Scribd
Upload
31 views

Tracking Using Optical Flow

This document contains code for analyzing video data to detect and track moving objects across multiple frames. It uses techniques like optical flow estimation, thresholding, morphological operations, and blob analysis to identify moving regions and bounding boxes around objects. The detected objects and motion vectors are displayed on separate video players for visualization and analysis.

Uploaded by

haris
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
31 views

Tracking Using Optical Flow

This document contains code for analyzing video data to detect and track moving objects across multiple frames. It uses techniques like optical flow estimation, thresholding, morphological operations, and blob analysis to identify moving regions and bounding boxes around objects. The detected objects and motion vectors are displayed on separate video players for visualization and analysis.

Uploaded by

haris
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 3

hbfr = video.BinaryFileReader('Filename', 'viptraffic.

bin'); % read the binary


video

hcr = video.ChromaResampler(...
'Resampling', '4:2:0 (MPEG1) to 4:4:4', ...
'InterpolationFilter', 'Pixel replication');

hcsc1 = video.ColorSpaceConverter('Conversion', 'YCbCr to RGB');


hcsc2 = video.ColorSpaceConverter('Conversion', 'RGB to intensity');

hidtc = video.ImageDataTypeConverter('OutputDataType', 'single');

hof = video.OpticalFlow( ...


'OutputValue', 'Horizontal and vertical components in complex form', ...
'ReferenceFrameDelay', 3);

hmean1 = video.Mean;
hmean2 = video.Mean('RunningMean', true);

hmedianfilt = video.MedianFilter2D;

hclose = video.MorphologicalClose('Neighborhood', strel('line',5,45));


hblob = video.BlobAnalysis( ...
'CentroidOutputPort', false, ...
'AreaOutputPort', true, ...
'BoundingBoxOutputPort', true, ...
'OutputDataType', 'double', ...
'NumBlobsOutputPort', false, ...
'MinimumBlobAreaSource', 'Property', ...
'MinimumBlobArea', 250, ...
'MaximumBlobAreaSource', 'Property', ...
'MaximumBlobArea', 3600, ...
'FillValues', -1, ...
'MaximumCount', 80);

herode = video.MorphologicalErode('Neighborhood', strel('square',2));

hshapeins1 = video.ShapeInserter( ...


'BorderColor', 'Custom', ...
'CustomBorderColor', [0 1 0]);
hshapeins2 = video.ShapeInserter( ...
'Shape','Lines', ...
'BorderColor', 'Custom', ...
'CustomBorderColor', [255 255 0]);

htextins = video.TextInserter( ...


'Text', '%4d', ...
'Location', [0 0], ...
'Color', [1 1 1], ...
'FontSize', 12);

hVideo1 = video.VideoPlayer('Name', 'Original Video');


hVideo1.Position(1) = round(0.4*hVideo1.Position(1)) ;
hVideo1.Position(2) = round(1.5*(hVideo1.Position(2))) ;
hVideo1.Position([4 3]) = [200 200];

hVideo2 = video.VideoPlayer('Name', 'Motion Vector');


hVideo2.Position(1) = hVideo1.Position(1) + 350;
hVideo2.Position(2) =round(1.5* hVideo2.Position(2));
hVideo2.Position([4 3]) = [200 200];

hVideo3 = video.VideoPlayer('Name', 'Thresholded Video');


hVideo3.Position(1) = hVideo2.Position(1) + 350;
hVideo3.Position(2) = round(1.5*(hVideo3.Position(2))) ;
hVideo3.Position([4 3]) = [200 200];

hVideo4 = video.VideoPlayer('Name', 'Results');


hVideo4.Position(1) = hVideo1.Position(1);
hVideo4.Position(2) = round(0.3*(hVideo4.Position(2))) ;
hVideo4.Position([4 3]) = [200 200];

% Initialize some variables used in plotting motion vectors.


MotionVecGain = 20;
line_row = 22;
borderOffset = 5;
decimFactorRow = 5;
decimFactorCol = 5;
firstTime = true;

while ~isDone(hbfr)
[y, cb, cr] = step(hbfr); % Read input video frame
[cb, cr] = step(hcr, cb, cr);
imrgb = step(hcsc1, cat(3,y,cb,cr)); % Convert image from YCbCr to RGB
image = step(hidtc, imrgb); % Convert image to single
I = step(hcsc2, image); % Convert color image to intensity
of = step(hof, I); % Estimate optical flow

% Thresholding and Region Filtering.


y1 = of .* conj(of);
% Compute the velocity threshold from the matrix of complex velocities.
vel_th = 0.5 * step(hmean2, step(hmean1, y1));

% Threshold the image and then filter it to remove fine speckle noise.
filteredout = step(hmedianfilt, y1 >= vel_th);

% Perform erosion operation to thin-out the parts of the road followed


% by the closing operation to remove gaps in the blobs.
th_image = step(hclose, step(herode, filteredout));

% Regional Filtering.

% Estimate the area and bounding box of the blobs in the thresholded
% image.
[area, bbox] = step(hblob, th_image);
% Select those boxes which are in our ROI.
Idx = bbox(1,:) > line_row;

% The next lines of code exclude other objects (like parts of the road)
% which are also segmented as blobs and select only cars. When the
% ratio between the area of the blob and the area of the bounding box
% is above 0.4 (40%), it is considered as a car and hence the bounding
% box for that object is used. Otherwise the bounding box is removed.
ratio = zeros(1, length(Idx));
ratio(Idx) = single(area(1,Idx))./single(bbox(3,Idx).*bbox(4,Idx));
ratiob = ratio > 0.4;
count = int32(sum(ratiob)); % Number of cars
bbox(:, ~ratiob) = int32(-1);

% Draw bounding rectangles around the tracked cars.


y2 = step(hshapeins1, image, bbox);

% Display the number of cars tracked and a white line showing ROI.
y2(22:23,:,:) = 1; % The white line.
y2(1:15,1:30,:) = 0; % Background for displaying count
image_out = step(htextins, y2, count);

% Generate the coordinate points for plotting motion vectors.


if firstTime
[R C] = size(of); % Height and width in pixels
RV = borderOffset:decimFactorRow:(R-borderOffset);
CV = borderOffset:decimFactorCol:(C-borderOffset);
[Y X] = meshgrid(CV,RV);
firstTime = false;
end

% Calculate and draw the motion vectors.


tmp = of(RV,CV) .* MotionVecGain;
lines = [X(:)';Y(:)';X(:)' + imag(tmp(:))';Y(:)' + real(tmp(:))'];
mv_video = step(hshapeins2, image, lines);

step(hVideo1, image); % Display Original Video


step(hVideo2, mv_video); % Display video with motion vectors
step(hVideo3, th_image); % Display Thresholded Video
step(hVideo4, image_out); % Display video with bounding boxes
end

You might also like