function multiObjectTracking()
% 创建用于读取视频、检测移动对象、% 并显示结果的系统对象
obj = setupSystemObjects();
tracks = initializeTracks(); % create an empty array of tracks
nextId = 1; % ID of the next track
% 检测移动物体，并在视频帧中跟踪它们
while ~isDone(obj.reader)
frame = readFrame();
[centroids, bboxes, mask] = detectObjects(frame);
predictNewLocationsOfTracks();
[assignments, unassignedTracks, unassignedDetections] = ...
detectionToTrackAssignment();
updateAssignedTracks();
updateUnassignedTracks();
deleteLostTracks();
createNewTracks();
displayTrackingResults();
end
%% 创建系统对象 % 创建用于读取视频帧、检测 % 前景对象和显示结果的系统对象。
function obj = setupSystemObjects()
% 初始化视频 I/O % 创建用于从文件中读取视频的对象，在每一帧中绘制跟踪的 % 对象，并播放视频。
vid = videoinput('winvideo', 1, 'YUY2_320x240');
src = getselectedsource(vid);
vid.FramesPerTrigger = 1;
% TriggerRepeat 从零开始，并且始终比触发器数量少 1%。
vid.TriggerRepeat = 899;
preview(vid);
start(vid);
stoppreview(vid);
savedvideo = getdata(vid);
% 创建一个视频文件阅读器
obj.reader = vision.VideoFileReader(savedvideo);
% 创建两个视频播放器，一个显示视频，一个显示前景遮罩
obj.videoPlayer = vision.VideoPlayer('Position', [20, 400, 700, 400]);
obj.maskPlayer = vision.VideoPlayer('Position', [740, 400, 700, 400]);
obj.detector = vision.ForegroundDetector('NumGaussians', 3, ...
'NumTrainingFrames', 40, 'MinimumBackgroundRatio', 0.7);
obj.blobAnalyser = vision.BlobAnalysis('BoundingBoxOutputPort', true, ...
'AreaOutputPort', true, 'CentroidOutputPort', true, ...
'MinimumBlobArea', 400);
end
function tracks = initializeTracks()
% 创建一个空的轨道数组
tracks = struct(...
'id', {}, ...
'bbox', {}, ...
'kalmanFilter', {}, ...
'age', {}, ...
'totalVisibleCount', {}, ...
'consecutiveInvisibleCount', {});
end
%% 读取一个视频帧 % 从视频文件中读取下一个视频帧。
function frame = readFrame()
frame = obj.reader.step();
end
function [centroids, bboxes, mask] = detectObjects(frame)
% 检测前景
mask = obj.detector.step(frame);
% 应用形态学运算来去除噪声并填充孔洞
mask = imopen(mask, strel('rectangle', [3,3]));
mask = imclose(mask, strel('rectangle', [15, 15]));
mask = imfill(mask, 'holes');
% 执行 blob 分析以查找连接的组件
[~, centroids, bboxes] = obj.blobAnalyser.step(mask);
end
%% 预测现有轨道的新位置 % 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)
bbox = tracks(i).bbox;
% 预测轨道的当前位置
predictedCentroid = predict(tracks(i).kalmanFilter);
% 移动边界框，使其中心位于 % 预测位置
predictedCentroid = int32(predictedCentroid) - bbox(3:4) / 2;
tracks(i).bbox = [predictedCentroid, bbox(3:4)];
end
end
function [assignments, unassignedTracks, unassignedDetections] = ...
detectionToTrackAssignment()
nTracks = length(tracks);
nDetections = size(centroids, 1);
% 计算将每个检测分配给每个轨道的成本
cost = zeros(nTracks, nDetections);
for i = 1:nTracks
cost(i, :) = distance(tracks(i).kalmanFilter, centroids);
end
% 解决分配问题
costOfNonAssignment = 20;
[assignments, unassignedTracks, unassignedDetections] = ...
