Machine Learning Algorithms for Image Processing and Recognition of Animal Objects in Photo-DVR Images
А. В. Ваганов, Константин Сергеевич Печененко, Lyubov A. Khvorova · Izvestiya of Altai State University · 2024
The article is devoted to the development of algorithms and software for processing and analyzing camera traps images to recognize animal objects in them. The algorithms are based on computer vision technologies, image mining, machine learning, and artificial intelligence, which helps increase the speed of image processing photographs and the reliability of object recognition. The software allows processing large amounts of image data, starting from the raw set of camera traps images and up to cataloging the images by types of objects in them. It aims at improving the quality and speed of processing images and increasing the reliability of object recognition. Thus, automatic classification of images with defects (images with damaged pixels, fogged and blurred images) and no defects is essentially the first step of the proposed algorithms. Next, a deep convolutional neural network is used to identify and classify images with animals and images with no animals. Images containing humans, vehicles, and objects of nature are identified and sorted out. The detailed .CSV file report is generated with the file names of images and labels of identified objects in them.