Algorithm And Implementation Of Fruit Detection Based On Region Convolution Network

China is a country rich in fruits,but at the same time facing the fact that postpartum technology is inconsistent with the present situation of the high yield,and now the deep learning methods have developed rapidly,driven by the big data and the rapid development of computer technology,which have very good practice in the field of target detection,If the methods of deep-learning are used in fruit target detection,it will play an important role and have practical significance in realization of automatic picking for fruit and fruit quality detection as well as production forecast.Convolution neural network(CNN)is one of the deep learning method,which has been widely used in image recognition technology,This paper start from the study of the deep CNN.According to the format of Pascal VOC,We finished the artificial markers and produced apricot and green orange data set for training and testing.We first selected typical CNN ZF net for classified training,analyze its extraction effect of fruit characteristics and finally recognition accuracy whether meet the testing requirements.On this basis,we make the research of CNN based on area,its core is using the CNN to extract the region proposal and do classification with sharing convolution.We choose two feature extraction network(ZF,VGG network)to set up regional convolution network,and achieve the target detection of the apricot and green orange.Furthermore,we adjust the training parameters according to the test result.In the end,we make the test results visualization and complete the GUI design of fruit detection system.