Research On The All-day Stereo Vision Ranging

Stereo vision is an important branch of research in the field of computer vision.By using two cameras to capture the same scene and analyzing the target pixel deviation in both images,the distance between the target and the camera can be calculated according to the stereo ranging mathematical model and the internal and external parameters of the camera.Normal visible light cameras used in traditional stereo vision ranging are easily affected by illumination condition and may not even be able to perform ranging tasks at night.Infrared camera can solve this problem and be able to perform ranging tasks under bad lighting conditions,but it is restricted by the level of technology and manufacturing.Compared with visible images,infrared camera has poor imaging effect so further infrared images enhancement is essential.In this paper,the stereo vision ranging technology has been deeply studied,and the stereo vision ranging platform which can be used in day and night condition is designed and realized.In order to solve the contradiction between ranging precision and speed,this paper improves and innovates the camera calibration,image processing,template matching and ranging.Firstly,aiming at the particularity of infrared camera imaging effect,an improved infrared camera calibration method is proposed.The introduction of the classical Zhang's calibration method into the infrared camera calibration process without the aid of additional electronic components,the calibration process is simple and efficient,the calibration results are ideal.Secondly,in order to solve the shortcoming of serious time consumption in the region template matching method based on normalized cross-correlation,an adaptive normalized cross-correlation with randomness is proposed.Through the random access to image pixels,pixel region with tags and the limitation of multiple adaptive conditions,new algorithm greatly improves the matching speed while maintaining a high matching accuracy.Thirdly,aiming at solves the problem of low contrast,fuzzy and noise of infrared images,an improved infrared image enhancement algorithm is proposed,new algorithm includes adaptive dual-threshold imgae segmentation algorithm and adaptive non-linear histogram equalization algorithm.Former algorithm adaptively divides the infrared image into three parts through adaptive dual-threshold.Then the latter algorithm adjusts the grayscale of each part in different style,and through the non-linear curve fitting of piecewise grayscale mapping function to obtain the final enhanced infrared image.Finally,ranging results and error analysis are done,through the use of sub-pixel level matching to solve the problem of low ranging accuracy caused by pixel-level matching.Based on the works shown above,the day and night stereo vision ranging system is built.Hardware platform includes dual-visible cameras,infrared camera and communication equipment;the program coded by Matlab can realize the complete functions of stereo vision ranging,meanwhile algorithms and user interface are also realized.Finally,the experiment verifies the feasibility of the algorithm and the experimental platform.The experimental results show that the ranging distance of all-day stereo vision ranging system proposed in this paper is significantly improved compared with other stereo vision ranging platforms,and the ranging precision is ideal.