from ultralytics import YOLO import cv2 import numpy as np import os # ------------------ 配置 ------------------ model_path = 'ultralytics_yolo11-main/runs/train/exp4/weights/best.pt' #img_folder = '/home/hx/yolo/ultralytics_yolo11-main/dataset1/test' img_folder = '/home/hx/yolo/test_image' output_mask_dir = 'output_masks1' os.makedirs(output_mask_dir, exist_ok=True) SUPPORTED_FORMATS = ('.jpg', '.jpeg', '.png', '.bmp', '.tif', '.tiff') # ------------------ 加载模型 ------------------ model = YOLO(model_path) model.to('cuda') # 使用 GPU（如有） def get_orientation_vector(contour): """ 使用 cv2.fitLine 计算轮廓的主方向（单位向量） 返回：主方向单位向量 (2,) """ if len(contour) < 5: return np.array([1.0, 0.0]) # 默认方向：沿 x 轴 [vx, vy, _, _] = cv2.fitLine(contour, cv2.DIST_L2, 0, 0.01, 0.01) direction = np.array([vx[0], vy[0]]) # 主轴方向 norm = np.linalg.norm(direction) return direction / norm if norm > 1e-8 else direction def get_contour_center(contour): """计算轮廓质心""" M = cv2.moments(contour) if M["m00"] == 0: return np.array([0, 0]) return np.array([int(M["m10"] / M["m00"]), int(M["m01"] / M["m00"])]) def calculate_jaw_opening_angle(jaw1, jaw2): """ 计算夹具开合角度，并返回修正后的方向向量 返回: (angle, dir1_final, dir2_final) """ center1 = get_contour_center(jaw1['contour']) center2 = get_contour_center(jaw2['contour']) fixture_center = np.array([(center1[0] + center2[0]) / 2.0, (center1[1] + center2[1]) / 2.0]) # ✅ 使用 fitLine 获取主方向 dir1_orig = get_orientation_vector(jaw1['contour']) dir2_orig = get_orientation_vector(jaw2['contour']) def correct_and_compute(d1, d2): """校正方向并计算夹角""" # 校正 jaw1 方向：应指向 fixture_center to_center1 = fixture_center - center1 if np.linalg.norm(to_center1) > 1e-6: to_center1 = to_center1 / np.linalg.norm(to_center1) if np.dot(d1, to_center1) < 0: d1 = -d1 # 反向 # 校正 jaw2 方向 to_center2 = fixture_center - center2 if np.linalg.norm(to_center2) > 1e-6: to_center2 = to_center2 / np.linalg.norm(to_center2) if np.dot(d2, to_center2) < 0: d2 = -d2 # 计算夹角 cos_angle = np.clip(np.dot(d1, d2), -1.0, 1.0) angle = np.degrees(np.arccos(cos_angle)) return angle, d1, d2 # 尝试原始方向 angle_raw, dir1_raw, dir2_raw = correct_and_compute(dir1_orig, dir2_orig) if angle_raw <= 170.0: return angle_raw, dir1_raw, dir2_raw print(f"⚠️ 初始角度过大: {angle_raw:.2f}°，尝试翻转 jaw2 方向...") angle_corrected, dir1_corr, dir2_corr = correct_and_compute(dir1_orig, -dir2_orig) print(f"🔄 方向修正后: {angle_corrected:.2f}°") # 数值兜底：若仍过大，取补角 if angle_corrected > 170.0: final_angle = 180.0 - angle_corrected print(f"🔧 数值修正: {angle_corrected:.2f}° → {final_angle:.2f}°") return final_angle, dir1_corr, dir2_corr return angle_corrected, dir1_corr, dir2_corr def process_image(img_path, output_dir): img = cv2.imread(img_path) if img is None: print(f"❌ 无法读取图像: {img_path}") return h, w = img.shape[:2] filename = os.path.basename(img_path) name_only = os.path.splitext(filename)[0] print(f"\n🔄 正在处理: {filename}") # 创建单通道掩码 composite_mask = np.zeros((h, w), dtype=np.uint8) results = model(img_path, imgsz=1280, conf=0.5) rotated_rects = [] for r in results: if r.masks is not None: masks = r.masks.data.cpu().numpy() boxes = r.boxes.xyxy.cpu().numpy() for