无人机手势识别

src/Drone_gesture_operation/main.py

@@ -10,6 +10,79 @@ class StableFPSHandRecognizer:
     def __init__(self, target_fps=30):
         # 1. 帧率锁定参数
         self.target_fps = target_fps
+        self.frame_interval = 1.0 / target_fps
+        self.last_frame_time = time.time()
+
+        # 2. 优化后的肤色检测阈值
+        self.skin_lower = np.array([0, 20, 70], np.uint8)  # 放宽下界
+        self.skin_upper = np.array([20, 255, 255], np.uint8)  # 调整上界
+        self.kernel = np.ones((5, 5), np.uint8)  # 更大的核去噪
+
+        # 3. 优化后的手指检测参数（降低阈值，提高识别率）
+        self.defect_depth_threshold = 10  # 降低深度阈值
+        self.min_defect_distance = 5  # 降低距离阈值
+        self.min_contour_area = 500  # 降低最小轮廓面积
+
+        # 4. 手势缓存&帧缓存
+        self.gesture_buffer = []
+        self.stable_gesture = "None"
+        self.frame_queue = []
+        self.queue_lock = threading.Lock()
+
+        # 5. 识别区域参数（仅显示边框）
+        self.recognition_area = None
+        self.area_color = (0, 255, 0)  # 边框颜色
+
+    def _init_recognition_area(self, frame_shape):
+        """初始化识别区域（调大尺寸，右侧更大范围）"""
+        h, w = frame_shape[:2]
+        x1 = int(w * 1.5 / 3)  # 左边界左移（从2/3改为1.5/3），扩大宽度
+        y1 = int(h * 0.05)  # 上边界上移（从0.1改为0.05），扩大高度
+        x2 = w - 10  # 右边界右移（从-20改为-10），减少右侧边距
+        y2 = int(h * 0.95)  # 下边界下移（从0.9改为0.95），减少底部边距
+        self.recognition_area = (x1, y1, x2, y2)
+
+    def _draw_recognition_area(self, frame):
+        """绘制识别区域（仅显示边框，无背景色）"""
+        if self.recognition_area is None:
+            self._init_recognition_area(frame.shape)
+        x1, y1, x2, y2 = self.recognition_area
+
+        # 仅绘制边框（移除半透明背景）
+        cv.rectangle(frame, (x1, y1), (x2, y2), self.area_color, 2)
+
+        # 添加区域提示文字（在边框上方）
+        cv.putText(frame, "Recognition Area", (x1 + 10, y1 - 10),
+                   cv.FONT_HERSHEY_SIMPLEX, 0.6, self.area_color, 2)
+        return frame
+
+    def _get_roi(self, frame):
+        """获取识别区域的ROI（确保坐标有效）"""
+        if self.recognition_area is None:
+            self._init_recognition_area(frame.shape)
+        x1, y1, x2, y2 = self.recognition_area
+
+        # 边界保护
+        x1 = max(0, x1)
+        y1 = max(0, y1)
+        x2 = min(frame.shape[1], x2)
+        y2 = min(frame.shape[0], y2)
+
+        return frame[y1:y2, x1:x2], (x1, y1)
+
+    def count_fingers(self, cnt):
+        """优化后的手指计数逻辑（更鲁棒）"""
+        try:
+            # 计算凸包（带坐标）和凸包缺陷
+            hull = cv.convexHull(cnt)
+            hull_indices = cv.convexHull(cnt, returnPoints=False)
+            defects = cv.convexityDefects(cnt, hull_indices)
+
+            if defects is None or len(defects) == 0:
+                return 0
+
+            finger_count = 0
+            # 遍历缺陷点
         self.frame_interval = 1.0 / target_fps  # 每帧间隔时间（秒）
         self.last_frame_time = time.time()
 
@@ -51,6 +124,27 @@ def count_fingers(self, cnt, frame_small):
                 end = tuple(cnt[e][0])
                 far = tuple(cnt[f][0])
 
+                # 计算缺陷深度（实际像素值）
+                depth = d / 256.0
+
+                # 计算角度（过滤误判的缺陷）
+                a = np.linalg.norm(np.array(end) - np.array(start))
+                b = np.linalg.norm(np.array(far) - np.array(start))
+                c = np.linalg.norm(np.array(end) - np.array(far))
+                angle = np.arccos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c)) * 180 / np.pi
+
+                # 有效缺陷：深度足够 + 角度小于90度
+                if depth > self.defect_depth_threshold and angle < 90:
+                    finger_count += 1
+
+            # 缺陷数+1=手指数量
+            return min(finger_count + 1, 5)
+        except Exception as e:
+            print(f"手指计数错误: {e}")
+            return 0
+
+    def capture_frames(self, cap):
+        """帧采集线程（稳定）"""
                 # 计算缺陷深度（转换为实际像素值）
                 depth = d / 256.0
 
