Depth hole filling and optimizing method based on binocular parallax image

Environment perception is one of the most vital function for autonomous robots while performing complex tasks in dynamic environment. In binocular stereo vision algorithm, calculation disparity image depends on matching and unmatched points form depth holes image. When encountering obstacles with less texture, such as blank walls, large area deep will often appear. Although existing filtering algorithm can solve problem small hole, it cannot be applied to hole. To this problem, a hole optimization proposed, which detects from globally, fills according value distribution, optimizes weighted least squares effect. The experimental results show that average time only 15 ms.