Object Detection and Depth Estimation of Real World Objects using Single Camera

This research paper proposes a single camera based depth estimation technique. The proposed technique takes images of walls in a room and detects objects of interest in a cluttered environment. Having detected different objects in a room the proposed technique calculates their areas. Based on training data and polynomial curve fitting approach the proposed technique estimates the distance of the camera from the objects. For a real world object one can determine a fixed equation which can then be used to find any random distance. The approach is efficient and can effectively be applied to any indoor navigation or motion planning algorithm. Based on the estimated distances from different objects the proposed algorithm estimates the accurate location of the camera (mounted on a robot) in a room. For detection we have used template matching technique. Algorithm compares the reference template with the objects of interest in a cluttered environment by using SURF (speeded up robust features). The proposed algorithm is tested on real world images and compared with the existing depth estimation techniques.