Research On Recognition Of Seedling Row Lines Based On Illumination Correction And Sub-regional Feature Points Clustering

With the intelligent development of agricultural machinery,the precise automatic navigation technology of rice transplanter has become a research hotspot of a large number of scholars at home and abroad.Compared with satellite navigation,machine vision with the advantages of low cost,wide range and strong adaptability has low requirements on restriction conditions and land scale,which is consistent with China's national conditions and agricultural status,and once considered as an important development direction of navigation technology in the future.As a key technology in the vision navigation system of rice transplanter,the accuracy of extracting the row line of rice seedling directly affects the accuracy of the automatic walking of transplanter.However,the visual characteristics of the images obtained by the field camera are easily affected by the light,which makes the accuracy of extracting the center line of seedling row decrease.In order to improve the adaptability of the vision navigation system of transplanter to the Illumination,this paper studies the technology of extracting seedling row lines in complex illumination environment.The main research contents are as follows:(1)Analysis on the basic theory of rice seedling image processingThe correlation of each component of RGB and HSV color spaces and illumination color characteristics are analyzed,respectively,and the color space model of non-uniform illumination image processing in complex environment is determined.Besides,the extraction method of illumination component is studied where the accuracy and time cost of three algorithms as Retinex,bilateral filtering and fast guided filtering algorithms are compared and analyzed.The analysis results show that the fast guided filtering algorithm has good edge preserving performance and low algorithm complexity and can be selected to extract the illumination component.(2)Research of image illumination correction algorithm based on illumination partition adaptive Gamma functionThe adaptability of traditional adaptive Gamma image illumination correction algorithm is study based on the characteristics of seedling image and an image illumination correction algorithm based on illumination partition adaptive Gamma is proposed.Firstly,the illumination component is extracted in the HSV color space model and partition thresholds are calculated,which are used to divide the image into high brightness,normal brightness and low brightness.Then,based on the theory of light reflection imaging model,the luminance component of the seedling image is obtained with fast guide filtering algorithm,and the partition adaptive Gamma correction function is constructed where the correction control value is adaptive according to the characteristics of light information distribution.The experimental results show that the algorithm can effectively improve the influence of uneven illumination on the seedling image segmentation.(3)Research on detection of seedling row lines based on sub-regional feature points clusteringBased on the characteristics of seedling planting,the pixel distribution characteristics of seedlings are analyzed and an algorithm of seedling row lines detection based on sub-regional feature points clustering is proposed.Firstly,rice seedling pixels are segmented by using the extra green feature and Otsu method.Then,sub-region feature points of seedling row are extracted and clustered based on seedling image characteristics.Finally,the least square method,the Standard Hough algorithm and the random Hough algorithm are used to fitting the row pixel information.The experimental results show that the algorithm has the advantages of time-consuming,high accuracy,and can provide reliable navigation information for the transplanter to walk autonomously.