Green Apple Recognition Method At Night For Apple Picking Robot

At present,most of the apple picking operations in my country are still manual picking,with low work efficiency and high labor intensity.However,since the reform and opening up,my country's rural labor force has been continuously outflowing,and there is often a shortage of manpower.Realizing automatic apple picking is of great significance for improving apple quality and picking efficiency.This subject takes green apple as the research object.On the basis of reading a large number of documents,a method for identifying green apples at night by an apple picking robot is proposed.Image acquisition,image denoising,image enhancement,image segmentation,and apple recognition of green apples at night were performed on computer vision related platforms,and the algorithm was verified through experiments.The main research contents and results of this article are as follows:(1)An experimental platform was built in the laboratory environment to simulate the orchard environment,and it was proposed to use bright field diffuse front lighting as a night light supplement method,which effectively solved the problem of light spots and shadows at night images.Based on the analysis of the difference image method,the noise type of the night image is Gaussian noise.Through the peak signal-to-noise ratio index,the optimal guide filter algorithm is selected from a variety of filtering methods to solve the noise problem of the night image.(2)The fusion-contrast-contrast histogram equalization algorithm and Gammacorrected night image enhancement algorithm are used to effectively solve the problems of dark images at night,poor contrast,unclear details and blurred edges.Experimental results show that the algorithm can effectively improve the image mean,standard deviation,average gradient and information entropy.(3)Aiming at the problem that green apples and leaves are similar in color and difficult to rely on color for segmentation,an image segmentation method based on superpixel classification is proposed.First,a simple linear iterative clustering algorithm is used to super-segment the image,which greatly reduces the number of pixels.Then the H and Cr components of the superpixels are extracted as color features,and the mean,standard deviation,consistency,smoothness and entropy based on the LBP operator are extracted as texture features.The support vector machine is trained by using the characteristic data set.Finally,the trained support vector machine is used to realize the segmentation of the green apple image.(4)According to the area threshold,remove the super pixel area that is misclassified as apple;use the open operation to smooth the edge of the binary template image;use the local parameter adaptive Hough circle for the defects of occlusion,overlap and size change in the image of apple The transformation method identifies the location of the apple.Experimental results show that the recognition rate of green apples at night is 87%,and the average running time is 520 ms.The above research results show that the apple picking robot night green apple recognition method proposed in this paper can effectively overcome the difficult conditions of night and the difficulty of color difference between apples and leaves,which is of great significance for the next step of identifying the automatic picking of green apples at night.