Artificial vision algorithm for the detection of Apodemus sylvaticus rodent pests in corn crops (Zea mays L.)

Abstract

The purpose of this work is to develop an artificial vision algorithm to detect the presence of Apodemus sylvaticus rodent pests in a Zea mays L. corn crop and implement the necessary hardware to guarantee the functionality of the algorithm in a crop in Laredo, since has seen that these rodents generate considerable economic losses due to contamination and nibbling of the fruit, which in turn cause diseases to people and animals that consume this product. For this research, Python software was used in a Python 3.8.0 programming language in a Visual Studio programming environment, a Yolov5 pre-trained convolutional neural network with 3615 illustrations of different rodents and; libraries such as base64, BytesIO, PIL import Image, time, torch and cv2. For the results, a sample of 225 images of 3 rodents detected in the culture (75 for each rodent) was considered, whose percentages of algorithm detection efficiencies are greater than 90%, that is, 97.33%, 98.67% and 100.00% for rodents 1, 2 and 3 respectively and; a total average efficiency of 98.67% with an error of 1.33%. In conclusion, the application of an artificial vision algorithm managed to detect the presence of Apodemus sylvaticus rodent pests in a Zea mays L corn crop.