Novel automatic scorpion-detection and -recognition system based on machine-learning techniques

Abstract

All species of scorpions have the ability to inoculate venom, some of them even with the possibility of killing a human. Therefore, early detection and identification is essential to minimize scorpion stings. In this paper, we propose a novel automatic system for the detection and recognition of scorpions using computer vision and machine learning approaches. Two complementary image processing techniques were used for the proposed detection method in order to accurately and reliably detect the presence of scorpions. The first based on the fluorescence characteristics of scorpions when are exposed to ultraviolet (UV) light, and the second on the shape features of the scorpions. On the other hand, three models based on machine learning algorithms for the image recognition and classification of scorpions have been compared. In particular, the three species of scorpions found in La Plata city (Argentina): Bothriurus bonariensis (of no sanitary importance), and Tityus trivittatus and Tityus confluence (both of sanitary importance), have been researched using the Local Binary Pattern Histogram (LBPH) algorithm and deep neural networks with transfer learning (DNN with TL) and data augmentation (DNN with TL and DA) approaches. Confusion matrix and Receiver Operating Characteristic (ROC) curve were used for evaluating the quality of these models. Results obtained show that the DNN with TL and DA model is the most efficient model to simultaneously differentiate between Tityus and Bothriurus (for health security) and between Tityus trivittatus and Tityus confluence (for biological research purposes).