COMPUTER VISION BASED PROJECTS
AR TAG DETECTION
The focus of this project is detecting a custom AR Tag (a form of fiducial marker), that is used for obtaining a point of reference in the real world, such as in augmented reality applications. There are two aspects to using an AR Tag, namely detection and tracking, both of which is implemented in this project. The detection stage involves finding the AR Tag from a given image sequence while the tracking stage involves keeping the tag in “view” throughout the sequence and performing image processing operations based on the tag’s orientation and position.
LANE DETECTION
In this project, we aimed to do simple Lane Detection to mimic Lane Departure Warning systems used in Self Driving Cars. The task was to design an algorithm to detect lanes on the road, as well as estimate the road curvature to predict car turns using histogram.
COLOR SEGMENTATION USING GMM
This project focused on the concept of color segmentation using Gaussian Mixture Models and Expectation Maximization techniques. The video sequence provided was captured underwater and showed three buoys of different colors, namely yellow, orange and green. The aim of this project was to obtain a tight segmentation of each buoy for the entire video sequence by applying a tight contour (in the respective color of the buoy being segmented) around each buoy.
LUCAS-KANADE (LK) TRACKER
In this project we implement the Lucas-Kanade (LK) template tracker for three video sequences from the Visual Tracker benchmark database: featuring a car on the road, a human walking, and a box on a table. The aim was to initialize the tracker by drawing a bounding box around the object to be tracked in the first frame of the video and for each of the subsequent frames, the tracker updated an affine transform that warped the current frame so that the template in the first frame was aligned with the warped current frame.
VISUAL ODOMETRY
Visual Odometry is a crucial concept in Robotics Perception for estimating the trajectory of the robot.
The concepts involved in Visual Odometry are quite the same for SLAM which needless to say is an integral part of Perception. In this project we were given frames of a driving sequence taken by a camera in a car, and the scripts to extract the intrinsic parameters.
We implement the different steps to estimate the 3D motion of the camera, and provide as output a plot of the trajectory of the camera.
TRAFFIC SIGN RECOGNITION
In this project, we aimed to do Traffic Sign Recognition, we performed the two steps of detection and recognition using existing OpenCV code (HOG feature detector, MSER feature detector, SVM routines) to create the complete pipeline. The challenge was in tuning the system to detect well.
