Research Of The Field Vision Navigation Line Extraction Algorithm Based On Close And Distant View Images

The application of intelligent equipment in agriculture is becoming more and more common with the improvement of science and technology.In order to make intelligent equipment operate in the field according to the expected path,the precision of navigation line extraction is also becoming more and more high.Therefore,the extraction of field navigation lines has become the focus of attention.At present,most of the field navigation lines are extracted in the image coordinate system,the vegetation image information is obtained via sensing equipment,and the vegetation image information is processed via the relevant algorithms,and the field navigation line is obtained in the image coordinate system.Some scholars or research teams have also designed and developed the navigation equipment,installed sensors on the navigation equipment to obtain vegetation image information,in order to obtain the walking direction of the navigation equipment.The extraction of navigation lines in the image coordinate system is not conducive to the improvement of the navigation line precision.In order to improve the precision of the navigation line extraction and be applied to the actual coordinate system of the ground,this paper based on the existing navigation line extraction algorithms,aiming at the complex nonlinear environment in the field,using close view image and distant view image mapping to the actual coordinate system on the ground,then study the navigation line extraction algorithm based on image feature fusion.The main contents are as follow:(1)Study on factors affecting extraction of field navigation line,this paper studied the three factors affecting of navigation line extraction,including image coverage region,light propagation mode,selection of feature points,according to the factors affecting navigation line extraction,the definition of close view image and distant view image is given.(2)Study on coordinate system transformation based on close view image and distant view image.In this paper,under the indoor and field environment,using the neural network algorithm to transform the coordinate system,make the feature points of the image coordinate system mapping to the actual coordinate system on the ground.In this paper,four cameras with the same parameters are used to collect the close view images and distant view images respectively.The feature points extracted from the close view image and distant view image are mapped to the same ground actual coordinate system,and then extraction of the field navigation line.(3)Study on feature point fusion based on close view image and distant view image.After mapping the feature points of the close view image and distant view image to theactual coordinate system on the ground,the feature points of multiple images are fused,and then the field navigation lines are extracted via using the fused feature points,and then the navigation lines after the fusion are revised with modifying the lateral displacement.The field experiment results show that,within the distance of 3m on the ground,the error of field navigation line extracted with the fusion feature points of four images was relatively small compared with the actual field navigation line,the minimum angle error was 2.73°,the minimum lateral displacement error was 4.214 cm.The navigation line extracted after fusion of four images feature points was revised,the minimum angle error was 0.41°,the minimum lateral displacement error was 0.247 cm.This will provide a theoretical basis for improving the accuracy of navigation line extraction.