|
|
how can i provide real time video input
Posted:
Mar 28, 2012 3:59 AM
|
|
%% Tracking Cars Using Gaussian Mixture Models
% This demo illustrates how to detect cars in a video sequence
% using foreground detection based on gaussian mixture models (GMMs).
% Copyright 2004-2010 The MathWorks, Inc.
% $Revision: 1.1.6.1 $ $Date: 2010/11/22 03:06:06 $
%% Introduction
% This demo illustrates the use of gaussian mixture models to detect
% foreground in a video. After foreground detection, the demo process
% the binary foreground images using blob analysis. Finally,
% bounding boxes are drawn around the detected cars.
%% Initialization
% Use these next sections of code to initialize the required variables and
% System objects.
%%
% Create a System object to read video from a binary file.
hbfr = vision.BinaryFileReader('Filename', 'viptraffic.bin');
%%
% Create a System object to upsample the chrominance components of the
% video.
hcr = vision.ChromaResampler(...
'Resampling', '4:2:0 (MPEG1) to 4:4:4', ...
'InterpolationFilter', 'Pixel replication');
%%
% Create color space converter System objects to convert the image from
% YCbCr to RGB format.
hcsc = vision.ColorSpaceConverter('Conversion', 'YCbCr to RGB');
%%
% Create a System object to detect foreground using gaussian mixture models.
hof = vision.ForegroundDetector(...
'NumTrainingFrames', 5, ... % only 5 because of short video
'InitialVariance', (30/255)^2); % initial standard deviation of 30/255
%%
% Create a blob analysis System object to segment cars in the video.
hblob = vision.BlobAnalysis( ...
'CentroidOutputPort', false, ...
'AreaOutputPort', true, ...
'BoundingBoxOutputPort', true, ...
'OutputDataType', 'single', ...
'NumBlobsOutputPort', false, ...
'MinimumBlobAreaSource', 'Property', ...
'MinimumBlobArea', 250, ...
'MaximumBlobAreaSource', 'Property', ...
'MaximumBlobArea', 3600, ...
'FillValues', -1, ...
'MaximumCount', 80);
%%
% Create and configure two System objects that insert shapes, one for
% drawing the bounding box around the cars and the other for drawing the
% motion vector lines.
hshapeins1 = vision.ShapeInserter( ...
'BorderColor', 'Custom', ...
'CustomBorderColor', [0 255 0]);
hshapeins2 = vision.ShapeInserter( ...
'Shape','Lines', ...
'BorderColor', 'Custom', ...
'CustomBorderColor', [255 255 0]);
%%
% Create and configure a System object to write the number of cars being
% tracked.
htextins = vision.TextInserter( ...
'Text', '%4d', ...
'Location', [0 0], ...
'Color', [255 255 255], ...
'FontSize', 12);
%%
% Create System objects to display the results.
sz = get(0,'ScreenSize');
pos = [20 sz(4)-300 200 200];
hVideoOrig = vision.VideoPlayer('Name', 'Original', 'Position', pos);
pos(1) = pos(1)+220; % move the next viewer to the right
hVideoFg = vision.VideoPlayer('Name', 'Foreground', 'Position', pos);
pos(1) = pos(1)+220;
hVideoRes = vision.VideoPlayer('Name', 'Results', 'Position', pos);
line_row = 22; % Define region of interest (ROI)
%% Stream Processing Loop
% Create a processing loop to track the cars in the input video. This
% loop uses the previously instantiated System objects.
%
% When the BinaryFileReader object detects the end of the input file, the loop
% stops.
while ~isDone(hbfr)
[y, cb, cr] = step(hbfr); % Read input video frame
[cb, cr] = step(hcr, cb, cr); % Upsample chroma to construct YCbCr 4:4:4
image = step(hcsc, cat(3,y,cb,cr)); % Convert image from YCbCr to RGB
% for display purposes
% Remove the effect of sudden intensity changes due to camera's
% auto white balancing algorithm.
y = im2single(y);
y = y-mean(y(:));
fg_image = step(hof, y); % Foreground
% Estimate the area and bounding box of the blobs in the foreground
% image.
[area, bbox] = step(hblob, fg_image);
Idx = bbox(1,:) > line_row; % Select boxes which are in the ROI.
% Based on dimensions, exclude objects which are not cars. When the
% ratio between the area of the blob and the area of the bounding box
% is above 0.4 (40%) classify it as a car.
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 detected cars.
y2 = step(hshapeins1, image, bbox);
% Display the number of cars tracked and a white line showing the ROI.
y2(22:23,:,:) = 255; % White line
y2(1:15,1:30,:) = 0; % Black background for displaying count
image_out = step(htextins, y2, count);
step(hVideoOrig, image); % Original video
step(hVideoFg, fg_image); % Foreground
step(hVideoRes, image_out); % Bounding boxes around cars
end
% Close the video file
release(hbfr);
%% Summary
% The output video displays the bounding boxes around the cars. It also
% displays the number of cars in the upper left corner of the video.
displayEndOfDemoMessage(mfilename)
Hi, The file loads a already present *.avi, or *.bin file to perform tracking....how can i make it in real time by providing video input?
I even tried giving 'videoinput' function....but its not supported by computer vision. Correct me if i'm wrong. Please suggested me, alternate to make this run real time.
Thanks.
|
|
