Vehicle Tracking Model

MATLAB add-on for vehicle detection using a pre-trained SVM model with HOG features. Simple, accurate, and easy to integrate into projects.

This MATLAB add-on enables vehicle detection in images using a pre-trained Support Vector Machine (SVM) classifier. The model is trained using Histogram of Oriented Gradients (HOG) features, providing high accuracy and reliable performance. With just a few lines of code, you can easily detect vehicles in images of various formats, including .jfif.

The add-on includes:

• A pre-trained SVM classifier for vehicle classification.
• A simple interface for image detection.
• Customization options for retraining the model with your own dataset.

Ideal for applications in autonomous driving, smart cities, or image analysis, this tool simplifies the integration of vehicle detection into your MATLAB projects.

After downloading unzip the file. Then detect vehicle from image with the following code:

load('vehicleClassifier.mat');
testImage=imread('testImage.jpg'); %Replace image with actual image
targetSize=[128 128];
testImageResized=imresize(testImage,targetSize);
if size(testImageResized,3)==3
    testImageGray=rgb2gray(testImageResized);
else
    testImageGray=testImageResized;
end
testFeatures=extractHOGFeatures(testImageGray,'CellSize',cellSize);
predictedLabel=predict(classifier,testFeatures);
disp(['The predicted class is: ',char(predictedLabel)]);
figure;
imshow(testImage);
title(['Predicted class: ', char(predictedLabel)]);

To detect vehicle from video:

% Load the trained classifier and parameters
load('vehicleClassifier.mat'); % Contains 'classifier', 'cellSize', 'targetSize'

% Specify the path to your video file
videoFile = 'testVideo.mkv'; % Replace with your video file path

% Create a VideoReader object
videoReader = VideoReader(videoFile);

% Create a VideoWriter object to write the processed video
outputVideoFile = 'processed_video_with_scoreboard.mp4'; % Output video file name
videoWriter = VideoWriter(outputVideoFile, 'MPEG-4');
open(videoWriter);

% Initialize vehicle counts
vehicleCounts = struct('Bus', 0, 'Car', 0, 'Motorbike', 0,'CNG',0,'Rickshaw',0,'Truck',0);

% Define scoreboard position and style
scoreboardPosition = [10, 10]; % Top-left corner of the scoreboard
scoreboardFontSize = 14;
scoreboardBoxColor = 'yellow';
scoreboardTextColor = 'black';

% Process each frame
while hasFrame(videoReader)
    % Read the next frame
    frame = readFrame(videoReader);
    
    % Convert the frame to grayscale
    grayFrame = rgb2gray(frame);
    
    % Enhance contrast if necessary
    grayFrame = histeq(grayFrame);
    
    % Detect edges using the Canny method
    edges = edge(grayFrame, 'Canny');
    
    % Dilate edges to connect adjacent edges
    se = strel('rectangle', [5, 5]);
    dilatedEdges = imdilate(edges, se);
    
    % Fill holes to create solid shapes
    filledRegions = imfill(dilatedEdges, 'holes');
    
    % Remove small objects (noise)
    cleanImage = bwareaopen(filledRegions, 1000); % Adjust threshold as needed
    
    % Label connected components
    [labelsMatrix, numObjects] = bwlabel(cleanImage);
    
    % Measure properties of image regions
    stats = regionprops(labelsMatrix, 'BoundingBox', 'Area');
    
    % Loop through each detected region
    for i = 1:numObjects
        % Filter out regions that are too small or too large
        if stats(i).Area > 2000 % Adjust area threshold as needed
            bbox = stats(i).BoundingBox;
            
            % Crop the detected region from the grayscale frame
            detectedRegion = imcrop(grayFrame, bbox);
            
            % Resize to match the size used during training
            detectedRegion = imresize(detectedRegion, targetSize);
            
            % Extract HOG features
            hogFeatures = extractHOGFeatures(detectedRegion, 'CellSize', cellSize);
            
            % Predict the class label
            predictedLabel = predict(classifier, hogFeatures);
            labelStr = char(predictedLabel);
            
            % Increment the corresponding count
            if isfield(vehicleCounts, labelStr)
                vehicleCounts.(labelStr) = vehicleCounts.(labelStr) + 1;
            else
                vehicleCounts.(labelStr) = 1;
            end
            
            % Draw bounding box and label on the frame
            frame = insertShape(frame, 'Rectangle', bbox, 'Color', 'green', 'LineWidth', 2);
            frame = insertText(frame, [bbox(1), bbox(2)-15], labelStr, 'FontSize', 12, 'BoxColor', 'yellow', 'TextColor', 'black');
        end
    end
    
    % Update the scoreboard text
    scoreboardText = sprintf('Bus: %d\nCar: %d\nMotorbike: %d\nCNG: %d\nRickshaw: %d\nTruck: %d', ...
        vehicleCounts.Bus, vehicleCounts.Car, vehicleCounts.Motorbike, vehicleCounts.CNG, vehicleCounts.Rickshaw, vehicleCounts.Truck);
    
    % Add the scoreboard to the frame
    frame = insertText(frame, scoreboardPosition, scoreboardText, ...
        'FontSize', scoreboardFontSize, ...
        'BoxColor', scoreboardBoxColor, ...
        'TextColor', scoreboardTextColor, ...
        'BoxOpacity', 0.6);
    
    % Display the processed frame
    imshow(frame);
    drawnow;
    
    % Write the processed frame to the output video
    writeVideo(videoWriter, frame);
end

% Close the video writer
close(videoWriter);

% Display the vehicle counts
disp('Total Vehicle Counts:');
disp(vehicleCounts);

This is primary versionof our model . We just detect with Bus, Car, CNG , Motorbike, Rickshaw, Truck with our model. But you can detect more type by adding folder like "Ambulance" or etc and download image of ambulance in the folder. Then run the dataStore.m file and your model will be updated. You can reduce the error of detecting vehicle by adding more sample image in respective folder.

In the time of running dataStore.m you may face some error . Here are some prerequisite to run smoothly

• MATLAB (Version 2020a or later): Ensure you are using a compatible version of MATLAB.
• Required Toolboxes:

Computer Vision Toolbox: For image processing, feature extraction, and visualization.

Statistics and Machine Learning Toolbox: Required for the fitcecoc function to train the SVM classifier.

License

This project is licensed under the MIT License.

MIT License

Copyright (c) 2024 - MD. Salehin Islam

Permission is hereby granted, free of charge, to any person obtaining a copy

of this software and associated documentation files (the "Software"), to deal

in the Software without restriction, including without limitation the rights

to use, copy, modify, merge, publish, distribute, sublicense, and/or sell

copies of the Software, and to permit persons to whom the Software is

furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all

copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE

SOFTWARE.

To reach out me:-

website: https://mdsalehinislam.netlify.app/

facebook: https://www.facebook.com/mohammad.salehin.islam.27

github: https://github.com/SALEHINISLAM

Thanks

Salehin

from Bangladesh