Base Paper on Estimation of Vehicle Orientation Using Stereo Vision

File Size:
1.54 MB
Volume 2, Issue 9 (September, 2016)
Publication No:
Aarti Soni, Dr. Raman Chadha, Sonam Soni
9 x

Nowadays, image processing is among rapidly growing technologies. Image processing basically includes the following three steps: Importing the image via image acquisition tools, Analyzing and manipulating the image, Output in which result can be altered image or report that is based on image analysis. In this paper we are going to estimate the pose of vehicle with stereovision. With stereo vision, we can see where things are in relation to our own bodies with much greater harshness --especially when those objects are moving toward or away from us in the depth component. Stereoscopic vision is what gives us the ability to see things with height, width, and depth.The estimation of attitude of a vehicle closely depends on the 3D correspondences or triangulation. This implies obtaining the correct estimate of the 3D point in space, which are used to obtain the correct estimate of the attitude. Hence an effort was made in order to minimize the triangulation error further.
A video sequence from a stereoscopic system, image features are detected and matched for the initial stereo pair, then reconstructed into 3D space and stored as 3D landmarks after outlier elimination. After that key features are tracked in the subsequent images and used to estimate the camera motion. Considering the two algorithms for which the optimization is performed, it has been found that the triangulation is far better than the disparity method. Observing the results, we may conclude that the problem of local minima has affected the output that has led to the convergence of the method onto some other point.

Indexed Terms:
Stereo vision, Triangulation, Visual Odometry, RANSAC, Vehicle Orientation