Research On Depth Level Of Robot Vision Based On Gray Gradient

The traditional method of robot operation is to follow the predetermined program after programming,and its forward path is a fixed path,and the position of the target object is also fixed.In the actual production process,due to various reasons,it is difficult to ensure that all the target objects are neatly arranged on the production line.If the offset distance of the target object is too large or other parameters change,it will hinder the robot to grasp it.Therefore,the production flexibility of the traditional programming control production line is relatively poor.In order to solve this problem,a method of using machine vision to judge the depth level of objects is proposed to replace the traditional programming for production control.In this paper,we use gauge blocks as objects on the actual production line to explore the depth changes of objects with different depths in the image.(1)Image acquisition.The image acquisition equipment adopts DH-HV1351UM-M compact USB interface CMOS digital camera.In order to avoid light interference,the acquisition is mainly carried out in darkroom.Different spectral light sources and different angle light sources were selected for the control experiment,and the distance of gauge block was adjusted to obtain different depth images.(2)Image preprocessing.Because of the external factors and the equipment's own conditions,the image will appear noise or the required image feature area is not obvious,so the image should be preprocessed first after obtaining the image.In this paper,the content of preprocessing includes image denoising and image enhancement.The former is to remove the noise in the image,and the latter is to enhance the characteristics of the target area,which is convenient for the analysis of the features behind the image.Since there is no general de-noising or enhancement algorithm,this paper selects the appropriate algorithm by comparing the effects of different algorithms.(3)Image depth information extraction.Image depth information extraction mainly includes two parts,which are the introduction of the existing depth estimation theory and the depth judgment based on the change of gray gradient in the image.The existing depth information extraction theory mainly introduces the passive depth estimation method.Then,the algorithm is designed for this topic.Assuming that the distribution of each pixel conforms to the Markov property,K-means clustering algorithm is used as the initialization tag to calculate the probability of each pixel belonging to a tag,and the depth level of the image is divided according to the change of gray level.Through the experiment to obtain the image and analyze,we can get the result of the division according to the change of gray level,analyze and count the number of pixels in the image of different depth levels and the gray distribution of different depth in this case.The experiment proves that the error of depth level measurement data judged by gray level change is small,and it is easy to install and use.It is suitable to solve the problem of insufficient production flexibility in the traditional production line based on programming control.