A Random Forest Classification Model for Transmission Line Image Processing
Zhang Bingzhen, Qiao Xiaoming, Yang Hemeng, Zhubo Zhou
Abstract
Random Forest algorithm is widely used in transmission line image processing, which is one of the important applications of classifier integration. But in the process of classification, when the classification results of each base classifier have similar error distribution, the final reduction of the classification effect. Aiming at this problem, this paper proposes a method of similarity measure of decision tree based on confusion Matrix. This method considers the number of trees in different categories, the classification of the correct and wrong cases, and the classification of the wrong classification. At the same time, combined with the classification performance of the decision tree, the model selection of random forest is completed by using the strategy of "deleting inferior". The experimental results show that the proposed method has higher average classification accuracy and higher stability than the original algorithm in three data sets. Therefore, the random forest image classification model based on confusion Matrix can improve the classification ability of random forest.