Research On The Identification System Of Fruit And Vegetable In Supermarket Based On Embedded Machine Vision

With the rapid development of agricultural e-commerce in China,fruit and vegetable sales model has undergone tremendous change.A growing number of supermarkets and chain of fruit and vegetable stores take the method that combined the traditional business model with the "order online,distribution offline" model to sell fruit and vegetable.Under the requirements of "more orders,fast delivery”,this new type of fruit and vegetable sales model is likely to be wrong when artificially identify fruit and vegetable.At the same time,fruit and vegetable food traceability system also hopes to be able to quickly and accurately identify the types of fruit and vegetable.Based on the above problems,this thesis uses the machine vision technology to study the method of fruit and vegetable identification,and applies the recognition algorithm to the embedded android system,and designs a set of identification system of fruit and vegetable in supermarket based on embedded machine vision.The main contents of this thesis are as follows:(1)Design of supermarket fruit and vegetable identification system based on embedded android development board.According to the functional requirements and performance requirements of the fruit and vegetable identification system,the hardware structure is designed and the hardware modules are selected,including the processor module,the image acquisition module,the WIFI module,the storage module and the electronic scale.And according to RS-232 interface protocol to solve the electronic scale and embedded development board between the serial communication problems.(2)Study on the algorithm of fruit and vegetable recognition based on machine vision.Firstly,a standard fruit and vegetable image database containing 16 kinds of fruit and vegetable was constructed.Then,the fruit and vegetable images were pretreated,including median filtering,image segmentation based on HSV color space and OTSU method,void filling,and finally extracting HSV color histogram features and LBP rotation invariant mode texture features.In this thesis,we experimentallyverify the recognition rate and recognition speed of three SVM multi-classification strategies to determine the multi-classification strategy of fruit and vegetable.A multi-classification strategy based on the main color hue and OAO-SVM is proposed to optimize the classifier.The results show that the average recognition rate can reach93.11%,and the recognition time also decreases slightly.(3)Design of software system for fruit and vegetable identification system.Firstly,this thesis study the construction of OpenCV development environment in android environment,and then design the database of fruit and vegetable identification system,fruit and vegetable identification software and information management software respectively.Finally,with the picture selection and camera shooting fruit and vegetable images to verify the identification rate and speed of fruit and vegetable identification system.