2020 ZhuHai WanShan International Intelligent Vessel Competition

Yat-Sen Robot from SYSU

无人船比赛：检测球及其颜色

（标签）一般使用只需看RELEASE ，调试需要参考DEBUG

RELEASE 编译

git clone https://github.com/sysu18364015/visual.git
cd visual
mkdir build && cd build
cmake ..
make

RELEASE 使用

在visual/build/文件夹下：

roscore
source <path/to/devel/setup.sh>
roslaunch hikvision_ros hik.launch

两个关键的ROS publisher:

image发布者:topic名称/hik_camera,topic数据类型 sensor_msgs::Image

color发布者:topic名称/ballcolor,topic数据类型 turtlesim::Color

DEBUG 关于相机实时图像回传

frame频率

使用rand降低处理频率，这部分代码在src/hik_cam.cpp中：

void HikvisionCamera::decodeCallback(int nPort, char *pBuf, int nSize, FRAME_INFO *pFrameInfo)
{
    int lFrameType = pFrameInfo->nType;

    if (lFrameType == T_YV12)
    {
        int x=rand()%16+1;  
        if(x!=1) return;

        cv::Mat picBGR;
        cv::Mat picYV12 = cv::Mat(pFrameInfo->nHeight * 3/2, pFrameInfo->nWidth, CV_8UC1, pBuf);
        cv::cvtColor(picYV12, picBGR, cv::COLOR_YUV2BGR_YV12);
        COLOR c = detect_pipeline(picBGR);
        ...

原图还是加了检测结果的图

detect_pipeline.h中定义是否画block颜色检测、矩形框、形状的圆形检测

DEBUG 改参数

params.h

颜色统计值 src/color.cpp--colorClassify

DEBUG 可执行文件

在visual/build/devel/lib/hikvision_ros/文件夹下可以看到编译出来的几个executable：

localtest

涉及主要的函数调用为：src/localtest.cpp中：decodeCallback--detect_pipeline

使用方法：

首先 图像所在文件夹下生成imgs.txt

ls *.jpg > imgs.txt

然后 在visual/build/devel/lib/hikvision_ros/下执行

./localtest path/to/images/

hik_cam_node

src/hik_cam.cpp中：decodeCallback--detect_pipeline

DEBUG 算法流程

find_horinzon_line

colorDetect

colorDetectBlock：得到属于各颜色的block的list

	colorClassify

	框内颜色占比例：这个GRID中不到1/3的像素点为这个颜色，则认为结果不对

	框内颜色一致性：这个GRID RGB STDERR大于阈值，则认为结果不对

seaskyline.cpp中的filterByLine

	取线的最高和最低点均值作为阈值，过滤掉海天线上方的

getColorCirclesRect：各颜色的block的list用矩形框住（GRID左上角坐标，目标框应包含至右下角）

	GRID数量太少则忽略

	去掉NUM_OUTIERS个最大最小值

	rectangle长宽比：如果太大认为它不是球，过滤掉
	
	矩形最上方距离海天线：如果大于图像高度的1/6，过滤
    
    矩形最上方如果在图像高度的下1/4之下，过滤

取面积最大的长方形对应的那个颜色作为返回值

可能用到的修改

parameters

You can specify some camera and steam parameters by hik.launch

<arg name="ip_addr" default="192.168.5.100"/>
<arg name="user_name" default="admin"/>
<arg name="password" default="admin"/>
<arg name="port" default="8000"/>
<arg name="channel" default="1"/>

Or in command line

roslaunch hikvision_ros hik.launch ip_addr:=192.168.5.100 password:=123456

support for camera_calibration

you can use camera_calibration to calibrate hikvision camera, hik_ros provides set_camera_info sevice for receiving and storing camera's calibration parameters.

example

# open camera
roslaunch hikvision_ros hik.launch

# see topic name
rostopic list

# [output]
# ➜  ~ rostopic list
# /hik_cam_node/camera_info
# /hik_cam_node/hik_camera
# ...

# calibrate
rosrun camera_calibration cameracalibrator.py --size 11x8 --square 0.03 image:=/hik_cam_node/hik_camera  camera:=/hik_cam_node/hik_camera

then begin calibration. After calibration you straightly press commit button, hik_ros has the ability to save the calibration parameter to camera_info_url, which is set in launch file OR use default path ( ~/.ros/camera_info )

# [output]
#[ INFO] [1551582484.454024618]: New camera info received
#[ INFO] [1551582484.454296067]: [ hik_camera ] Write camera_info to ~/.ros/camera_info/hik_camera.yaml success.