Crop Pest Image Classification and Recognition Based on Significance Detection and BOW Model

Abstract

In order to reduce the interference of background information on target recognition, and improve the accuracy of classification and recognition, this paper proposes a method based on significance detection and BOW model to classify and identify 22 kinds of crop pest images. First, smooth the image by bilateral filtering. Then use GBVS (graphic-based visual saliency) method to calculate the saliency map, and extract the surf feature from the region of interest and built the BOW model. Finally, use SVM (support vector machine) to train classifier. The experimental results show that the classification accuracy of the proposed method is nearly 10% higher than that of the traditional BOW model.