Application of Superpixel to identify Maggots and their larval stages

Abstract

Flies maggots are used to estimate the postmortem interval (PMI) through their developmental time in forensic entomology. Maggots identification is hard since they often have similar morphologies. Computer vision techniques and machine learning seem to be a good alternative to solve this problem. The aim is to create maggots microscopic images database and apply the dataset with an algorithm to automate the maggots identification. Although that approach could be used in forensic entomological identification and criminal expertise, this paper focuses on comparing the image classifications with IBK, J48, Random Forest and Random Tree classifiers. The Random Forest algorithm achieved the best performance, which was above 80% in most tests using the precision metric (P).