assignDetectionsToTracks(cost, costOfNonAssignment);
end
function updateAssignedTracks()
numAssignedTracks = size(assignments, 1);
for i = 1:numAssignedTracks
trackIdx = assignments(i, 1);
detectionIdx = assignments(i, 2);
centroid = centroids(detectionIdx, :);
bbox = bboxes(detectionIdx, :);
% 使用新的检测纠正对物体位置的估计
correct(tracks(trackIdx).kalmanFilter, centroid);
% 用检测到的 % 边界框替换预测的边界框
tracks(trackIdx).bbox = bbox;
% 更新曲目的年龄
tracks(trackIdx).age = tracks(trackIdx).age + 1;
% 更新可见性
tracks(trackIdx).totalVisibleCount = ...
tracks(trackIdx).totalVisibleCount + 1;
tracks(trackIdx).consecutiveInvisibleCount = 0;
end
end
%% 更新未分配的曲目 % 将每个未分配的曲目标记为不可见，并将其年龄增加 1。
function updateUnassignedTracks()
for i = 1:length(unassignedTracks)
ind = unassignedTracks(i);
tracks(ind).age = tracks(ind).age + 1;
tracks(ind).consecutiveInvisibleCount = ...
tracks(ind).consecutiveInvisibleCount + 1;
end
end
function deleteLostTracks()
if isempty(tracks)
return;
end
invisibleForTooLong = 10;
ageThreshold = 8;
% 计算它可见的轨道年龄的比例
ages = [tracks(:).age];
totalVisibleCounts = [tracks(:).totalVisibleCount];
visibility = totalVisibleCounts ./ ages;
% 找到“丢失”曲目的索引
lostInds = (ages < ageThreshold & visibility < 0.6) | ...
[tracks(:).consecutiveInvisibleCount] >= invisibleForTooLong;
% 删除丢失的曲目
tracks = tracks(~lostInds);
end
function createNewTracks()
centroids = centroids(unassignedDetections, :);
bboxes = bboxes(unassignedDetections, :);
for i = 1:size(centroids, 1)
centroid = centroids(i,:);
bbox = bboxes(i, :);
% 创建卡尔曼滤波器对象
kalmanFilter = configureKalmanFilter('ConstantVelocity', ...
centroid, [200, 50], [100, 25], 100);
% 创建一个新轨道
newTrack = struct(...
'id', nextId, ...
'bbox', bbox, ...
'kalmanFilter', kalmanFilter, ...
'age', 1, ...
'totalVisibleCount', 1, ...
'consecutiveInvisibleCount', 0);
% 将其添加到曲目数组中
tracks(end + 1) = newTrack;
% 增加下一个 id
nextId = nextId + 1;
end
end
function displayTrackingResults()
% 将帧和掩码转换为 uint8 RGB
frame = im2uint8(frame);
mask = uint8(repmat(mask, [1, 1, 3])) .* 255;
minVisibleCount = 8;
if ~isempty(tracks)
% 噪声检测往往会导致短暂的轨道 % 仅显示已可见超过 % 最小帧数的轨道。
reliableTrackInds = ...
[tracks(:).totalVisibleCount] > minVisibleCount;
reliableTracks = tracks(reliableTrackInds);
% 显示对象。如果在此帧中未检测到对象，则显示其预测的边界框。
if ~isempty(reliableTracks)
% 获取边界框
bboxes = cat(1, reliableTracks.bbox);
% 获取 ID
ids = int32([reliableTracks(:).id]);
% 为对象创建标签，指示我们显示预测而不是实际 % 位置的对象
labels = cellstr(int2str(ids'));
predictedTrackInds = ...
[reliableTracks(:).consecutiveInvisibleCount] > 0;
isPredicted = cell(size(labels));
isPredicted(predictedTrackInds) = {' predicted'};
labels = strcat(labels, isPredicted);
% 在框架上绘制
frame = insertObjectAnnotation(frame, 'rectangle', ...
bboxes, labels);
% 画在面具上
mask = insertObjectAnnotation(mask, 'rectangle', ...
bboxes, labels);
end
end
% 显示蒙版和边框
obj.maskPlayer.step(mask);
obj.videoPlayer.step(frame);
end
displayEndOfDemoMessage(mfilename)
end