一、前言
区别于之前的直接驱动相机，这里改为读取图像话题进行处理，原因是如果opencv驱动相机后只能单一使用，就限制了其他识别功能（除非将原始图像发布出来），所以这里改成可以读取任意相机图像话题的方法，增加实用性。
二、原理
利用image_transport::ImageTransport
void imageCallback(const sensor_msgs::ImageConstPtr& msg)
{// ...
}ros::NodeHandle nh;
image_transport::ImageTransport it(nh);
image_transport::Subscriber sub = it.subscribe(cam_topic, 1, imageCallback);
在回调函数中转化ros图像信息为opencv格式
try{cv_ptr = cv_bridge::toCvCopy(msg, sensor_msgs::image_encodings::BGR8);frame = cv_ptr->image;}catch (cv_bridge::Exception& e){ROS_ERROR("cv_bridge exception: %s", e.what());return;}
这里frame为Mat形式，相当于opencv的帧
之后就可以对frame进行图像处理了
三、使用
launch文件的使用
<?xml version="1.0"?>
<launch><node pkg="color_detect" name="color_topic" type="color_topic" output="screen"/><param name="image_view" type="bool" value="true"/><param name="cam_topic" type="string" value="/usb_cam/image_raw"/><param name="H" type="int" value="100"/><param name="S" type="int" value="100"/><param name="V" type="int" value="100"/></launch>通过更改相机话题名称来读取相机图像话题 。
将图像处理的代码放进回调函数，而不是在while中，防止图像处理有延迟和重复。
void imageCallback(const sensor_msgs::ImageConstPtr& msg)
{try{cv_ptr = cv_bridge::toCvCopy(msg, sensor_msgs::image_encodings::BGR8);frame = cv_ptr->image;}catch (cv_bridge::Exception& e){ROS_ERROR("cv_bridge exception: %s", e.what());return;}Scalar lower_red(H - 30, S - 30, V - 30);Scalar upper_red(H + 30, S + 30, V + 30); // 定义红色的HSV范围if(!frame.empty()) //画面是否正常{ /*对图片二次处理*/cvtColor(frame, imghsv, COLOR_BGR2HSV);// 将图像转换为HSV颜色空间split(imghsv, hsvSplit);equalizeHist(hsvSplit[2], hsvSplit[2]);merge(hsvSplit, imghsv);inRange(imghsv, lower_red, upper_red, mask);//二值化红色部分Mat kernel = getStructuringElement(MORPH_RECT, Size(5, 5));morphologyEx(mask, mask, MORPH_OPEN, kernel);//开运算morphologyEx(mask, mask, MORPH_CLOSE, kernel);//闭运算GaussianBlur(mask, mask, Size(5, 5), 0);//高斯滤波Canny(mask, mask, 150, 100);//canny算子边缘检测vector<vector<Point> > contours;vector<Vec4i> hierarchy;findContours(mask,contours,hierarchy,RETR_EXTERNAL,CHAIN_APPROX_SIMPLE,Point()); //ROS_INFO("个数为%d",int(contours.size()));std::vector<std::vector<cv::Point> >::const_iterator itc = contours.begin();std::vector<std::vector<cv::Point> >::const_iterator max_c = contours.begin();if( (!contours.empty() && !hierarchy.empty())){//寻找最大面积的轮廓while (itc != contours.end()){if( cv::contourArea(*itc) > cv::contourArea(*max_c)) {max_c = itc;}itc++;}Rect boundRect = boundingRect(*max_c);circle(frame, Point(boundRect.x + boundRect.width/2, boundRect.y + boundRect.height/2), 5, Scalar(0,0,255), -1);rectangle(frame, Point(boundRect.x, boundRect.y), Point(boundRect.x + boundRect.width, boundRect.y + boundRect.height), Scalar( 0, 0, 255), 2);//ROS_INFO("x:%d,y:%d",boundRect.x+ boundRect.width/2, boundRect.y + boundRect.height/2);detect_msg.Class = "red";detect_msg.xmin = boundRect.x;detect_msg.xmax=boundRect.x + boundRect.width;detect_msg.ymin=boundRect.y;detect_msg.ymax= boundRect.y + boundRect.height;}else{} if(image_view){namedWindow("video");startWindowThread();imshow("video", frame);setMouseCallback("video", on_mouse, 0);// int c = waitKey(1);// if (c == 27) {//key:esc// ros::shutdown();// }}//opencv转rostry{ros_msg = cv_bridge::CvImage(std_msgs::Header(), "bgr8", frame).toImageMsg(); }catch (cv_bridge::Exception& e){ROS_ERROR("cv_bridge exception:%s",e.what());}image_pub.publish(ros_msg); boundingbox_pub.publish(detect_msg);} }
所有效果和之前的功能包相同
Opencv+ROS实现特定物品识别_opencv 训练模型 特定物体识别-CSDN博客文章浏览阅读913次，点赞5次，收藏4次。ros实现特定物品识别，基于opencv的hsv值颜色识别算法为原理。_opencv 训练模型 特定物体识别https://blog.csdn.net/2301_76165902/article/details/144225683具体代码可以去这篇博客查看。
感谢观看。
参考：ROS image_transport使用笔记_image callback-CSDN博客
