A vegetable identification method and device

Abstract

The invention provides a vegetable identification method and device. The method includes the steps of establishing the first convolution neural network model, collecting the images of each vegetable in different scenes, building the corresponding input dataset, iteratively training the first convolution neural network model according to a learning rate adaptively updated gradient descent algorithmand the input data set, obtaining the second convolution neural network model, carrying out the vegetable recognition according to the second convolution neural network model, and improving the efficiency of the vegetable recognition is improved. The first convolution neural network model is trained by a stochastic gradient descent algorithm, so that the second convolution neural network model obtained by the training can accurately identify the vegetable species under the complex background.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a vegetable recognition method and device. Background technique [0002] Against the backdrop of economic globalization, agricultural product markets are gradually opening up around the world, and product types and transaction volumes have surged, triggering a historic change in the sales terminals of agricultural products. At present, in the identification of vegetables in supermarkets and farmers’ markets, most of the classification work must be done manually, relying on human eyes to identify vegetables. In the past three years, my country’s market supervision departments have tried to implement traceability scales at agricultural product sales points. During the sales process , the vendor enters the PLU code of the product, and will print out a small ticket with information such as "traceability code" and send the information to the upstream server for management and...

AI Technical Summary

Problems solved by technology

At present, in the identification of vegetables in supermarkets and farmers’ markets, most of the classification work must be done manually, relying on human eyes to identify vegetables. In the past three years, my country’s market supervision departments have tried to implement traceability scales at agricultural product sales points. During the sales process , the vendor enters the PLU code of the product, and will print out a small ticket with information such as "traceability code" and send the information to the upstream server for management and traceability. However, the implementation effect is not obvious. Compared with ordinary scales, Traceable scales require vendors to manually enter the PLU code according to the products they sell, which is cumbersome and inefficient. In this way, not only a lot of labor is consumed, but also the work efficiency is low, which seriously affects the development speed of vegetable commercialization

[0003] The existing Chinese patent (application number: 201510004587.8) discloses an integrated scale based on the recognition of fruits and vegetables, which mainly uses the NCC template matching algorithm for image recognition, focusing on category recognition and fruit and vegetable traceability. The main problem with the template matching algorithm is the requirement The template can be as similar as possible to the image to be recognized, but the vegetable packaging is updated quickly, so the template image also needs to be updated continuously to improve the accuracy, and this algorithm has the problem of being unable to recognize vegetable images with complex backgrounds

[0004] The existing Chinese patent (application number: 201711133538.X) discloses an image recognition method and a traceable scale terminal. It mainly uses the deep residual neural network algorithm to identify the vegetable category on the initial image after image enhancement. Network "residual learning" principle, but this algorithm is relatively complicated, mainly reflected in the subclass recognition of the initial image, which needs to be identified first, and then continue to identify, so there is a problem of low recognition efficiency

Images