AUTOMATIC PEST COUNTING METHOD BASED ON COMBINATION OF MULTI-SCALE FEATURE FUSION NETWORK AND POSITIONING MODEL

摘要

An automatic pest counting method based on the combination of a multi-scale feature fusion network and a positioning model, comprising the following steps: step 1, establishment of training samples: obtaining several pest images in a natural field environment as training images, and marking pests in the images to obtain training samples; step 2, construction of a pest detection and counting model: constructing a positioning model and a multi-scale feature fusion network, extracting a candidate region of the training samples by using the positioning model, extracting features in the candidate region by means of the multi-scale feature fusion network and then classifying the features, and outputting coordinate values in corresponding images; step 3, obtaining of images to be counted: obtaining pest images captured in the field and preprocessing said images to obtain images to be counted; and step 4, obtaining of the number of pests: inputting the images to be counted into the pest detection and counting model to obtain the number of pests in the images. The method can precisely determine the number and specific locations of pests, and has high robustness and accuracy.