Jhony-Heriberto Giraldo-Zuluaga

Jhony-Heriberto Giraldo-Zuluaga, Augusto Salazar, Alexander Gomez et al.

The segmentation and classification of animals from camera-trap images is due to the conditions under which the images are taken, a difficult task. This work presents a method for classifying and segmenting mammal genera from camera-trap images. Our method uses Multi-Layer Robust Principal Component Analysis (RPCA) for segmenting, Convolutional Neural Networks (CNNs) for extracting features, Least Absolute Shrinkage and Selection Operator (LASSO) for selecting features, and Artificial Neural Networks (ANNs) or Support Vector Machines (SVM) for classifying mammal genera present in the Colombian forest. We evaluated our method with the camera-trap images from the Alexander von Humboldt Biological Resources Research Institute. We obtained an accuracy of 92.65% classifying 8 mammal genera and a False Positive (FP) class, using automatic-segmented images. On the other hand, we reached 90.32% of accuracy classifying 10 mammal genera, using ground-truth images only. Unlike almost all previous works, we confront the animal segmentation and genera classification in the camera-trap recognition. This method shows a new approach toward a fully-automatic detection of animals from camera-trap images.

Jhony-Heriberto Giraldo-Zuluaga, Augusto Salazar, Juan M. Daza

Animal biometrics is an important requirement for monitoring and conservation tasks. The classical animal biometrics risk the animals' integrity, are expensive for numerous animals, and depend on expert criterion. The non-invasive biometrics techniques offer alternatives to manage the aforementioned problems. In this paper we propose an automatic segmentation and identification algorithm based on artificial vision algorithms to recognize Diploglossus millepunctatus. Diploglossus millepunctatus is an endangered lizard species. The algorithm is based on two stages: automatic segmentation to remove the subjective evaluation, and one identification stage to reduce the analysis time. A 82.87% of correct segmentation in average is reached. Meanwhile the identification algorithm is achieved with euclidean distance point algorithms such as Iterative Closest Point and Procrustes Analysis. A performance of 92.99% on the top 1, and a 96.82% on the top 5 is reached. The developed software, and the database used in this paper are publicly available for download from the web page of the project.