Research On Recognition Method Of Green Apple Target In Natural Environment

China is the country with the largest apple acreage and the highest total output in the world.Due to the increase of social labor cost in recent years in our country,the cost of fruit harvesting is high.With the rapid development of automation technology,the birth of agricultural robots has made it possible to reduce the cost of fruit harvesting and improve the level of agricultural automation.In the course of picking robots,the fruit identification is the most crucial step.The accuracy of the identification directly affects whether the robots can finish the picking quickly and accurately.Therefore,it is very important to develop an efficient and accurate identification algorithm to realize the automatic fruit picking.Green apple was chosen as the research object in this paper.In order to solve the problems of the identification of green apples in the natural environment,the main research contents and conclusions are as follows:(1)In order to solve the spot problem caused by uneven illumination in natural environment,vector median filter is used to smooth and de-noisy the image.The characteristics of spot are analyzed according to the dichromatic reflection model.The method can be designed by using spot characteristics which can be used to extract the spot area.Then the image restoration algorithm can repair the image spot removal area.(2)Aiming at the recognition of non-overlapping fruit in natural environment,this paper proposes a recognition method based on the Support Vector Machine(SVM),which also combine with the texture,color and shape features of the image to realize the recognition of the green apple in the natural environment.(3)The method was proposed to locate overlapping apple targets in natural environment.The improved spectral clustering algorithm is used to segment the image,and Random Hough Transform(RHT)is used to identify and locate the fruit.In order to solve the problem of large number of traditional spectral clustering operations and slow computing speed,this paper improves the traditional spectral clustering algorithm based on Mean Shift(MS)and sparse matrix principle.Firstly,the MS is used to segment the image.The prior information provided by the segmentation result is combined with the color feature and the texture feature of the image to establish the sparse similarity matrix.The data processing capacity of the spectral clustering algorithm is greatly reduced through the improvement,which improves the algorithm running speed.In this paper,the problem of uneven illumination in natural environment,the identification of non-overlapping fruit and overlapping fruit are studied separately,and the corresponding identification methods are proposed.Finally,the collected test images are respectively tested based on the matlab R2013 a platform and the experimental results verify the effectiveness of the algorithm.