基于计算机视觉的水稻杂株识别研究

Identification of Rice Hybrids Based on Computer Vision

水稻杂株是品种形成过程中的干扰因素,对水稻产业具有较大的危害。水稻杂株的防除以识别为前提,但目前的识别方法消耗大量人力,识别的效率也不理想。计算机视觉是一种图像分析处理技术,在农业领域的应用较广。为此,设计了基于计算机视觉的水稻杂株识别方法,拍摄图像后依次进行预处理、灰度化和二值化,最后根据外观特征采用阈值分割法将杂株识别并提取出来。试验结果表明:秧苗期水稻杂株的性状特征最少,导致计算机视觉的识别效果较差;计算机视觉在抽穗期的识别率最高,误识率最低,具有良好的识别效果。因此,这种识别方法最适合在水稻的抽穗期使用,可以为水稻的品种形成提供技术支撑。

Rice hybrids are the interfering factors in the formation of varieties and have great harm to the rice industry.The control of rice hybrids is premised on identification,but the current identification method consumes a lot of manpower,and the efficiency of recognition is not satisfactory.Computer vision is an image analysis processing technology that is widely used in agriculture.In this paper,a method based on computer vision for rice hybrid identification was designed.After the images were taken,pre-processing,gray-scale and binarization were carried out.Finally,the hybrids were identified and extracted according to the appearance characteristics.The results showed that the characteristics of rice hybrids in seedling stage were the least,which led to poor recognition of computer vision.Computer vision has the highest recognition rate at the heading stage,the lowest misrecognition rate,and a good recognition effect.Therefore,this identification method is most suitable for use in the heading stage of rice,and can provide technical support for the formation of rice varieties.