Non-Invasive Monitoring: Panthera Onca Recognition Based on Image Processing

Abstract

Currently, Panama has projects to monitor endangered species such as the jaguar; for which they have been mainly using camera traps. These projects generate a large number of images that exceed, more often than not, the manpower capacity to label them. A faster way to do the tagging labor is to use new technologies associated with computer vision, which can reduce the time it takes to check whether or not a jaguar exists among the captured images. In this article, an algorithm has been developed using Matlab to perform image processing and recognition of jaguars based on background subtraction and the characteristic shape of jaguar spots. The algorithm achieves a high degree of precision of 80% and an accuracy rate of 79%, demonstrating that positive results can be obtained using computer vision for camera trap image analysis.