Gaussian mixed model-based motion detection and shadow elimination algorithm research

摘要

In computer vision field, regarding sequence image’s moving target detection is one of its researches important orientations; it has wide research prospects in each field of life. On this basis, the paper goes deeper analysis of complicated scenes moving target detection, provides Gaussian model improvement forms, applies fixed learning rate to learn variance, and sets up lower limit threshold value, targeted at the new type algorithm, according to different confusion scope, it adopts different updating ways, finally by experiment verification, we can get new type algorithm handling quality and speed are obviously faster than traditional algorithm. Combine Gaussian mixed model with HSV color space shadow elimination method, and modifies Gaussian mixed model’s parameters, let its shadow elimination efficiency to be greatly promoted, and gets that shadow elimination method purely carrying on in HSV color space will appear great deviation, while adopt Gaussian mixed model learning way to combine with HSV color space shadow elimination ways then it will get closer to practice, so the paper proposed algorithm has good effectiveness and timeliness.