基于ComⅥ和双阈值OTSU算法的农作物图像识别

针对农作物图像中依附泥土和杂质噪声呈现不规则性和复杂性特点,提出了一种基于植被指标合成双阈值OTSU算法的农田作物图像识别方法.该方法根据农作物充分显露和部分被遮盖2类图片特点,将图像识别过程分为3个阶段：首先利用植被指标合成获取农作物图像灰度图,然后根据双阈值OTSU自适应算法进行二值化处理与图像分割,再进行正常的形态学运算,将3个阶段所分割的图像叠加形成最终的农作物与土壤识别图像,并将该算法与双阈值迭代设定法进行了对比.试验研究表明该算法克服了传统灰度图算法和阈值迭代算法的缺点,能有效提取和识别过渡区域的边缘,图像识别的准确率为92.7%以上.最后,采用Visual Basic2010和Matlab 2012软件设计了农作物图像识别系统,从应用角度实现了图像识别的可视化与自动化.

Crop image recognition based on ComVI and double threshold OTSU algorithm

A new recognition method of crop image with soil and other noise was proposed, which was used by combined vegetative indices（ComVI） and double threshold OTSU algorithm. The method has three stages of image processing according to unmasked and masked crops image characters. Firstly, the gray -scale image was adopted with combined vegetative indices, then the binaryzation and image segmentation were processed by way of double threshold OTSU adaptive method, and finally, post - treatment to the segmented binary images by morphology operation was also implemented. Therefore, crops and coil were recognized by superimposing the segmentation image in the three stages. Experimental results shows that it is a high performance algorithm which can overcome drawbacks of traditional gray -scale image and threshold iterative method, and can effectively extract and recognize transitional edge image. The image recognition accuracy is more than 92.7 %. Finally, tion system was designed to achieve the image recognition of the visualization and Basic 2010 and Matlab 2012 software. crop image recogniautomation by Visual Basic 2010 and Matlab 2012 software.