A New Motion Obstacle Detection Based Monocular-Vision Algorithm

As for real-time changes in the illumination and external environment in indoor structured environment, this paper proposed an advanced moving obstacle detection algorithm based on monocular vision. It cut the image with a monocular camera, did the gray scale conversion and median filtering, combined the background subtraction method of Gaussian mixture model and inter-frame subtraction method, then conducted morphological processing and connectivity analysis on binarized difference image with the area information included, which effectively solved the hole problem, and obtained complete information on moving targets. The result showed that this method can quickly and accurately detect moving obstacles within the region observed, of good real time performance and robustness, able to adapt to light and other changes in external conditions.