Research On Segmentation And Recognition Of Green Apple In Complex Background

With the rapid development of the field of artificial intelligence,automatic picking technology has been continuously improved.It is of great significance for Apple’s quality management and market optimization to realize the target recognition of green apple in complex background.However,in the complex background,the recognition rate of green apple is affected by light,shadow,fruit overlap and occlusion.In order to solve the interference caused by the complex background,this paper takes the green apple image as the research object,and proposes the segmentation and recognition method of the green apple object.The specific work is as follows:(1)In view of the noise interference in the process of green apple image acquisition,a variety of filtering methods are used for comparative test.Through subjective evaluation and objective evaluation,the filtering algorithm is analyzed.Finally,Gauss Filtering algorithm is selected as the denoising algorithm in this paper.(2)In order to solve the problem that it is difficult to segment the green apple image with high overlap and occlusion in complex background accurately,a method combining multi-color space threshold segmentation,morphological open operation and fusion segmentation results is proposed to extract the green apple region.To solve the problem that Hough transform is difficult to fit the apple contour accurately,a method based on the prior template information is proposed to realize the effective segmentation and recognition of green apple.The recognition rate and true positive rate are 83.33% and 94.99% respectively.(3)Using the method of Simple Linear Iterative Cluster(SLIC)and Support Vector Machine(SVM),the high-precision segmentation of green apple in complex background is realized.To solve the problem that the size and shape of the super pixel blocks are different and the texture features are difficult to extract,the method of Rotation Invariant Local Binary Pattern(RILBP)and improved gray level cooccurrence matrix is used to extract the texture features of the super pixel blocks.Combined with the color features extracted from RGB and HSI color space,the super pixel block classification is realized.Finally,Hough transform is used to detect the green apple target.The true positive rate was 95.7% and the false positive rate was 1.68%.The recognition rate was 91.89%.