大豆病斑智能识别无损预处理及其特征提取方法的研究

Study on the method of non-loss pre-processing and feature extraction for intelligent recognition of soybean diseased spots

以大豆病叶为例,针对在采集叶片图像过程中出现的几何失真问题,提出了基于投影模型的植物叶片校正算法,测量准确性只与校正方法有关,不受其他背景因素影响。在精确校正的基础上,又实现了叶片病斑区域的特征提取与计算,实验结果表明,该方法校正精度达99%,病斑区域提取精度达100%,为进一步的病害诊断奠定了先期基础。

By using soybean as an example,this paper proposes a registration Algorithm named Projection Model,aiming to avoid Geometric distortion when collecting lamina images.The accuracy is related only to registration Algorithm.Furthermore,the diseased spots have been extracted and calculated,The results showed that the registration accuracy and extraction accuracy can reach 99% and 100% respectively.It is hoped this model provides a preliminary basis for disease diagnosis.