% function obj = setupSystemObjects(file_name)
% Initialize Video I/O
% Create objects for reading a video from a file, drawing the tracked
% objects in each frame, and playing the video.
% Inputs:
% file_name: string
% Outputs:
% obj sturct
function obj = setupSystemObjects(file_name)
global obj;
global tracks;
% Create a video file reader.
obj.reader = vision.VideoFileReader(file_name);
% Create two video players, one to display the video,
% and one to display the foreground mask.
obj.maskPlayer = vision.VideoPlayer('Position', [740, 400, 700, 400]);
obj.videoPlayer = vision.VideoPlayer('Position', [20, 400, 700, 400]);
% Create System objects for foreground detection and blob analysis
% The foreground detector is used to segment moving objects from
% the background. It outputs a binary mask, where the pixel value
% of 1 corresponds to the foreground and the value of 0 corresponds
% to the background.
obj.detector = vision.ForegroundDetector('NumGaussians', 3, ...
'NumTrainingFrames', 40, 'MinimumBackgroundRatio', 0.7);
% Connected groups of foreground pixels are likely to correspond to moving
% objects. The blob analysis System object is used to find such groups
% (called 'blobs' or 'connected components'), and compute their
% characteristics, such as area, centroid, and the bounding box.
obj.blobAnalyser = vision.BlobAnalysis('BoundingBoxOutputPort', true, ...
'AreaOutputPort', true, 'CentroidOutputPort', true, ...
'MinimumBlobArea', 600);
end
% function [centroids, bboxes, mask] = detectObjects(frame)
% This function is to dectect moving object
% Inputs:
% frame: frame
% Outputs:
% centroids: array
% bboxes: array
% mask: array
function [centroids, bboxes, mask] = detectObjects(frame)
global obj;
global tracks;
% Detect foreground.
mask = obj.detector.step(frame);
% Apply morphological operations to remove noise and fill in holes.
mask = imopen(mask, strel('rectangle', [3,3]));
mask = imclose(mask, strel('rectangle', [15, 15]));
mask = imfill(mask, 'holes');
% Perform blob analysis to find connected components.
[~, centroids, bboxes] = obj.blobAnalyser.step(mask);
end
% function predictNewLocationsOfTracks(mask,centroids)
% Use the Particle filter to predict the centroid of each
% track in the current frame, and update its bounding box accordingly.
% Inputs:
% mask: array
% centroids: array
% Outputs:
%
function predictNewLocationsOfTracks(mask,centroids)
global obj;
global tracks;
for i = 1:length(tracks)
bbox = tracks(i).bbox;
% Predict the current location of the track.
tracks(i).particles = pfDiffusion(tracks(i).particles,mask);
tracks(i).particles = pfResample(tracks(i).particles,mask,centroids);
predictedCentroid = mean(tracks(i).particles);
% Shift the bounding box so that its center is at
% the predicted location.
predictedCentroid = int32(predictedCentroid(1:2)) - bbox(3:4) / 2;
tracks(i).bbox = [predictedCentroid, bbox(3:4)];
end
end
% function [assignments, unassignedTracks, unassignedDetections] = ...
% detectionToTrackAssignment(centroids)
% compute an assignment which minimizes the total cost
% Inputs:
% centroids: array
% Outputs:
% assignments: array
% unassignedTracks: array
% unassignedDetections: array
function [assignments, unassignedTracks, unassignedDetections] = ...
detectionToTrackAssignment(centroids)
global obj;
global tracks;
nTracks = length(tracks);
nDetections = size(centroids, 1);
% Compute the cost of assigning each detection to each track.
cost = zeros(nTracks, nDetections);
for i = 1:nTracks
cost(i, :) = distance(mean(tracks(i).particles), centroids);
end
% Solve the assignment problem.
costOfNonAssignment = 20;
[assignments, unassignedTracks, unassignedDetections] = ...
assignDetectionsToTracks(cost, costOfNonAssignment);
end
% function updateAssignedTracks(assignments,centroids, bboxes)
% updates each assigned track with the corresponding detection
% Inputs:
% assignments: array
% centroids: array
% bboxes: array
% Outputs:
%
function updateAssignedTracks(assignments,centroids, bboxes)
global obj;
global tracks;
numAssignedTracks = size(assignments, 1);
for i = 1:numAssignedTracks
trackIdx = assignments(i, 1);
detectionIdx = assignments(i, 2);
centroid = centroids(detectionIdx, :);
bbox = bboxes(detectionIdx, :);
% Correct the estimate of the object's location
% using the new detection.
tracks(trackIdx).particles= pfCorrect(tracks(trackIdx).particles,
centroid);
% Replace predicted bounding box with detected
% bounding box.
tracks(trackIdx).bbox = bbox;
% Update track's age.
tracks(trackIdx).age = tracks(trackIdx).age + 1;
% Update visibility.
tracks(trackIdx).totalVisibleCount = ...
tracks(trackIdx).totalVisibleCount + 1;
tracks(trackIdx).consecutiveInvisibleCount = 0;
end
end
% function tracks=updateUnassignedTracks(unassignedTracks)
% Mark each unassigned track as invisible, and increase its age by 1.
% Inputs:
% unassignedTracks: array
% Outputs:
% tracks: struct
function tracks=updateUnassignedTracks(unassignedTracks)
global obj;
global tracks;
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()
% deletes tracks that have been invisible for too many consecutive frames
% Inputs:
%
% Outputs:
%
function deleteLostTracks()
global obj;
global tracks;
if isempty(tracks)
return;
end
invisibleForTooLong = 30;
ageThreshold = 8;
% Compute the fraction of the track's age for which it was visible.
ages = [tracks(:).age];
totalVisibleCounts = [tracks(:).totalVisibleCount];
visibility = totalVisibleCounts ./ ages;
% Find the indices of 'lost' tracks.
lostInds = (ages < ageThreshold & visibility < 0.4) | ...
