Research On Classification Technology And System Implementation Of Red Fuji Apple Based On Machine Vision

Apple has always been loved by consumers for its rich nutrition.With the continuous improvement of people’s living standard,in order to meet consumers’ consumption demands,it has become a key step to grade the external quality of apple efficiently and accurately.At present,the classification of external quality of apple in China mainly consists of artificial classification and mechanical classification.These two classification methods have the problems of high labor cost,single classification index and low classification efficiency.The use of machine vision technology can perform efficient and accurate automatic grading detection of Apple external quality.Therefore,This article uses machine vision technology to classify the external features of Red Fuji Apple and design the Red Fuji Apple online classification system.The main research contents are:Researched an external quality grading method for red Fuji apples based on ordinary image processing.Firstly,Set up the image acquisition system,the collected image of red Fuji apple is perform pre-processing operations,including the image grayscale processing,image filtering,image segmentation.After image segmentation,the image is further processed by morphological opening.Secondly,extract the red Fuji apple image for size,fruit shape and color features,and statistical analysis of the extracted data information.This study adopted the minimum circumferential circle get apple size features;adopted roundness value is used to describe the shape of apple,get apple shape features;Use the H component in HSV space to extract apple red pixels to obtain color features.Finally,the extracted external characteristic parameters of apples are compared with the Red Fuji Apple National Classification Standard to produce a comprehensive classification result.Researched a quality classification method of red Fuji apples based on convolutional neural network.Built two network models of VGG-16 and Res Net-50.A total of 2335 red Fuji apple image datasets were used to train and test the model.Using PYQT5 to achieve the interface development of the system.Completed the Red Fuji Apple online grading system based on VGG-16 and Resnet-50 network models.The results showed that:The overall accuracy of the external quality grading method of Red Fuji Apple based on ordinary image processing reached 82%.Classification of Red Fuji Apples Based on Convolutional Neural Network,the accuracy of the VGG-16 model reached 94%,and the accuracy of the Res Net-50 model reached 97%.Red Fuji Apple’s classification system based on machine vision has completed accurate and efficient automatic classification of apples,providing a theoretical research basis for the external quality classification of apples in China.