Projects
Path planning for formations of two robots with three degree of freedom
The aim of this project was to do planning for two robots with three degrees of freedom (position & orientation) from a point A to a point B in the environment with a constraint that both the robots have to be within certain distance of each other. In this project, the traditional A* algorithm was extended to include a third dimension in planning, the orientation. Also, the non-holonomicity of a differential drive system was been taken into account while planning for the robots. Project Report
Retargeting of 2D images
Retargeting of 2D images was done based on the method - "Seam Carving for Content-Aware Image Resizing" by Avidan et al. Retargeting means to resize an image while adapting to the image content and layout. Automatic as well as manually controlled retargeting was achieved. In manually controlled retargeting the user is asked to select a portion of the image which they want to preserve or erase in the image. Coding was done in Matlab.
Automatic Retargeting : Video1 Video2
Manual Retargeting : Video1
Automatic Face Morphing
A set of n number of images were taken. Feature detection in each image was performed based on the complexity map of the facial image. Feature matching was performed using the feature descriptors. Morphing sequence was calculated based on the similarity of images. Finally the images were morphed using TPS algorithm and a video was generated. Coding was done in Matlab.
Automatic Face Morphing Result : Video
Automatic 2D Image Stitching
Image stitching was achieved using Harris corners as interest points. Adaptive Non-Maxima suppression was used to minimize and uniformly distribute the number of points. A bias/gain normalized feature descriptor was extracted for each corner point, and then the descriptors of the images to be matched were compared. These matches were further refined with RANSAC and the best homography was selected to stitch the images together. Coding was done in Matlab.
Output : Final stitched image
Autonomous Robot Slalom
Objective of this project was to develop an Autonomous Robot system which traverses from the start point to the end gate while passing through each gate without touching any of the red coloured pylons. The relative distances between the different gates were arbitrary. The camera the only input to the system, was mounted on the bot. Live motion-Jpeg was transmitted from the camera to the Computer via wireless network. Motion of the bot from the current state was calculated in Matlab and transmitted to the bot as velocity signals via a different wireless network.
Sample Run : Video
Corner Detection in 2D images
In an alternate corner detection method (other than Harris corner), a window was scanned across the image and all the edge pixels were found in that window. A short line segment was traced through the edge pixels, using edge orientation estimated by corresponding gradients. The closest common intersection point to all these line segments was considered the best fit to local corner. Coding was done in Matlab.
Input Image : Image
Output : Image
Hand Gesture Recognition System for Industrial/Domestic Control
Image acquisition of hand gestures was done via USB port by MATLAB video handle. After thresholding and Image segmentation, features were extracted. Back propagation network was trained using input features and target vectors and resulting weight matrices were saved. These matrices were then used to test the real-time inputs during actual run. A 4 wheel cart was taken, which was controlled by 2 motors and a gear arrangement. Based on the classification performed by the neural network, control outputs were sent through serial port communication to a motor driver IC. Project Report
Calculation of Heart Risk Factor (HRF) through vital Cardiovascular Signal Processing
In this project, the P, QRS and T components of the heart wave were separated using QRST template formation and QRS and T cancellation technique. After isolating P, QRS and T components they were analysed separately. Heart signal processing and analysis was done with the help of Matlab. From the analysis of P, QRS and T components, the percentage of respective Fibrillation was derived. These were combined to calculate Heart Risk Factor and the Heart Risk to patient were shown on the Heart Risk Scale. Project Report