Projects-Image Processing

• Image Processing in Road Traffic Analysis :   
• 1.   Traffic Signal Monitoring System: Nowadays in many metropolitan cities, traffic is the main problem. As traffic controller is based on the principle of timer rather than vehicles density, so it increases the number of accidents. This project addresses the application of digital image processing techniques in Traffic Surveillance and provides a solution for the above problems where in it measures the relative density of cars from all the sides of road (junction)and sends the signal and also controls the delay of traffic lights  depending on the density. The project uses the image processing techniques such as lane masking, background elimination, edge detection, contour linking, contour labeling, contour estimation and depending on the number of cars, indication is sent to the traffic signal and particular time is given according to its number. These algorithms are implemented using DevC++ with Open Source Computer Vision library. 
• 2. Traffic Density and Detection of Vehicles:  In the present days of freeways and express highways it is very difficult for traffic authorities to control the traffic and maintain accurate statistics of the same. Here we propose a novel method that detects vehicles, counts them and classifies them as cars, buses and trucks on the fly. The statistics obtained from the above analysis can be used to reduce congestion on roads, plan the roads more effectively and provide accurate statistics of the traffic situation on that road. We assume that the camera is fixed at an elevated position near the highway such that the vehicles are coming towards the camera. First we detect the vehicles in the video and put bounding boxes around them. Then once the vehicle crosses a predefined boundary we increment the count of vehicles by 1 and classify the vehicle as car, bus or truck and it is done by SURF Features and KMeans Algorithm.

• Image Inpainting: It is the conversion of the any given image to its respective paint. It is implemented using DevC++ with OpenCV library. The algorithm is involves the geometric transformation of an image in one direction, taking the average of the input and transformed images and quantizing to the certain levels to get painted image. Such application is used in tools like Gimp, Picasa, etc.