A speed adaptive ego-motion detection system using EDGE-histograms produced by variable graduation method

An ego-motion detection system has been developed, where edge features are extracted from an image and analyzed to detect any motion from the camera holder instead of using the conventional method of comparing pixel intensities. Such edge-based feature representation scheme reduces the computational complexity and increases the accuracy, thus being better suited to hardware implementation due to its simplicity. In this paper we have enhanced the reliability and flexibility of the system by introducing a new pre-processing scheme in edge detection and an automatic speed adaptation capability in local motion detection. The pre-processing improves the local motion detection accuracy by only highlighting the apparent general contour edges, while filtering out insignificant features in the background which may lead to misjudgement. The automatic speed adaptation capability improves the system and renders it more flexible to accommodate to more complex motion patterns with variable speeds. The system performance has been demonstrated by simulation experiments and the robustness against disturbing moving objects in the scene has also been shown.