i, mask in enumerate(masks): x1, y1, x2, y2 = map(int, boxes[i]) x1, y1 = max(0, x1), max(0, y1) x2, y2 = min(w, x2), min(h, y2) obj_mask = np.zeros((h, w), dtype=np.uint8) mask_resized = cv2.resize(mask, (w, h)) obj_mask[y1:y2, x1:x2] = (mask_resized[y1:y2, x1:x2] * 255).astype(np.uint8) contours, _ = cv2.findContours(obj_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE) if len(contours) == 0: continue largest_contour = max(contours, key=cv2.contourArea) area = cv2.contourArea(largest_contour) if area < 100: continue rect = cv2.minAreaRect(largest_contour) rotated_rects.append({ 'rect': rect, 'contour': largest_contour, 'area': area }) composite_mask = np.maximum(composite_mask, obj_mask) # 创建三通道可视化掩码 vis_mask = np.stack([composite_mask] * 3, axis=-1) vis_mask[composite_mask > 0] = [255, 255, 255] # 白色前景 if len(rotated_rects) < 2: print(f"⚠️ 检测到的对象少于2个（共{len(rotated_rects)}个）: {filename}") mask_save_path = os.path.join(output_dir, f'mask_{name_only}.png') cv2.imwrite(mask_save_path, composite_mask) print(f"✅ 掩码已保存（无足够夹具）: {mask_save_path}") return # 按面积排序，取前两个 rotated_rects.sort(key=lambda x: x['area'], reverse=True) jaw1, jaw2 = rotated_rects[0], rotated_rects[1] # 计算角度和方向 opening_angle, dir1_final, dir2_final = calculate_jaw_opening_angle(jaw1, jaw2) print(f"✅ 最终夹具开合角度: {opening_angle:.2f}°") # ------------------ 可视化 ------------------ center1 = get_contour_center(jaw1['contour']) center2 = get_contour_center(jaw2['contour']) fixture_center = ((center1[0] + center2[0]) // 2, (center1[1] + center2[1]) // 2) # 绘制最小外接矩形 box1 = cv2.boxPoints(jaw1['rect']) box1 = np.int32(box1) cv2.drawContours(vis_mask, [box1], 0, (0, 0, 255), 2) # jaw1: 红色 box2 = cv2.boxPoints(jaw2['rect']) box2 = np.int32(box2) cv2.drawContours(vis_mask, [box2], 0, (255, 0, 0), 2) # jaw2: 蓝色 # 绘制主方向箭头（绿色） scale = 60 end1 = center1 + scale * dir1_final end2 = center2 + scale * dir2_final cv2.arrowedLine(vis_mask, tuple(center1), tuple(end1.astype(int)), (0, 255, 0), 2, tipLength=0.3) cv2.arrowedLine(vis_mask, tuple(center2), tuple(end2.astype(int)), (0, 255, 0), 2, tipLength=0.3) # 标注夹具中心（青色） cv2.circle(vis_mask, fixture_center, 5, (255, 255, 0), -1) cv2.putText(vis_mask, "Center", (fixture_center[0] + 10, fixture_center[1]), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 0), 1) # 标注角度 cv2.putText(vis_mask, f"Angle: {opening_angle:.2f}°", (20, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2) # 保存结果 vis_save_path = os.path.join(output_dir, f'mask_with_angle_{name_only}.png') cv2.imwrite(vis_save_path, vis_mask) print(f"✅ 带角度标注的掩码图已保存: {vis_save_path}") # ------------------ 主程序 ------------------ if __name__ == '__main__': if not os.path.isdir(img_folder): print(f"❌ 图像文件夹不存在: {img_folder}") exit() image_files = [f for f in os.listdir(img_folder) if f.lower().endswith(SUPPORTED_FORMATS)] if len(image_files) == 0: print(f"⚠️ 未找到支持的图像文件") exit() print(f"✅ 发现 {len(image_files)} 张图像，开始处理...") for image_file in image_files: image_path = os.path.join(img_folder, image_file) process_image(image_path, output_mask_dir) print(f"\n🎉 所有图像处理完成！结果保存在: {output_mask_dir}")