✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。
🍎个人主页：Matlab科研工作室
🍊个人信条：格物致知。
更多Matlab仿真内容点击👇
⛄ 内容介绍
运动目标检测算法是从视频图像序列中提取感兴趣目标的一项技术,是目标分类、目标跟踪、目标行为等其他视频分析技术的前提条件,也是机器视觉和数字图像处理所研究的重点问题之一,因此对实时视频的运动目标检测算法的研究有着重要的研究价值和意义。
⛄ 部分代码
function [Xpoints1, Ypoints1, Xpoints2, Ypoints2, tms, yc] = ProcessVideo(videoFilePath)
if nargin < 1
videoFilePath = fullfile(pwd, 'video/video.avi');
end
% clc; clear all; close all;
% videoFilePath = fullfile(pwd, 'video/video.avi');
time_start = cputime;
[pathstr, name, ext] = fileparts(videoFilePath);
foldername = fullfile(pwd, sprintf('%s_images', name));
T = 1;
P = 5;
W1 = [75 95];
L1 = [360 17];
W2 = [55 55];
L2 = [35 1565];
Xpoints1 = [];
Ypoints1 = [];
Xpoints2 = [];
Ypoints2 = [];
Xpointst1 = [];
Ypointst1 = [];
Xpointst2 = [];
Ypointst2 = [];
figure('Position', get(0, 'ScreenSize'));
hg1 = subplot(1, 2, 1);
hg2 = subplot(1, 2, 2);
for frame = 1:146
filename = fullfile(foldername, sprintf('%04d.jpg', frame));
R = imread(filename);
Imi = R;
xc1 = 0;
yc1 = 0;
xc2 = 0;
yc2 = 0;
if frame > 72
I = rgb2hsv(Imi);
I = I(:,:,1);
I = roicolor(I, 0.1, 0.17);
MeanConverging1 = 1;
while MeanConverging1
M00 = 0.0;
for i = L1(1)-P : (L1(1)+W1(1)+P)
for j = L1(2)-P : (L1(2)+W1(2)+P)
if i > size(I,1) || j > size(I,2) || i < 1 || j < 1
continue;
end
M00 = M00 + double(I(i,j));
end
end
M10 = 0.0;
for i = L1(1)-P : (L1(1)+W1(1)+P)
for j = L1(2)-P : (L1(2)+W1(2)+P)
if i > size(I,1) || j > size(I,2) || i < 1 || j < 1
continue;
end
M10 = M10 + i * double(I(i,j));
end
end
M01 = 0.0;
for i = L1(1)-P : (L1(1)+W1(1)+P)
for j = L1(2)-P : (L1(2)+W1(2)+P)
if i > size(I,1) || j > size(I,2)|| i < 1 || j < 1
continue;
end
M01 = M01 + j * double(I(i,j));
end
end
xc1 = round(M10 / M00);
yc1 = round(M01 / M00);
oldL = L1;
L1 = [floor(xc1 - (W1(1)/2)) floor(yc1 - (W1(2)/2))];
if abs(oldL(1)-L1(1)) < T || abs(oldL(2)-L1(2)) < T
MeanConverging1 = 0;
end
end
s = round(1.1 * sqrt(M00));
W1 = [ s floor(1.2*s) ];
L1 = [floor(xc1 - (W1(1)/2)) floor(yc1 - (W1(2)/2))];
Xpoints1 = [Xpoints1 xc1];
Ypoints1 = [Ypoints1 yc1];
yc1t = yc1+randi(2,1,1)*25;
xc1t = xc1+randi(2,1,1)*25;
Xpointst1 = [Xpointst1 xc1t];
Ypointst1 = [Ypointst1 yc1t];
else
Xpoints1 = [Xpoints1 NaN];
Ypoints1 = [Ypoints1 NaN];
Xpointst1 = [Xpointst1 NaN];
Ypointst1 = [Ypointst1 NaN];
end
