技术栈：
FFmpe，appium,OBS,opencv
思路：
- 通过自动化录制的测试视频
- 利用FFmpe选择兴趣区域进行截取
- 拿到兴趣区域进行视频前10s，中间10s，后10s 视频
- 把三段10s的视频进行每50ms一张图片
- 通过opencv进行图片的分析
主要依据人的视线规则是相同的图片持续200ms，就是认为是卡顿的。我根据这个规则，通过opencv比较步骤四的图片。第n个图片对比n+1的图片比较像素相似度。然后相似度在多少范围内持续了4张图片（50ms*4=200ms）就认为这段视频是卡段的
计算出卡顿率
上代码
通过自动化录制的测试视频这个就不写了，大致就是通过OBS启动虚拟摄像头
然后学生端显示老师的摄像头就是虚拟摄像头投射的测试视频
通过利用FFmpe选择兴趣区域进行截取
注：crop:ow[:oh[:x[:y:[:keep_aspect]]]]
crop的使用
ffmpe -i D:\Users\admin\Desktop\test\丰金莉分享的视频.mp4 -vf crop=327:184:641:200 D:\Users\admin\Desktop\test\test1.mp4
拿到兴趣区域进行视频前10s，中间10s，后10s 视频\
ffmpe -ss 00:00:00 -i D:\Users\admin\Desktop\test\test1.mp4 -vcodec copy -acodec copy -t 00:00:10 D:\Users\admin\Desktop\test\before10.mp4
把三段10s的视频进行每50ms一张图片
ffmpe -i D:\\Users\\admin\\Desktop\\test\\before10.mp4 -f image2 -vf fps=fps=20 D:\\Users\\admin\\Desktop\\test\\before\\%d.png
通过opencv进行图片的分析
package com.abcnull.tools;
import java.awt.HeadlessException;
import java.awt.image.BufferedImage;
import java.io.ByteArrayInputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.UnsupportedEncodingException;
import java.text.DecimalFormat;
import java.util.ArrayList;
import java.util.List;
import javax.imageio.ImageIO;
import org.opencv.core.Core;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.MatOfByte;
import org.opencv.core.MatOfPoint;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.utils.Converters;
public class ImageCompare {
private boolean compareResult = false;
private String mark = "_compareResult";
/**
* 比较两张图片，如不同则将不同处标记并输出到新的图片中
* @param imagePath1 图片1的路径
* @param imagePath2 图片2的路径
*/
public Integer CompareAndMarkDiff(String imagePath1, String imagePath2)
{
Mat mat1 = readMat(imagePath1);
Mat mat2 = readMat(imagePath2);
mat1 = Imgcodecs.imdecode(mat1, Imgcodecs.IMREAD_UNCHANGED);
mat2 = Imgcodecs.imdecode(mat2, Imgcodecs.IMREAD_UNCHANGED);
/*Mat mat1 = Imgcodecs.imread(imagePath1, Imgcodecs.IMREAD_UNCHANGED);
Mat mat2 = Imgcodecs.imread(imagePath2, Imgcodecs.IMREAD_UNCHANGED);*/
if(mat1.cols() == 0 || mat2.cols() == 0 || mat1.rows() == 0 || mat2.rows() == 0)
{
System.out.println("图片文件路径异常，获取的图片大小为0，无法读取");
return 0;
}
if(mat1.cols() != mat2.cols() || mat1.rows() != mat2.rows())
{
System.out.println("两张图片大小不同，无法比较");
return 0;
}
mat1.convertTo(mat1, CvType.CV_8UC1);
mat2.convertTo(mat2, CvType.CV_8UC1);
Mat mat1_gray = new Mat();
Imgproc.cvtColor(mat1, mat1_gray, Imgproc.COLOR_BGR2GRAY);
Mat mat2_gray = new Mat();
Imgproc.cvtColor(mat2, mat2_gray, Imgproc.COLOR_BGR2GRAY);
mat1_gray.convertTo(mat1_gray, CvType.CV_32F);
mat2_gray.convertTo(mat2_gray, CvType.CV_32F);
double result = Imgproc.compareHist(mat1_gray, mat2_gray, Imgproc.CV_COMP_CORREL);
if(result == 1)
{
System.out.println("两个图片完全相同");
compareResult = true;//此处结果为1则为完全相同
return 100;
}
int a= new Double(result*100).intValue();;
System.out.println("相似度数值为:"+a+"%");
// Mat mat_result = new Mat();
// //计算两个灰度图的绝对差值，并输出到一个Mat对象中
// Core.absdiff(mat1_gray, mat2_gray, mat_result);
// //将灰度图按照阈值进行绝对值化
// mat_result.convertTo(mat_result, CvType.CV_8UC1);
// List<MatOfPoint> mat2_list = new ArrayList<MatOfPoint>();
// Mat mat2_hi = new Mat();
// //寻找轮廓图
// Imgproc.findContours(mat_result, mat2_list, mat2_hi, Imgproc.RETR_EXTERNAL, Imgproc.CHAIN_APPROX_SIMPLE);
