物体检测软件

Question

我想创建一个运动检测软件，通过 Live Stream Camera 获取输入视频。该软件必须执行以下操作：

1- 当检测到移动时，它会在有移动的区域周围绘制一个矩形。 - 对于绘制矩形，我想使用 vision.CascadeObjectDetector，因为有一个他们用来进行人脸识别的示例，但问题是我不知道 CascadeObjectDetector 是否适用于我制作的代码（视频输入）。应该用imaqVideoDevice来做吗？-

2- 记录在恒定时间内屏幕上有多少次移动，并返回一分钟内有多少次移动。就像每分钟计算一次动作一样。 - 正如您在代码中看到的那样，我使用 tic-toc 函数来制作计数器和“cont”变量来计算注册的运动。我想知道，我需要做什么才能完成柜台...-

虽然我对 Matlab 有点陌生，但我知道它的库充满潜力，我愿意进一步了解它。 这是我的代码。你们可能想忽略注释部分，因为它是巴西葡萄牙语。

提前致谢。

---

我为vision.ForegroundDetector 重建了我的代码，但现在我遇到了问题...我不知道如何在subplot(1,2,1); imshow(FrameRGB); title('Imagem Recebida') 上重现它的结果，我唯一知道的是我没有使用imshow...

%O botão Fechar, deve finalizar o código e parar a câmera
uicontrol('String', 'Fechar',...
          'Callback', 'stop(camera)',...
          'Callback', 'close');
camera = videoinput('winvideo', 1, 'MJPG_320x240');

Detect = vision.ForegroundDetector(...
       'NumTrainingFrames', 100, ... % 5 because of short video
       'InitialVariance', 30*30); % initial standard deviation of 30
hblob = vision.BlobAnalysis(...
       'CentroidOutputPort', false, 'AreaOutputPort', false, ...
       'BoundingBoxOutputPort', true, ...
       'MinimumBlobAreaSource', 'Property', 'MinimumBlobArea', 250);
hsi = vision.ShapeInserter('BorderColor','White');

set(camera,'TriggerRepeat', Inf);
camera.FrameGrabInterval = 1.5;

cont = 0; %cont é a variável que vai armazenar o movimento
A = 0;
timerEnd = 0;

start(camera)
timerStart = tic;
hsnk = vision.VideoPlayer();

while(camera.FramesAcquired>=0)

        FrameRGB = getsnapshot(camera);
        fgMask = step(Detect, FrameRGB);
        bbox   = step(hblob, fgMask);
        out    = step(hsi, FrameRGB, bbox); % draw bounding boxes around cars
        %drawnow;
        subplot(1,2,1); imshow(hsnk); title('Imagem Recebida')
        text(cont, timerEnd,...
        ['Contador:', num2str(cont/timerEnd)],...
        'HorizontalAlignment', 'left');


        IM = getdata(camera,2);
        IMGray = rgb2gray(IM(:,:,1:3));
        IMbw=im2bw(IMGray,0.5);
        IMneg=imadjust(IMGray,[0 1],[1 0]);

        i1=IM(:,:,:,1);
        i2=IM(:,:,:,2);
        i1=rgb2gray(i1(:,:,1:3));
        i2=rgb2gray(i2(:,:,1:3));
        m=abs(double(i1)-double(i2))/256;
        subplot(1,2,2);imshow(m);title('Detecção de Movimentos')



        if sum(sum(m))>700

            cont = cont+1;
            A = mod(cont,0);
        end

        timerEnd = toc(timerStart);
        fprintf('Em %d minutos e %f segundos, obteve-se %14.0f movimentos  \n', floor(timerEnd/60), rem(timerEnd,60), A);

end

stop(camera)

---

%O botão Fechar, deve finalizar o código e parar a câmera
uicontrol('String', 'Fechar',...
          'Callback', 'stop(camera)',...
          'Callback', 'close');
camera = videoinput('winvideo', 1, 'MJPG_320x240');

Detect = vision.ForegroundDetector(...
       'NumTrainingFrames', 10, ... % 5 because of short video
       'InitialVariance', 4.5*4.5); % initial standard deviation of 30

hblob = vision.BlobAnalysis(...
       'CentroidOutputPort', false, 'AreaOutputPort', false, ...
       'BoundingBoxOutputPort', true, ...
       'MinimumBlobAreaSource', 'Property', 'MinimumBlobArea', 250);
hsi = vision.ShapeInserter('BorderColor','White');

set(camera,'TriggerRepeat', Inf);
camera.FrameGrabInterval = 1.5;

cont = 0; %cont é a variável que vai armazenar o movimento
A = 0;


start(camera)
timerStart = tic;
% hsnk = vision.VideoPlayer();


while(camera.FramesAcquired>=0)

        FrameRGB = getsnapshot(camera);
        fgMask = step(Detect, FrameRGB);
        bbox   = step(hblob, fgMask);
        out    = step(hsi, FrameRGB, bbox); % draw bounding boxes around cars
        subplot(1,2,1);imshow(out);title('Imagem x')


        IM = getdata(camera,2);
        IMGray = rgb2gray(IM(:,:,1:3));
        IMbw=im2bw(IMGray,0.5);
        IMneg=imadjust(IMGray,[0 1],[1 0]);

        i1=IM(:,:,:,1);
        i2=IM(:,:,:,2);
        i1=rgb2gray(i1(:,:,1:3));
        i2=rgb2gray(i2(:,:,1:3));
        m=abs(double(i1)-double(i2))/256;
        subplot(1,2,2);imshow(m);title('Detecção de Movimentos')



        if sum(sum(m))>700

            cont = cont+1;
            A = mod(cont,0); 
        end

         if toc(timerStart) == 15.0000
            plot(A, timerEnd, 'c');
            cont = 0;      
            A = 0;
        end 

        timerEnd = toc(timerStart);
        fprintf('Em %f segundos, obteve-se %14.0f movimentos  \n',rem(timerEnd,60), A);
        %fprintf('Em %d minutos e %f segundos, obteve-se %14.0f movimentos  \n', floor(timerEnd/60), rem(timerEnd,60), A);

end


stop(camera)

剩下的唯一问题是计数器，我试图通过这一行来构建它：

 if timerEnd == 15.0000
                plot(A, timerEnd, 'c');
                cont = 0;      
                A = 0;
            end 

它有什么问题，还是只是所在的地方无法正常工作？ 再次感谢！

Answer 1

vision.CascadeObjectDetector 用于检测特定的对象类别，例如面部、眼睛、鼻子等。对于运动检测，您应该使用vision.ForegroundDetector，它会给您一个二进制掩码，将像素标记为背景（静止）或前景（移动）。但是，这仅在您的相机不移动时才有效。

查看example了解更多详情。

请参阅example，了解如何将 vision.VideoPlayer 放入自定义 UI。