Research On Vision Navigation Technology Of Weeding Robot Based On ROS

Large-scale promotion and use of agricultural intelligent equipment is an important guarantee for the realization of modern fine agriculture.As one of the intelligent agricultural equipments,weeding robots can effectively reduce the amount of herbicides and realize the protection of ecological environment,and it is becoming a hotspot in agricultural robot research.This paper studies the visual navigation technology of weeding robot,including farmland image processing technology and navigation lineextraction technology.Farmland images contain a wealth of information that can be made easier to obtain after preprocessing.The preprocessing steps of farmland image includes color space selection,image grayscale,image enhancement and filtering.In this paper,the farmland images in different color spaces are compared and analyzed,and the RGB color space is selected for gray-scale operation of farmland images.Analysis and comparison of several commonly used farmland image grayscale method,because the 2g-r-b gray method can better overcome the influence of light intensity on the image quality,so choose to use this method to gray the image.In order to effectively remove the noise and protect the detailed information of the farmland image,the improved adaptive median filter algorithm is used to filter the farmland image by analyzing and comparing the commonly used image filtering algorithm.The algorithm uses the mean filtering which has a strong anti-noise ability to improves the adaptive median filter algorithm so that it can effectively remove the noise and reduce the damage of the image detail information.In farmland image,the difference of color between crop and soil is obvious,according to this characteristic,in this paper,fuzzy c-means clustering segmentation method,iterative threshold method,the maximum entropy threshold method and Otsu threshold segmentation method were studied and compared,Otsu threshold segmentation method has good segmentation effect and real-time,so choose the method tosegment the farmlandimage.This paper improves the gray vertical scanning algorithm,so that it can accurately detect the information of the crop column in the farmland image.After obtaining the column number of the crop and the edge information of the column,the edge information of the column is used to obtain the known point.On the basis of deep research on the navigation line detection algorithm of known points,this paper analyzes the shortcomings of this algorithm and improves it,which effectively overcomes the defects in the original algorithm and accurately detects the navigation line in the image.Combined with the camera calibration results,how to get navigation parameters from the navigation line was studied in this paper.The function of image acquisition and navigation line detection are implemented and tested in theROS environment.The results show that the navigation line can be accurately detected by using the algorithm researched and improved in this paper.