Abstract
All around the world, poisonous scorpions are still considered as a public health issue. The scorpion’s species can be determined by its physical characteristics. Different methods have been applied to differentiate among different insects, such as bugs, bees and moths. However, none have been applied to distinguish between different scorpion species. This paper presents a procedure to distinguish between two different species of scorpions (Centruroides limpidus and Centruroides noxius) using image processing techniques and three different machine-learning methods. First, the live scorpion is distinguished from the photograph image using a dynamic separation threshold obtaining its area and contour. A shape vector is obtained from both, area and contour, calculating the following features: aspect ratio, rectangularity, compactness, roundness, solidity and eccentricity. Finally, artificial neuronal network, classification and regression tree, and random forest classifiers are used to differentiate between both species. All three classifiers were evaluated by accuracy, sensitivity and specificity. Experimental results are reported and discussed. The best performance was obtained from the Random Forest algorithm with 82.5 percentage of accuracy.
Figure 1A: Photographs of the two studies scorpions.
Figure 1B: Photographs of the two studies scorpions.
Figure 2A: Color distribution of scorpions.
Figure 2B: Color distribution of scorpions.
Figure 2B: Error rate between the number of tree in th RF..
Citation
J. C. Urteaga-Reyesvera and A. Possani-Espinosa, “Scorpions: Classification of poisonous species using shape features,” 2016 International Conference on Electronics, Communications and Computers (CONIELECOMP), Cholula, Mexico, 2016, pp. 125-129, doi: 10.1109/CONIELECOMP.2016.7438563.
@INPROCEEDINGS{urteaga2016scorpions,
author={Urteaga-Reyesvera, J. Carlos and Possani-Espinosa, Andre},
booktitle={2016 International Conference on Electronics, Communications and Computers (CONIELECOMP)},
title={Scorpions: Classification of poisonous species using shape features},
year={2016},
volume={},
number={},
pages={125-129},
keywords={Artificial neural networks;Insects;Shape;Support vector machines;Feature extraction;Vegetation;Scorpions;species classification;shape feature;random forest},
doi={10.1109/CONIELECOMP.2016.7438563}
}