基于机器学习的小流域土壤侵蚀量检测方法

Soil Erosion Detection Method for Small Watershed Based on Machine Learning

目前的土壤侵蚀量检测方法对于土壤侵蚀程度的分类精度较低,因此提出一种基于机器学习的小流域土壤侵蚀量检测方法。首先对采集到的图像进行预处理,主要包括灰度化、去噪、分割、形态学处理等一系列操作,为后期的支持向量机分类打下基础。提取出小流域土壤侵蚀图像特征及类型因子,在检测过程中引入支持向量机构建一种新土壤侵蚀分类模型,利用最优超平面对数据进行归一化,实现土壤流失情况分类。为验证所设计的检测方法的有效性,选择某研究区进行实验。侵蚀量分类检测的实验过程根据研究区的相关参数完成建模并进行分类检测,为验证设计检测方法的有效性,选择传统检测方法作为对照。实验结果表明,设计方法的检测分类结果更加接近实际值,验证了设计方法在小流域土壤侵蚀量分类检测中的有效性。

Current soil erosion detection methods have a low classification accuracy for soil erosion degree,so a small watershed soil erosion detection method based on machine learning is proposed.Firstly,the collected image is preprocessed,including grayscale,denoising,segmentation,morphology processing and a series of operations,which lays a foundation for the later classification of support vector machine.In addition,a new soil erosion classification model was built by introducing support vector mechanism in the detection process,and the optimal hyperplane was used to normalize the data to realize the classification of soil loss.Designs for the validation of the method of testing validity,choose some experiments in the study area,erosion classification test experiment is accomplished according to the related parameters in the study area classified modeling and testing,to verify the validity of the design method,the traditional detection method by contrast.The experimental results show that the design method of detecting classification result is more close to the actual value,to verify the design method is an effective means to detect in the classification of soil erosion of small watershed.