基于根系图像处理的番茄枯萎病检测研究

Research on Tomato Fusarium Wilt Detection Based on Root Image Processing

摘要: 番茄枯萎病是番茄病害中最严重的一种，枯萎病的早期识别具有重要意义。本研究以患枯萎病番茄的根部为实验对象，通过图像处理技术，首先将番茄根部用扩展高斯差分(XDoG)进行边缘检测，在HSV色彩空间中对番茄枯萎病进行检测。对于根系没有颜色变化的样本，提取与病害相关的根部形状参数，并结合从根系扫描仪获取的参数，建立随机森林(RF)检测模型，识别率为92.64%。为了缩短该方法的运行时间并提高准确率，引入主成分分析法(PCA)，建立PCA-RF模型，该模型的运行时间提高了62.13%，平均识别率提高了2.62%。结果表明，与常用的识别算法相比，PCA-RF模型具有更高的检测准确率。本研究为番茄枯萎病识别提供了一种高效稳定的方法。

Abstract: Tomato Fusarium wilt is one of the most serious tomato diseases, and early identification of tomato Fusarium wilt is of great significance. In this study, the roots of tomatoes with Fusarium wilt were used as the experimental object. Through image processing technology, the tomato roots were edge detected by using the extended difference of Gaussian (XDoG), and the tomato Fusarium wilt was detected in the HSV color space. For samples with no color change at the root, the root shape parameters related to the disease were extracted, and combined with the parameters obtained from the root system scanner, a Random Forest (RF) detection model was established, and the recognition rate was 92.64%. In order to shorten the running time of the method and improve the accuracy, Principal Component Analysis (PCA) was introduced and a PCA-RF model was established. The running time of the model was increased by 62.13% and the average recognition rate was increased by 2.62%. The results show that the PCA-RF model has higher detection accuracy than the commonly used recognition algorithms. This study provides an efficient and stable method for tomato Fusarium wilt identification.