写出代码用matlab提取出视频前景目标，实现对运动物体的目标识别与跟踪，并用bounding box标识目标位置，然后导出视频，不使用背景建模和连通组

好的，以下是一个简单的基于帧差法的前景目标提取、目标跟踪和标识算法的Matlab代码：

% 读取视频
v = VideoReader('input_video.mp4');

% 初始化
frame = readFrame(v);
prev_frame = rgb2gray(frame);
prev_prev_frame = prev_frame;
h = size(frame, 1);
w = size(frame, 2);
foreground = zeros(h, w);
bounding_boxes = [];

% 逐帧处理
while hasFrame(v)
    % 读取当前帧
    frame = readFrame(v);
    curr_frame = rgb2gray(frame);
    
    % 计算当前帧和前一帧的差分图像
    diff = abs(double(curr_frame) - double(prev_frame)) > 20;
    diff_prev = abs(double(prev_frame) - double(prev_prev_frame)) > 20;
    diff = diff & diff_prev;
    
    % 消除噪声
    se = strel('disk', 2);
    diff = imopen(diff, se);
    diff = imclose(diff, se);
    
    % 寻找连通区域
    [L, num] = bwlabel(diff);
    stats = regionprops(L, 'Area', 'BoundingBox');
    
    % 筛选目标
    for i = 1:num
        if stats(i).Area > 100 && stats(i).Area < 5000
            bounding_boxes(end+1, :) = stats(i).BoundingBox;
            foreground(stats(i).BoundingBox(2):stats(i).BoundingBox(2)+stats(i).BoundingBox(4)-1, ...
                       stats(i).BoundingBox(1):stats(i).BoundingBox(1)+stats(i).BoundingBox(3)-1) = ...
                foreground(stats(i).BoundingBox(2):stats(i).BoundingBox(2)+stats(i).BoundingBox(4)-1, ...
                           stats(i).BoundingBox(1):stats(i).BoundingBox(1)+stats(i).BoundingBox(3)-1) | L(stats(i).BoundingBox(2):stats(i).BoundingBox(2)+stats(i).BoundingBox(4)-1, ...
                                                                                                               stats(i).BoundingBox(1):stats(i).BoundingBox(1)+stats(i).BoundingBox(3)-1) == i;
        end
    end
    
    % 显示结果
    frame(foreground > 0) = 255;
    for i = 1:size(bounding_boxes, 1)
        frame = insertShape(frame, 'Rectangle', bounding_boxes(i, :), 'LineWidth', 2);
    end
    imshow(frame);
    
    % 更新状态
    prev_prev_frame = prev_frame;
    prev_frame = curr_frame;
    bounding_boxes = [];
    foreground = zeros(h, w);
end

这个算法基于帧差法，首先计算当前帧和前一帧的差分图像，然后消除噪声，寻找连通区域，筛选出符合条件的目标（根据面积大小筛选），最后在原始图像上用bounding box标识目标位置。这个算法比较简单，可能对于复杂场景效果不佳，但对于简单场景还是比较有效的。