需求：小车通过摄像头识别到stop标志，使小车停止
实现框架：
#!/usr/bin/env python3 import rclpy from rclpy.node import Node from sensor_msgs.msg import Image from std_msgs.msg import Bool from cv_bridge import CvBridge from ultralytics import YOLO class YoloStopDetector(Node): def __init__(self): super().__init__('yolo_stop_detector') self.bridge = CvBridge()#用于ros图片和cv图片转换 self.image_sub = self.create_subscription(Image,'/ascamera/camera_publisher/rgb0/image',self.image_callback,10)#图片订阅 self.result_image_pub = self.create_publisher(Image,'/yolo_result_image',10)#检测图片发布 self.stop_pub = self.create_publisher(Bool,'/stop_detected',10)#停止消息发布 self.model = YOLO('/home/ubuntu/ros2_ws/yolo_models/yolov8n.pt')#加载推理模型 self.cv_image_count = 0 self.get_logger().info("YOLO停止标志检测器已启动") self.last_stop = False def image_callback(self,msg): cv_image = self.bridge.imgmsg_to_cv2(msg,'bgr8') self.cv_image_count += 1 # 每5帧检测一次 if self.cv_image_count % 5 != 0: return result = self.model.predict(cv_image,imgsz=416,conf=0.4,verbose=False)[0] # 可用的属性和方法 # result.boxes # 检测框信息 # result.boxes.xyxy # 框的坐标 [x1, y1, x2, y2] # result.boxes.cls # 类别 ID（整数） # result.boxes.conf # 置信度（0~1） # result.boxes.data # 全部数据 [x1, y1, x2, y2, conf, cls] # result.names # 类别名字典 {0: 'person', 11: 'stop sign', ...} # result.plot() # 画检测结果的图片 annotated_image = result.plot() self.publish_result_image(annotated_image)#发布检测画面 stop=False for cls, conf, box in zip(result.boxes.cls,result.boxes.conf,result.boxes.xyxy): if int(cls.item()) == 11 and conf.item()>0.4: # item() 把 PyTorch 张量（tensor）转换成 Python 数字 x1,y1,x2,y2 = map(int,box.tolist()) area=(x2-x1)*(y2-y1) h,w=cv_image.shape[:2] if area > 0.02*w*h: #检测框大于图片一定面积生效 stop=True break msg=Bool() msg.data=stop self.stop_pub.publish(msg) if stop and not self.last_stop: self.get_logger().info("YOLO检测到停止标志") self.last_stop = stop def publish_result_image(self, image): result_msg = self.bridge.cv2_to_imgmsg(image,encoding='bgr8') self.result_image_pub.publish(result_msg) def main(): rclpy.init() node=YoloStopDetector() rclpy.spin(node) node.destroy_node() rclpy.shutdown() if __name__=="__main__": main()实现过程：新增一个目标检测节点，加载模型并进行推理，根据对图片处理结果，比如识别类型，置信度，识别目标占图面积决定标志位状态，发布bool类型的消息，c++巡线节点根据收到的消息更改标志位
最终实现效果：
巡线小车实测2
巡线小车实测3
巡线小车实测4
巡线小车实测5
