Research On Image Retrieval Method In Dark Light For Visual Location

In recent years,because of the rapid development of computer technology,the research of robot is more in-depth.Mobile robot mainly solves three problems: location,destination and path planning.The first problem is related to robot positioning.In the research of robot localization,most scholars use multi-sensor fusion.Because of the low cost of camera and rich perceptual information,most intelligent devices use visual positioning.In the field of visual location,visual location methods are divided into two categories according to whether a priori visual map is used for location.There are many technologies involved in the localization method based on prior visual map.Image retrieval technology is to find the most similar image to the current robot location in the prior map to match the current robot coordinates.However,in some special scenes,such as dark illumination,the captured images are dark and poor quality,so that the image retrieval accuracy is not high and cannot be accurately located.Image retrieval in dark light is an important problem to be solved in the field of visual location.Based on the previous research of image retrieval technology,this paper studies the common problem of image retrieval in dark light,and proposes an image retrieval algorithm with high accuracy in dark light.The algorithm proposed in this paper uses the convolution neural network method which is commonly used in the field of image retrieval in recent years,and proposes an innovative image retrieval framework for the special dark light scene.It integrates the illumination enhancement method in the field of image processing,designs the illumination enhancement network for image retrieval,and designs two kinds of descriptor extraction methods in image retrieval The two loss functions are compared and trained by the combination of illumination enhancement and image retrieval.Compared with the existing image retrieval methods,the retrieval accuracy is greatly improved in dark light.In order to verify the retrieval accuracy of the dark light image retrieval algorithm and the feasibility of practical application,this study is verified on the public data set and the actual campus data set.On the public data set used in this paper,compared with the research results of other image retrieval,the algorithm designed in this paper has higher accuracy in dark light image retrieval.At the same time,this paper also validates the method of whether to add lighting processing to the designed network,and obtains the network framework of adding lighting information,which effectively improves the accuracy of dark light image retrieval.In this project,the actual campus scene data set is verified,and the more accurate image retrieval results are obtained.The experimental results show that the innovative dark light image retrieval network framework proposed in this paper has high practicability for image retrieval under dark light conditions.The research of this topic plays a positive role in promoting the development of dark light image retrieval.