这篇文章主要介绍了C++基于ros怎么将文件夹中的图像转换为bag包的相关知识,内容详细易懂,操作简单快捷,具有一定借鉴价值,相信大家阅读完这篇C++基于ros怎么将文件夹中的图像转换为bag包文章都会有所收获,下面我们一起来看看吧。
一、前期工作创建工作空间
二、创建工作包
创建完成后,文件夹的格式为:
三、准备编译文件和代码
3.1 更换编译文件中的内容
将上图中的,CMakeLists.txt文件中的内容,替换为下面的内容
cmake_minimum_required(VERSION 3.0.2)project(create_bag) ## Compile as C++11, supported in ROS Kinetic and newer# add_compile_options(-std=c++11) ## Find catkin macros and libraries## if COMPONENTS list like find_package(catkin REQUIRED COMPONENTS xyz)## is used, also find other catkin packages# 寻找OpenCV库find_package( OpenCV REQUIRED )# 添加头文件include_directories( ${OpenCV_INCLUDE_DIRS} ) find_package(catkin REQUIRED COMPONENTS cv_bridge rosbag roscpp rospy std_msgs) catkin_package(# INCLUDE_DIRS include# LIBRARIES imgtobag# CATKIN_DEPENDS cv_bridge rosbag roscpp rospy std_msgs# DEPENDS system_lib) ############# Build ############# ## Specify additional locations of header files## Your package locations should be listed before other locationsinclude_directories(# include ${catkin_INCLUDE_DIRS} ${catkin_INCLUDE_DIRS} ${rosbag_INCLUDE_DIRS} ${OpenCV_INCLUDE_DIRS}) add_executable(node src/torosbag.cpp)target_link_libraries(node ${catkin_LIBRARIES} ${PCL_LIBRARIES} ${rosbag_LIBRARIES} ${OpenCV_LIBS})
3.2 准备主程序
leedarson@leedarson-desktop:~/catkin_ws/src/create_bag/src$ touch torosbag.cpp
创建一个cpp的文件夹,cpp文件中的内容为:
#include <string>#include <ros/console.h>#include <rosbag/bag.h>#include <cv_bridge/cv_bridge.h>#include <iostream>#include <vector>#include <sys/types.h>#include <dirent.h>#include <unistd.h>#include <opencv2/core/core.hpp>#include <opencv2/highgui/highgui.hpp>using namespace std;using namespace cv; void GetFileNames(string path,vector<string>& filenames, string con);void GetFileNamesByGlob(cv::String path,vector<cv::String>& filenames, string con);bool read_images(string path, vector<string> &image_files);int main(int argc, char **argv){ //输入文件和输出文件路径 string base_dir = "/home/leedarson/catkin_ws/src/create_bag/data/"; string img_dir = base_dir + "img/"; std::cout<<"image path is"<<img_dir<<std::endl; string output_bag=base_dir +"Human2.bag"; string img_format = ".jpg";//格式 vector<string> img_names; //GetFileNames(img_dir, img_names,".jpg"); read_images(img_dir, img_names); cout<<"图片读取完成"<<endl; cout<<"----"<<endl; ros::Time::init(); rosbag::Bag bag; bag.open(output_bag, rosbag::bagmode::Write); int seq = 0; vector<string>::iterator it; for(it = img_names.begin(); it != img_names.end();it++)//todo 之后改成图片数量的多少 { string tmp = *it; std::cout<<"tmp path is"<<tmp<<std::endl; //cout<<tmp<<endl; //string strImgFile = img_dir + tmp + img_format; string strImgFile = tmp; usleep(200000);//4hz ros::Time timestamp_ros = ros::Time::now(); // --- for image ---// cv::Mat img = cv::imread(strImgFile); if (img.empty()) cout<<"图片为空: "<<strImgFile<<endl; cv_bridge::CvImage ros_image; sensor_msgs::ImagePtr ros_image_msg; ros_image.image = img; ros_image.encoding = "bgr8"; //cout<<"debug_______"<<endl; //ros::Time timestamp_ros2 = ros::Time::now(); ros_image_msg = ros_image.toImageMsg(); ros_image_msg->header.seq = seq; ros_image_msg->header.stamp = timestamp_ros; ros_image_msg->header.frame_id = "/image_raw"; bag.write("/camera/color/image_raw", ros_image_msg->header.stamp, ros_image_msg); cout<<"write frame: "<<seq<<endl; seq++; } cout<<"---end---"<<endl; return 0;} //con:文件格式 form:文件命名形式void GetFileNames(string path,vector<string>& filenames, string con){ DIR *pDir; struct dirent* ptr; string filename, format, name, name2; if(!(pDir = opendir(path.c_str()))) return; int num=0; while((ptr = readdir(pDir))!=0) { //跳过.和..文件 if(strcmp(ptr->d_name, ".") == 0 || strcmp(ptr->d_name, "..") == 0) continue; filename = ptr->d_name; format = filename.substr(filename.find("."), filename.length()); //name = filename.substr(0, filename.find(".")); name = filename.substr(0, filename.find(".")); cout<<filename<<"\t"<<name<<"\t"<<format<<endl; if(format == con)//也可以添加对文件名的要求 { filenames.push_back(name); num++; } } std::cout<<"file size of:"<<filenames.size()<<"****"<<num<<std::endl; closedir(pDir);} //cv::glob(路径,要放置路径下文件定义的容器,false)//复制子字符串substr(所需的子字符串的起始位置,默认值为0 , 复制的字符数目)返回值:一个子字符串,从其指定的位置开始//按图片名升序排列bool read_images(string path, vector<string> &image_files){ //fn存储path目录下所有文件的路径名称,如../images/0001.pngvector<cv::String> fn; cv::glob(path, fn, false); size_t count_1 = fn.size(); if (count_1 == 0) { cout << "file " << path << " not exits"<<endl; return -1; } //v1用来存储只剩数字的字符串 vector<string> v1; for (int i = 0; i < count_1; ++i) { //cout << fn[i] << endl; //1.获取不带路径的文件名,000001.jpg(获取最后一个/后面的字符串) string::size_type iPos = fn[i].find_last_of('/') + 1; string filename = fn[i].substr(iPos, fn[i].length() - iPos); //cout << filename << endl; //2.获取不带后缀的文件名,000001 string name = filename.substr(0, filename.rfind(".")); //cout << name << endl; v1.push_back(name); } //把v1升序排列 sort(v1.begin(), v1.end(),[](string a, string b) {return stoi(a) < stoi(b); }); string v = ".jpg"; size_t count_2 = v1.size(); for(int j = 0; j < count_2; ++j) { string z = path + v1[j] + v; image_files.push_back(z);//把完整的图片名写回来 }return true;}
四、编译及执行
4.1 编译
4.2 执行
1,首先打开一个终端,输入roscore,启动ros
2,打开新的终端,进入工作空间,执行以下语句
leedarson@leedarson-desktop:~/catkin_ws$ source devel/setup.bash
leedarson@leedarson-desktop:~/catkin_ws$ rosrun create_bag node
通过以上操作就可以将文件夹中的图像转换为bag包。
4.3 检测录制的bag包的话题和信息
rostopic list
rostopic echo
关于“C++基于ros怎么将文件夹中的图像转换为bag包”这篇文章的内容就介绍到这里,感谢各位的阅读!相信大家对“C++基于ros怎么将文件夹中的图像转换为bag包”知识都有一定的了解,大家如果还想学习更多知识,欢迎关注编程网行业资讯频道。