[tracks(:).consecutiveInvisibleCount] >= invisibleForTooLong;
% Delete lost tracks.
tracks = tracks(~lostInds);
end
% function nextId = createNewTracks(centroids, unassignedDetections,
bboxes,nextId)
% Create new tracks from unassigned detections
% Inputs:
% centroids array
% unassignedDetections: array
% bboxes: array
% nextId: integer
% Outputs:
% nextId: array
function nextId = createNewTracks(centroids, unassignedDetections,
bboxes,nextId)
global obj;
global tracks;
centroids = centroids(unassignedDetections, :);
bboxes = bboxes(unassignedDetections, :);
for i = 1:size(centroids, 1)
centroid = centroids(i,:);
bbox = bboxes(i, :);
% Create a Particle filter object.
particles=ones(100,2)*0.5*[centroid;centroid];
% Create a new track.
newTrack = struct(...
'id', nextId, ...
'bbox', bbox, ...
'particles', particles, ...
'age', 1, ...
'totalVisibleCount', 1, ...
'consecutiveInvisibleCount', 0, ...
'showId', 0);
% Add it to the array of tracks.
tracks(end + 1) = newTrack;
% Increment the next id.
nextId = nextId + 1;
end
end
% function showId = displayTrackingResults( frame,mask,showId)
% draws a bounding box and label ID for each track on the video frame and the
foreground mask.
% It then displays the frame and the mask in their respective video players
% Inputs:
% frame frame
% mask: array
% showId: integer
% Outputs:
% showId: integer
function showId = displayTrackingResults( frame,mask,showId)
global obj;
global tracks;
% Convert the frame and the mask to uint8 RGB.
frame = im2uint8(frame);
mask = uint8(repmat(mask, [1, 1, 3])) .* 255;
minVisibleCount = 8;
if ~isempty(tracks)
% Noisy detections tend to result in short-lived tracks.
% Only display tracks that have been visible for more than
% a minimum number of frames.
reliableTrackInds = ...
[tracks(:).totalVisibleCount] > minVisibleCount;
reliableTracks = tracks(reliableTrackInds);
for i=1:length(tracks)
if reliableTrackInds(i) == 1 && tracks(i).showId == 0
tracks(i).showId = showId;
showId = showId + 1;
end
end
% Display the objects. If an object has not been detected
% in this frame, display its predicted bounding box.
if ~isempty(reliableTracks)
% Get bounding boxes.
bboxes = cat(1, reliableTracks.bbox);
% Get ids.
ids = int32([reliableTracks(:).showId]);
% Create labels for objects indicating the ones for
% which we display the predicted rather than the actual
% location.
labels = cellstr(int2str(ids'));
predictedTrackInds = ...
[reliableTracks(:).consecutiveInvisibleCount] > 0;
isPredicted = cell(size(labels));
isPredicted(predictedTrackInds) = {' predicted'};
labels = strcat(labels, isPredicted);
% Draw the objects on the frame.
for i=1:length(labels)
frame = insertObjectAnnotation(frame, 'circle', ...
[tracks(i).particles(:,1),tracks(i).particles(:,2)
ones(size(tracks(i).particles,1),1)*1],labels(i));
end
frame = insertObjectAnnotation(frame, 'rectangle', ...
bboxes, labels);
% Draw the objects on the mask.
mask = insertObjectAnnotation(mask, 'rectangle', ...
bboxes, labels);
end
end
% Display the mask and the frame.
obj.maskPlayer.step(mask);
obj.videoPlayer.step(frame);
end
% function Particles = pfPredict(Particles)
% This function performs particle diffusion
% Inputs:
% Particles: structure
% mask: array
% Outputs:
% Particles: structure
function Particles = pfDiffusion(Particles,mask)
Particles= Particles + randn(size(Particles,1) ,2)*4;
for i = 1: size(Particles,1)
if(Particles(i,1)>size(mask,2)-1)
Particles(i,1) = size(mask,2);
end
if(Particles(i,2)>size(mask,1)-1)
Particles(i,2) = size(mask,1);
end
if(Particles(i,1)<1)
Particles(i,1) = 1;
end
if(Particles(i,2)<1)
Particles(i,2) = 1;
end
end
end
% function Particles = pfresample(Particles)
% This function performs systematic re-sampling
% Inputs:
% Particles: structure
% mask: array
% centroids: array
% Outputs:
% Particles: structure
function Particles = pfresample(Particles,mask,centroids)
W = zeros(1,100)';
for i=1:size(Particles,1)
W(i) = mask(round(Particles(i,2)),round(Particles(i,1)));
end
if ~sum(W)
return;
end
W=W/sum(W);
cdf = cumsum(W);
M = size(Particles,1);
r_0 = rand / M;
for m = 1 : M
i = find(cdf >= r_0,1,'first');
Particles(m,1) = Particles(i,1);
Particles(m,2) = Particles(i,2);
r_0 = r_0 + 1/M;
end
end
% function Particles = pfCorrect(Particles)
% This function performs particle Correct
% move particles into centroid
% Inputs:
% Particles: structure
% centroid: array
% Outputs:
% Particles: structure
function Particles = pfCorrect(Particles, centroid)
Particles=repmat(centroid,size(Particles,1),1);
end
