # 执行检测
    faces = detector(gray, 0)
    results = []
    for face in faces:
        shape = predictor(gray, face)
        landmarks = np.array([[p.x, p.y] for p in shape.parts()])
        results.append(landmarks)

    return results, image