@@ -74,6 +168,63 @@ def capture_frames(self, cap):
             if not ret:
                 break
             with self.queue_lock:
+                self.frame_queue = [frame]  # 只保留最新帧
+            time.sleep(self.frame_interval * 0.5)
+
+    def process_frame(self, frame):
+        """优化后的帧处理逻辑"""
+        # 镜像翻转
+        frame = cv.flip(frame, 1)
+        # 绘制识别区域（仅边框）
+        frame = self._draw_recognition_area(frame)
+        # 获取ROI
+        roi, (roi_x, roi_y) = self._get_roi(frame)
+        current_gesture = "None"
+
+        if roi.size > 0:  # 确保ROI有效
+            # 预处理：缩小+转HSV+肤色掩码
+            roi_small = cv.resize(roi, (320, 240))  # 适度放大ROI
+            hsv = cv.cvtColor(roi_small, cv.COLOR_BGR2HSV)
+            mask = cv.inRange(hsv, self.skin_lower, self.skin_upper)
+
+            # 形态学操作（去噪+填充）
+            mask = cv.morphologyEx(mask, cv.MORPH_OPEN, self.kernel)
+            mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, self.kernel)
+            mask = cv.dilate(mask, self.kernel, iterations=2)
+
+            # 查找轮廓
+            contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
+            if contours:
+                # 取最大轮廓
+                cnt = max(contours, key=cv.contourArea)
+                area = cv.contourArea(cnt)
+
+                if area > self.min_contour_area:
+                    # 计算密实度
+                    hull = cv.convexHull(cnt)
+                    hull_area = cv.contourArea(hull)
+                    solidity = area / hull_area if hull_area > 0 else 0
+
+                    # 手指计数
+                    finger_count = self.count_fingers(cnt)
+
+                    # 可视化调试
+                    cnt_scaled = cnt * (roi.shape[1] / roi_small.shape[1], roi.shape[0] / roi_small.shape[0])
+                    cnt_scaled = cnt_scaled.astype(np.int32)
+                    cnt_scaled[:, :, 0] += roi_x
+                    cnt_scaled[:, :, 1] += roi_y
+                    cv.drawContours(frame, [cnt_scaled], -1, (255, 0, 0), 2)
+
+                    # 手势判断逻辑
+                    if solidity > 0.8:  # 握拳
+                        current_gesture = "stop"
+                    elif finger_count == 2:  # 食指+中指
+                        current_gesture = "front"
+                    elif finger_count >= 4:  # 手掌张开
+                        current_gesture = "back"
+                    # 其他情况（1/3指）归为None
+
+        # 手势缓存稳定
                 # 只保留最新1帧，避免堆积
                 self.frame_queue = [frame]
             # 采集线程限速，匹配目标帧率
@@ -272,6 +423,7 @@ def main():
         if len(set(self.gesture_buffer)) == 1:
             self.stable_gesture = self.gesture_buffer[0]
 
+        # 绘制UI
         # 5. 绘制极简UI（显示修改后的手势文本）
         # 5. 绘制极简UI（仅保留手势和FPS显示）
         # 5. 绘制极简UI（仅保留手势和FPS显示，移除手指数量）
@@ -281,11 +433,43 @@ def main():
         cv.putText(frame, f"FPS: {self.target_fps}", (10, 80),
                    cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2)
 
+        # 拉伸
         # 拉伸显示（保持清晰）
         frame_show = cv.resize(frame, (640, 480))
         return frame_show
 
     def run(self):
+        """主运行逻辑"""
+        # 摄像头初始化
+        cap = cv.VideoCapture(0)
+        cap.set(cv.CAP_PROP_FRAME_WIDTH, 640)  # 提高摄像头分辨率
+        cap.set(cv.CAP_PROP_FRAME_HEIGHT, 480)
+        cap.set(cv.CAP_PROP_FOURCC, cv.VideoWriter_fourcc(*'MJPG'))
+        cap.set(cv.CAP_PROP_BUFFERSIZE, 1)
+        cap.set(cv.CAP_PROP_FPS, self.target_fps)
+
+        # 启动采集线程
+        capture_thread = threading.Thread(target=self.capture_frames, args=(cap,), daemon=True)
+        capture_thread.start()
+
+        # 提示信息
+        print("=" * 50)
+        print(f"✅ 帧率锁定 {self.target_fps} 帧 | ESC退出")
+        print("💡 调试提示：")
+        print("   1. 把手放在右侧绿色边框的识别区域内（已扩大范围）")
+        print("   2. 握拳 → stop | 食指+中指 → front | 手掌张开 → back")
+        print("   3. 蓝色轮廓表示检测到的手部区域")
+        print("=" * 50)
+
+        # 主循环
+        while cap.isOpened():
+            # 帧率控制
+            current_time = time.time()
+            elapsed = current_time - self.last_frame_time
+            if elapsed < self.frame_interval:
+                time.sleep(self.frame_interval - elapsed)
+
+            # 读取帧
         """主运行逻辑，帧率锁死"""
         # 1. 摄像头初始化（硬件级优化）
         cap = cv.VideoCapture(0)
@@ -325,6 +509,9 @@ def run(self):
 
             # 处理并显示
             frame_show = self.process_frame(frame)
+            cv.imshow("Hand Gesture Recognition", frame_show)
+
+            # 更新时间戳
             cv.imshow("Stable FPS Gesture", frame_show)
 
             # 更新时间戳，确保下一帧同步
@@ -340,6 +527,9 @@ def run(self):
 
 
 if __name__ == '__main__':
+    # 可降低帧率（如15）提高稳定性
+    recognizer = StableFPSHandRecognizer(target_fps=20)
+    recognizer.run()
     # 实例化并运行，锁定30帧（可改20/15帧，更低更稳）
     recognizer = StableFPSHandRecognizer(target_fps=30)
     recognizer.run()