if frame > 90 && frame < 146
R = Imi;
I = rgb2ycbcr(R);
I = I(:,:,1);
I = mat2gray(I);
I = roicolor(I, 0.05, 0.3);
MeanConverging2 = 1;
while MeanConverging2
M00 = 0.0;
M00 = 0.0;
for i = L2(1)-P : (L2(1)+W2(1)+P),
for j = L2(2)-P : (L2(2)+W2(2)+P),
if i > size(I,1) || j > size(I,2) || i < 1 || j < 1
continue;
end
M00 = M00 + double(I(i,j));
end
end
M10 = 0.0;
for i = L2(1)-P : (L2(1)+W2(1)+P),
for j = L2(2)-P : (L2(2)+W2(2)+P),
if i > size(I,1) || j > size(I,2) || i < 1 || j < 1
continue;
end
M10 = M10 + i * double(I(i,j));
end
end
M01 = 0.0;
for i = L2(1)-P : (L2(1)+W2(1)+P),
for j = L2(2)-P : (L2(2)+W2(2)+P),
if i > size(I,1) || j > size(I,2)|| i < 1 || j < 1
continue;
end
M01 = M01 + j * double(I(i,j));
end
end
xc2 = round(M10 / M00);
yc2 = round(M01 / M00);
oldL = L2;
L2 = [floor(xc2 - (W2(1)/2)) floor(yc2 - (W2(2)/2))];
if abs(oldL(1)-L2(1)) < T || abs(oldL(2)-L2(2)) < T
MeanConverging2 = 0;
end
end
s = round(1.1 * sqrt(M00));
W2 = [ s floor(1.2*s) ];
L2 = [floor(xc2 - (W2(1)/2)) floor(yc2 - (W2(2)/2))];
Xpoints2 = [Xpoints2 xc2];
Ypoints2 = [Ypoints2 yc2];
yc2t = yc2+randi(2,1,1)*25;
xc2t = xc2+randi(2,1,1)*25;
Xpointst2 = [Xpointst2 xc2t];
Ypointst2 = [Ypointst2 yc2t];
else
Xpoints2 = [Xpoints2 NaN];
Ypoints2 = [Ypoints2 NaN];
Xpointst2 = [Xpointst2 NaN];
Ypointst2 = [Ypointst2 NaN];
end
axes(hg1); cla;
imshow(Imi, []); hold on;
if xc1 > 0 && yc1 > 0
plot(yc1, xc1, 'go', 'MarkerFaceColor', 'g');
plot(yc1t, xc1t, 'g+', 'MarkerFaceColor', 'g');
end
if xc2 > 0 && yc2 > 0
plot(yc2, xc2, 'bo', 'MarkerFaceColor', 'b');
plot(yc2t, xc2t, 'b+', 'MarkerFaceColor', 'b');
end
hold off; title(sprintf('%04d帧', frame));
bg = true(size(Imi,1), size(Imi,2));
axes(hg2); cla; imshow(bg);
hold on; box on;
plot(Ypoints1, Xpoints1, 'go-', 'MarkerFaceColor', 'g');
plot(Ypoints2, Xpoints2, 'bo-', 'MarkerFaceColor', 'b');
hold off; title(sprintf('%04d帧', frame));
pause(0.001);
end
time_end = cputime;
tms = time_end - time_start;
yc.Xpointst1 = Xpointst1;
yc.Ypointst1 = Ypointst1;
yc.Xpointst2 = Xpointst2;
yc.Ypointst2 = Ypointst2;
⛄ 运行结果
⛄ 参考文献
[1] 张学荣, 雷国伟. 基于灰度直方图的背景差目标检测算法[J]. 计算机安全, 2010(02):45-47.
[2] 张学荣. 直方图在视频监控系统中的应用[J]. 电脑知识与技术, 2009, 005(030):8520-8521.
[3] 戴鑫, 黄愐, 张进. 帧间差法在视频目标检测的仿真应用[J]. 电子技术与软件工程, 2021, 000(021):P.110-111.
[4] 邱祯艳. 基于实时视频的运动目标检测算法[D]. 中国计量学院.
