Research And Application Of Tea Colour Analysis Based On Color Features

Tea is closely linked with the life of Chinese people, and is one of China’s majorexport trade agricultural products. In addition, it is also one of China’s importanteconomic crops. The last procedure in the production of tea is to pick up. The qualityof tea is mainly judged by its colour and luster. At the present time，the selection oftea is still by traditional low-end artificial means. Apparently，this is a time-wastingand energy-consuming work, and there are great human factors. Therefore，the paperproposes the machine vision technology combined with digital image processingtechnology in order to seek a rapid tea quality testing method.In this paper，we study the fast and efficient extraction algorithm that is suitablefor agricultural image, which is the core question of agricultural product surfacetesting technology--feature extraction，based on digital image processing techniques.Like natural images，even the same kind of agricultural product images differ inthings such as color，texture and shape，and this is the basis of image featuresextraction. Take the image of Fujian tea “Tieguanyin” for example， this paperstudies the tea image feature extraction algorithm. Mainly based on RGB and HSVcolor characteristics analyzing，we construct feature vector function and carry onclustering algorithm research，and finally complete the classification of the color andlustre of tea image.The main research work includes:1. Through reading a lot of literature and learning digital image processingknowledge，the author researches BMP file structure and access.2. The author adopts improved seed algorithm to realize the regional extractionof characteristics of tea image color， and designs a set of applications program(color separator sampling system). in Microsoft Visual c++2005.3. The paper introduces RGB and HSV color space and its mutual transformationrelation，and analyses the distribution of tea color characteristics in R， G， B，(R-G)，(R-B)， H， S， V, discusses several characteristic vectors to differentiatethe quality of tea.4. The paper structures feature vector function based on Fisher discriminantanalysis theory, uses K-means clustering algorithm，and finally completes theclassification of the colour and lustre of tea image.