Research On An Object Depth Measurement Method For Monocular Vision

With the continuous development of computer vision technology and upgrading of digital imaging equipment,image depth measurement methods are widely used in the fields of intelligent robots,transportation assistance,3D modeling and 3D video production.By contrast laser,infrared light,ultrasonic,binocular vision and monocular vision depth measurement methods,the passive depth measurement method of monocular vision has become a research hotspot because of its simple operation,low cost and small space and load.This method has not only theoretical significance,but also practical value.This paper mainly uses the feature of the target object in the image to obtain the absolute depth information,including the following modules.Image preprocessing operation based on grayscale linear transformation,smoothing and improved LBF model image segmentation.By fusing the improved Harris algorithm with SIFT algorithm,the Harris-SIFT algorithm is proposed and used to extract the feature information of the target object in the image.Using the convex hull principle,the linear segment is selected to calculate the image scaling rate to obtain the object depth information.This paper optimizes the research content from the following two points:1)Improving the image segmentation algorithm of LBF model.Obtaining the properties of Gaussian function as kernel function by analyzing the standard LBF model,the complex function with small computation is used to simulate the Gaussian function so as to improve the evolution speed of the level set.The image enhancement operator,is introduced to correct,reduce noise and enhance contrast,so as to improve the segmentation efficiency of the parts with similar gray values in the image.Experimental results show that the improved LBF model in this paper has fewer iterations than the original model in image segmentation,thus reducing the segmentation time and improving the performance of the algorithm.2)By fusing the improved Harris algorithm and the SFIT algorithm,the Harris-SIFT algorithm suitable for this paper is proposed.On one hand,feature points extracted by Harris algorithm in multi-scale space are used to replace feature point detection in SIFT algorithm,which effectively avoids detecting redundant feature points by SIFT algorithm.On the other hand,the feature descriptors of SIFT corner detection algorithm are reconstructed to improve the accuracy and efficiency of the algorithm by reducing the dimension of the descriptor and distinguishing the pixels with the same gradient value but different gray value.In this paper,the improved Harris-SIFT corner detection algorithm is used to extract the feature information of the target object in the image,to obtain the depth information.Testing multiple methods for depth measurement of multiple objects,the error rate between the actual distance and the measured distance by this method is less than 3.5%,which is better than the other methods.Experimental results show that the depth measurement method based on monocular vision can accurately measure the depth information of objects in a short distance.