通过matlab视觉工具箱实现移动目标跟踪并绘制其运动轨迹,包含仿真操作录像,中文注释

function MouseTracking() obj = setupSystemObjects();%读取视频，检测移动对象，并显示结果的系统对象obj tracks = initializeTracks(); %初始化轨道数组 nextId = 1; a=1; SE=strel('square',50);%用于前期形态学腐蚀使用 while ~isDone(obj.reader)%使用卡尔曼滤波器跟踪预测目标 %以下函数在下面均有说明 frame = readFrame(); [centroids, bboxes, mask] = detectObjects(frame); predictNewLocationsOfTracks(); [assignments, unassignedTracks, unassignedDetections] = detectionToTrackAssignment(); updateAssignedTracks(); updateUnassignedTracks(); deleteLostTracks(); createNewTracks(); [cenb1x,cenb1y,cenb2x,cenb2y]=displayTrackingResults();%显示视频时返回两个小鼠的坐标点并储存，用于后期plot绘制小鼠轨迹 if a==1 cent1x=cenb1x; cent1y=cenb1y; cent2x=cenb2x; cent2y=cenb2y; else cent1x=[cent1x cenb1x]; cent1y=[cent1y cenb1y]; cent2x=[cent2x cenb2x]; cent2y=[cent2y cenb2y]; end a=a+1; end %绘制小鼠的轨迹图 plot(cent1x,cent1y); hold on; plot(cent2x,cent2y,'-r'); set(gca,'YDir','reverse'); legend('一号小鼠','二号小鼠',3); title('小鼠轨迹图'); %读取视频 function obj = setupSystemObjects() obj.reader = vision.VideoFileReader('mousevideo.mp4'); obj.videoPlayer = vision.VideoPlayer(); end %创建轨道数组用于跟踪小鼠目标 function tracks = initializeTracks() tracks = struct(... 'id', {}, ...%轨道的id,用于标记轨道 'bbox', {}, ...%小鼠外围的边界框 'kalmanFilter', {}, ...%卡尔曼滤波器用于预测下一个点的位置 'age', {}, ...%自首次检测到轨道以来的帧数 'totalVisibleCount', {}, ...%检测到轨道的帧总数（可见） 'consecutiveInvisibleCount', {});%未检测到轨道的连续帧数（不可见）。 end %读取视频帧 function frame = readFrame() frame = obj.reader.step(); end %检测前景，获取前景连通区域的中心点和边界框 function [centroids, bboxes, mask] = detectObjects(frame) frame1=im2uint8(frame); mask =zeros(1080,1920,3,'uint8'); for i=100:980 for j=464:1340 mask(i,j,:)=frame1(i,j,:); end end mask=im2bw(mask,0.2); mask=bwareaopen(mask, 200); mask=imfill(mask,'holes'); mask = imopen(mask, strel('rectangle', [40,40])); if a>270&&a<283 mask=imerode(mask,SE); end STATS = regionprops(mask,'Centroid','BoundingBox'); centroids=cat(1,STATS.Centroid); bboxes=cat(1,STATS.BoundingBox); end %使用卡尔曼滤波器预测当前帧中每个轨道的质心，并相应地更新其边界框 function predictNewLocationsOfTracks() for i = 1:length(tracks)%遍历已跟踪的轨迹 bbox = tracks(i).bbox; predictedCentroid = predict(tracks(i).kalmanFilter);%使用卡尔曼滤波器进行预测，得到预测中心 %移动边界框，使其中心位于预测位置 predictedCentroid = int32(predictedCentroid) - int32(bbox(3:4) / 2); tracks(i).bbox = [predictedCentroid, bbox(3:4)]; end end %通过最小化成本来完成将当前帧中的对象检测分配给现有轨道。 function [assignments, unassignedTracks, unassignedDetections] = detectionToTrackAssignment() %获取轨迹和新检测的目标的个数 nTracks = length(tracks); nDetections = size(centroids, 1); cost = zeros(nTracks, nDetections);%损失函数矩阵，行代表轨迹，列代表新检测目标 for i = 1:nTracks cost(i, :) = distance(tracks(i).kalmanFilter, centroids);%对于每个轨迹，使用它们的卡尔曼滤波器预测的结果，与每个新检测的目标中心计算欧式距离，存入损失函数矩阵（cost矩阵） end costOfNonAssignment = 40;%设置阈值，当损失函数矩阵中的某个值小于40时，则不分配轨迹 [assignments, unassignedTracks, unassignedDetections] = assignDetectionsToTracks(cost, costOfNonAssignment);%使用匈牙利匹配算法根据cost矩阵和阈值分配矩阵并检测目标 end function updateAssignedTracks()%更新已分配的轨迹 numAssignedTracks = size(assignments, 1); for i = 1:numAssignedTracks trackIdx = assignments(i, 1); detectionIdx = assignments(i, 2); centroid = centroids(detectionIdx, :); bbox = bboxes(detectionIdx, :); correct(tracks(trackIdx).kalmanFilter, centroid);%根据轨迹对应的检测目标位置中心修正它 的卡尔曼滤波器 tracks(trackIdx).bbox = bbox; tracks(trackIdx).age = tracks(trackIdx).age + 1;%轨迹年龄增加1 tracks(trackIdx).totalVisibleCount = tracks(trackIdx).totalVisibleCount + 1;%轨迹可见帧数增加1 tracks(trackIdx).consecutiveInvisibleCount = 0;%轨迹连续不可见帧数清零 end end function updateUnassignedTracks()%更新未分配轨迹 for i = 1:length(unassignedTracks) ind = unassignedTracks(i); tracks(ind).age = tracks(ind).age + 1;%轨迹年龄加1 tracks(ind).consecutiveInvisibleCount = tracks(ind).consecutiveInvisibleCount + 1;%轨迹连续不可见帧数加1 end end function deleteLostTracks()%删除丢失的轨迹 if isempty(tracks) return; end invisibleForTooLong = 20;%连续不可见帧数大于等于20时丢弃轨迹 ageThreshold = 8;%当年龄小于8时，若总可见帧数与年龄的比值小于0.6，则丢弃轨迹；否则再根据连续不可见帧数丢弃轨迹 ages = [tracks(:).age]; totalVisibleCounts = [tracks(:).totalVisibleCount]; visibility = totalVisibleCounts ./ ages; lostInds = (ages < ageThreshold & visibility < 0.6) | ... [tracks(:).consecutiveInvisibleCount] >= invisibleForTooLong; tracks = tracks(~lostInds); end function createNewTracks()%创建新轨迹 centroids = centroids(unassignedDetections, :); bboxes = bboxes(unassignedDetections, :); for i = 1:size(centroids, 1) centroid = centroids(i,:); bbox = bboxes(i, :); kalmanFilter = configureKalmanFilter('ConstantAcceleration', centroid, [1 1 1]*1e5,[25,10,10],25);%创建新的二阶加速度卡尔曼滤波器 newTrack = struct(... 'id', nextId, ... 'bbox', bbox, ... 'kalmanFilter', kalmanFilter, ... 'age', 1, ... 'totalVisibleCount', 1, ... 'consecutiveInvisibleCount', 0); tracks(end + 1) = newTrack; nextId = nextId + 1; end end function [cen1x,cen1y,cen2x,cen2y]=displayTrackingResults()%显示跟踪结果并返回轨迹中心点 frame = im2uint8(frame); mask = uint8(repmat(mask, [1, 1, 3])) .* 255; minVisibleCount = 1;%可见帧数阈值，大于阈值才显示 if ~isempty(tracks) reliableTrackInds = ... [tracks(:).totalVisibleCount] > minVisibleCount; reliableTracks = tracks(reliableTrackInds); if ~isempty(reliableTracks) bboxes = cat(1, reliableTracks.bbox); ids = int32([reliableTracks(:).id]); labels = cellstr(int2str(ids')); predictedTrackInds = ... [reliableTracks(:).consecutiveInvisibleCount] > 0; isPredicted = cell(size(labels)); isPredicted(predictedTrackInds) = {' predicted'}; labels = strcat(labels, isPredicted); frame = insertObjectAnnotation(frame, 'rectangle', bboxes, labels,'Color',{'cyan','yellow'},'FontSize',22); end %获取中心点坐标（有两只小鼠，两个中心点坐标） cena1x=bboxes(1,1)+bboxes(1,3)/2; cena1y=bboxes(1,2)+bboxes(1,4)/2; cena2x=bboxes(2,1)+bboxes(2,3)/2; cena2y=bboxes(2,2