// Mat mat_result1 = mat1;
// Mat mat_result2 = mat2;
//使用红色标记不同点
//System.out.println(mat2_list.size());
// for (MatOfPoint matOfPoint : mat2_list)
// {
// Rect rect = Imgproc.boundingRect(matOfPoint);
// Imgproc.rectangle(mat_result1, rect.tl(), rect.br(), new Scalar(0, 0, 255),2);
// Imgproc.rectangle(mat_result2, rect.tl(), rect.br(), new Scalar(0, 0, 255),2);
// }
// String fileName1 = getFileName(imagePath1);
// String targetPath1 = getParentDir(imagePath2)+File.separator+fileName1.replace(".", mark+".");
// String fileName2 = getFileName(imagePath2);
// String targetPath2 = getParentDir(imagePath2)+File.separator+fileName2.replace(".", mark+".");
// System.out.println(targetPath1);
// System.out.println(targetPath2);
//图片一的带标记的输出文件；
// Imgcodecs.imwrite(targetPath1, mat_result1);
//图片二的带标记的输出文件；
// Imgcodecs.imwrite(targetPath2, mat_result2);
//writeImage(mat_result1, targetPath1);
//writeImage(mat_result2, targetPath2);
return a;
}
private void writeImage(Mat mat, String outPutFile)
{
MatOfByte matOfByte = new MatOfByte();
Imgcodecs.imencode(".png", mat, matOfByte);
byte[] byteArray = matOfByte.toArray();
BufferedImage bufImage = null;
try {
InputStream in = new ByteArrayInputStream(byteArray);
bufImage = ImageIO.read(in);
ImageIO.write(bufImage, "png", new File(outPutFile));
} catch (IOException | HeadlessException e)
{
e.printStackTrace();
}
}
private String getFileName(String filePath)
{
File f = new File(filePath);
return f.getName();
}
private String getParentDir(String filePath)
{
File f = new File(filePath);
return f.getParent();
}
private Mat readMat(String filePath)
{
try {
File file = new File(filePath);
FileInputStream inputStream = new FileInputStream(filePath);
byte[] byt = new byte[(int) file.length()];
int read = inputStream.read(byt);
List<Byte> bs = convert(byt);
Mat mat1 = Converters.vector_char_to_Mat(bs);
return mat1;
} catch (UnsupportedEncodingException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
}
return new Mat();
}
private List<Byte> convert(byte[] byt)
{
List<Byte> bs = new ArrayList<Byte>();
for (int i = 0; i < byt.length; i++)
{
bs.add(i, byt[i]);
}
return bs;
}
public static void main(String[] args) {
List<Integer> list=new ArrayList<>();
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
ImageCompare imageCompare=new ImageCompare();
for (int i = 1; i < 200; i++) {
System.out.println("out"+i+".png"+"比"+"out"+(i+1)+".png"+"结果：");
int similarity = imageCompare.CompareAndMarkDiff("D:\\Users\\admin\\Desktop\\test\\before\\" + i + ".png", "D:\\Users\\admin\\Desktop\\test\\before\\" + (i + 1) + ".png");
System.out.println(similarity);
list.add(similarity);
}
list.add(0);
System.out.println(list.toString());
int sum=1;
List<Integer> num100=new ArrayList<>();
for(Integer num :list){
if(98>num){
if(sum>4){
num100.add(sum);
System.out.println("sum="+sum);
}
sum =1;
}else {
sum+=1;
}
}
int sum1=0;
for (int i = 0; i < num100.size(); i++) {
sum1=num100.get(i)+sum1;
}
System.out.println(sum1);
System.out.println(sum1*50);
System.out.println(num100.size()*200);
int aa=sum1*50;
int bb=num100.size()*200;
int cc=aa-bb;
int dd=aa/100;
System.out.println("卡顿率为："+dd+"%");
}
}
网友评论