how to count number of cars,bikes in an image using matlab

hi this ravi,

currently i am working on the project to count number of cars and bikes in an image using matlab. I have done it by finding area of the vehicle but it is not applicable for all images. So any one please help me.

input is the image that has vehicles like cars,bikes.

output must be number of cars and number of bikes in input image

thanks in advance

Answer 1

[Image Analyst]

This is far too challenging a project to be answered in a forum such as this. With different angles of view of the bicycle, with partial or total occlusions by cars, poles, people, etc., with bicycles entering and leaving the field of view, and with different configurations of bicycles (cruiser, racing, tricycle, with and without racks or trailers, etc.), this could be a Ph.D. project. What if all you see if half of the front tire sticking in from the edge of the image, or all you see is the bicyclists head and handlebars sticking above a car or wall? Good luck with it though.

[ravi]

Tell me is there any methods to identify the car or bike in an image and to draw the rectangle over that car or bike for finding the area of the car or bike.

thanks in advance

[Image Analyst]

Well it sounds like you've already seen the car detection and tracking capability of the Computer Vision System Toolbox on this page. If that is not robust and accurate enough for you then you will need to embark on a comprehensive improvement of it. Of course this will be tens of thousands of lines of code - far too much to consult with you on here in the Answers forum. I also assume you've gone over the papers published on this topic here in the tracking section of Vision Bib. Those are people who have devoted a good amount of time to solving problems like yours. If you can't find, or adapt, an algorithm from there, then perhaps you can hire one of those groups to develop the application for you.

Answer 2

[Anand]

Here's an approach you could try:

1. Use the trainCascadeObjectDetector on a set of labelled training data to train a detector that can detect cars. Do the same for bikes. Use <http://www.mathworks.com/help/vision/ug/train-a-cascade-object-detector.html> page as a guide to help training your detectors.

2. Once you've trained the two detectors, use the vision.cascadeObjectDetector to detect cars and bikes in the image. This will give you bounding boxes at expected locations of the cars and bikes.

Note that like Image Analyst mentioned, this is an involved problem and will need a lot of work beyond blindly following this approach.

Good luck!

[ravi]

hi Anand,

The link you has given as a guide is not opening giving some error while loading the page. So please give me the correct url to refer it will be useful to me.

Thanks Anand.

[Anand]

http://www.mathworks.com/help/vision/ug/train-a-cascade-object-detector.html

[maria ahmed]

if i just want to count the no cars in an image then what is the code?

[Image Analyst]

Code is probably given in the examples/demos that come with the Computer Vision System Toolbox: http://www.mathworks.com/products/computer-vision/description5.html

[NAGESWARARAO NAIK]

Thanks for your response, here they used boxLabeling, but i need to use pixelLabel to train.