Monovision based automated navigation and object detection

Paper proposes a new computer vision technique for automatic navigation and object detection. Automated navigation using a single still camera (mono-vision) where depth information is not available directly is a challenging task. Marking path instead of objects is the technique used here. This is based on human perception. Proposed `Next Path Method´ (NPM) uses pattern matching of the paths using cross-correlation which yields obstacle free traversal path. Object detection is performed by using proposed `sparse division´. Most of the objects are composed of pixels of similar values. Division of images based on similar pattern creates large number of tiny images. These are combined to form an object. The proposed algorithms were implemented on the hardware and were tested in varied and cluttered environments. We obtained satisfactory results in all the real-time experiments conducted.