Technology and Application of Water Target Ranging Based on Vision
Binocular vision has the advantages of high short-range ranging accuracy and rich image content.It is currently a research hotspot in the field of unmanned driving.Accurately acquiring the parallax of binocular images is the key technology of bin-ocular vision ranging.However,the water surface ranging faces irregular target shaking,which increases the scale and brightness difference of the water target between the binocular images,which is easy to cause ambiguity in matching.In order to improve the ac-curacy of stereo matching,a method combining Yolov5 for object segmentation and an improved SIFT feature matching approach are proposed.By extracting and matching the affine invariant feature points of the target in the nonlinear scale space,the image par-allax can be accurately measured,and then the accurate target distance can be obtained.Experiments show that the improved model SIFT can obtain the parallax value of binocular image very accurately,the accuracy of the model is 97.1%,the matching time is on-ly 1.3 s,which is 10.4%higher than the original SIFT and the matching time is shortened by 4 s,which is 11.3%higher than that of the model SURF,and the matching time is shorted by 0.5 s.
water targetbinocular rangingfeature extractionSIFTYolov5
万方数据
维普
