No abstract

No abstract

Camera trapping is a technique to study wildlife using automatic triggered cameras. However, camera trapping collects a lot of false positives (images without animals), which must be segmented before the classification step. This paper presents a Multi-Layer Robust Principal Component Analysis (RPCA) for camera-trap images segmentation. Our MultiLayer RPCA uses histogram equalization and Gaussian filter as pre-processing, texture and color descriptors as features, and morphological filters with active contour as postprocessing. The experiments focus on computing the sparse and low-rank matrices with different amounts of camera-trap images. We tested the Multi-Layer RPCA in our camera-trap database. To our best knowledge, this paper is the first work proposing Multi-Layer RPCA and using it for camera-trap images segmentation.

No abstract

No abstract

Societal Impact Statement Tropical dry forests (TDF) underpin the wellbeing of millions, mostly rural populations; yet have suffered from severe clearing in Colombia, triggering cascading effects such as desertification. By engaging scientists, society, and institutions in the establishment of platforms for monitoring biodiversity and ecosystem functioning, crucial knowledge gaps will be bridged, helping to find a path toward sustainable development. Science‐led but socially and economically anchored information on biodiversity will help to incorporate nature's contributions to people into the society's cultural values. Ultimately, these transformative actions will translate into the comprehensive management of TDF through a greater impact in decision making. Summary Thousands of permanent plots have been established across the tropics with the purpose of monitoring tree communities. Research outcomes from these platforms, however, have been mainly directed toward the academic community, and their contribution to society has been limited so far. Here, we show how generating robust data on biodiversity has supported decision making in Colombian tropical dry forests (TDF), where less than 8% of their original cover remains. As a first step to build a national dialogue around the critical status of this ecosystem, a national collaborative network on TDF research and monitoring was born in 2014, the Red de Investigación y Monitoreo del Bosque Seco Tropical en Colombia (Red BST‐Col). Our main goal is to generate scientifically sound information that feeds into the comprehensive management of this ecosystem. To do so, a set of biodiversity monitoring platforms has been established across the country, which have already served to answer socio‐ecological questions related with deforestation drivers, citizen science, or the valuation of ecosystem services. Overall, this research agenda has nurtured the four lines that underpin the Program for the comprehensive management of dry forests in Colombia (knowledge management, preservation, restoration, and sustainable use), formulated by the Humboldt Institute, the United Nations Development Programme, and the Ministry of Environment in 2019. Many challenges are ahead, however, for a complex territory where multiple social actors and productive sectors coexist. The ultimate goal is to integrate all the dimensions of biodiversity to achieve a synthetic understanding of the functioning of the most endangered ecosystem in Colombia, and its relationship with local communities' wellbeing.

regiones biogeográficas y es poco conocida. Por lo tanto se emplearon técnicas de fototrampeo para el registro de los mamíferos medianos y grandes asociados al bosque ribereño en la cuenca alta y media del río Bita, Vichada, Colombia. Se estimó la diversidad, el índice de abundancia relativa y los patrones de actividad. Se instalaron 77 cámaras trampa a lo largo de 200 km del río en dos periodos hidroclimáticos (baja y alta precipitación). Se obtuvieron en total 16439 imágenes, con un esfuerzo de muestreo de 7700 días-trampa. Se registraron 24 especies de mamíferos, agrupados en 17 familias y 9 órdenes. El índice de diversidad de Shannon-Wiener obtenido a través de este método fue H´= 1,995, las especies mejor representadas a través del índice de abundancia relativa fueron Tayassu pecari (IAR = 0,176, n = 1360), Cuniculus paca (IAR = 0,174, n = 1346), Tapirus terrestris (IAR = 0,144, n = 1114) y Dasyprocta fuliginosa (IAR = 0,116, n=900). En términos de patrones de actividad Dasyprocta fuliginosa presentó patrones de actividad diurnos, Leopardus pardalis catemeral, Tayassu pecari crepuscular matutino, Cuniculus paca y Didelphis marsupialis nocturna y Tapirus terrestris crepuscular vespertino. Esta información puede ser de utilidad para la generación de programas para el manejo y conservación de la mastofauna en la cuenca del río Bita. Palabras claves. Cámaras trampa. Conservación. Escudo Guayanés. Orinoquia colombiana.

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.