Research On Classification Of Tea Quality Based On Computer Vision Technology

At present,the method of classification of tea quality at home and abroad is still the main method that based on sensory evaluation which is existing subjective judgment and large external disturbance.And Physical and chemical methods are difficult to get rid of the complex process,high cost and heavy workload.In order to make up for the shortcomings of the existing evaluation methods,this study used green tea as the research object,using computer vision technology to achieve rapid nondestructive detecting and classification of tea quality.In this study,the computer vision hardware system was built independently,and the samples of different quality were collected.The original image was pretreated,the shape features and color characteristics of the tea images were extracted,and the rapid evaluation model of the tea grade was established.A set of intelligent grades system for tea quality based on computer vision system were developed,providing a technical support for online identification of tea quality grades.The main contents and conclusions are as follows:(1)The influence of various parts of computer vision system on image quality was analyzed and the computer vision classification hardware system for this study was set up.By using the different pretreatment methods on the original image preprocessing,it is known that the 3×3 size template for median filter could be used in the image smoothing method and the 3×3 size template for Laplacian operator could be used in the image sharpening method,which can eliminate the original image noise information.The experimental results show that the visual system can obtain high quality images,and the pretreatment method can improve the image quality.(2)This paper explored the effective extraction of the shape and color characteristics of tea image by computer vision technology.12 color features of tea images were extracted by using the RGB and HSI color models,and 22 texture features of tea images were extracted based on gray level co-occurrence matrix and gray statistic moment method.The extracted characteristic parameters were consistent with the actual observed tea characteristics,which can effectively reflect the shape information for tea leaf.(3)The tea grading model that based on computer vision,PCA-GA-BP neural network and PCA-PSO-LSSVM support vector machine was established.By using the principal component analysis method to extract the principal component of the characteristic parameters of tea image,the PCA-GA-BP neural network and the tea quality grading model of PCA-PSO-LSSVM were established.In this model,the accuracy of PCA-GA-BP neural network model is 92.5% for Biluochun tea,and the accuracy rate of PCA-PSO-LSSVM is 90%.The accuracy rate of PCA-PSO-LSSVM support vector machine for the identification of Wuyuan green tea is 91.37%,and the accuracy rate of PCA-GA-BP neural network is 87.50%.Compared with the traditional neural network and support vector machine classifier,the model is faster convergence and higher accuracy.(4)A set of tea intelligent grading system was developed,which based on Open CV,Visual Studio 2010 platform and computer vision.The system supports the camera to collect images online,get the image offline and provides image preprocessing,such as the color component display of tea image,filtering,sharpening and so on,the feature extraction of tea image,the data analysis statistics and the assessment model are established.In this system,various types of tea quality grades are identified by intelligence.The system is simple,convenient and has strong portability.