基于机器视觉的花生品种识别系统研究

摘要

为了实现花生品种自动识别，收集了12个花生品种的600粒籽仁作为实验样本，利用计算机视觉和图像处理技术对图像进行处理；然后，对每幅样本图像提取形态、颜色和纹理等3大类共48个特征，建立人工神经网络（ ANN ）和支持向量机（ SVM ）识别模型对这些特征进行分析识别，并基于 Visual C＋＋6．0环境构建识别系统。运行结果表明，SVM 方法识别效果比较稳定，对12个花生品种自我识别率达到100％，测试识别率达到83％；另外，基于Visual C＋＋的识别系统在识别效果与效率方面比Matlab 都有了较大的提高。该花生品种识别系统对于花生品种识别具有积极意义。%Based on image processing and computer vision technology , we obtained 600 images of 12 varieties,Among which We take 45 images as training sample images , and the remaining 5 as test images .For each image , we have ac-quired 48 characteristics .Then we constructed the artificial neural network models and support vector machine model to identify different species based on Visual C++6 .0 .Results show that the recognition effect of the SVM method is more stability and the overall self-recognition performance can reach 100%.At the same time , recognition system based on Visual C++has better recogniton effect and higher efficiency than the system based on Matlab .Methods and conclusions of this paper have positive significance to the detection of Peanut varieties .