Automated identification of animal species in camera trap images

Abstract

Image sensors are increasingly being used in biodiversity monitoring, with each study generating many thousands or millions of pictures. Efficiently identifying the species captured by each image is a critical challenge for the advancement of this field. Here, we present an automated species identification method for wildlife pictures captured by remote camera traps. Our process starts with images that are cropped out of the background. We then use improved sparse coding spatial pyramid matching (ScSPM), which extracts dense SIFT descriptor and cell-structured LBP (cLBP) as the local features, that generates global feature via weighted sparse coding and max pooling using multi-scale pyramid kernel, and classifies the images by a linear support vector machine algorithm. Weighted sparse coding is used to enforce both sparsity and locality of encoding in feature space. We tested the method on a dataset with over 7,000 camera trap images of 18 species from two different field cites, and achieved an average classification accuracy of 82%. Our analysis demonstrates that the combination of SIFT and cLBP can serve as a useful technique for animal species recognition in real, complex scenarios.

Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networks

“…A few previous works proposed solutions for this problem. Yu et.al [4] manually cropped and selected images, which contain the whole animal body. This conditioning allowed then to obtain 82% of accuracy classifying 18 animal species in their own dataset.…”

On-Board Video Recording Unravels Bird Behavior and Mortality Produced by High-Speed Trains

“…Additionally, artificial intelligence systems under development are expected to enable extraction of useful wildlife data from images taken in nature by automated cameras (e.g., Yu et al, 2013;Longmore et al, 2017). This extraction capacity will be a key advance to enable data analysis of highspeed recordings from train fronts, similar to the extraction of useful biological monitoring data from continuous recording webcams or sensor networks (Benson et al, 2010;Porter et al, 2010;Evans et al, 2015).…”

“…In any case, automating data interpretation of the on-board recordings will be a challenge. The rails, posts, and catenary that comprise the infrastructure limit the analysis focus, but the constant (relative) movement of the background makes it difficult to distinguish the targets (birds) from the landscape elements in the recordings (Yu et al, 2013;Longmore et al, 2017). A solution for this would be to combine high-speed cameras with complementary sensors (e.g., sound/impact) in the front of the train, to precisely record collision impacts, and in combination with automated recording systems capture specifically the video fragments for the moments just before a collision occurs.…”

“…), and (iii) data review and interpretation to enable extraction of only relevant data from the huge volume of data collected. Technology advances have already addressed the first two requirements and the emergence of an artificial intelligence system may soon enable automated analysis of the generated images (Yu et al, 2013;Longmore et al, 2017).…”

Wild life passer species recognition from a technical passage through data fusion of a wireless sensor network

“…Tagged animals are bears equipped with either commercial or custom radio-colars, developed in the ALPINE project. As far as we are aware, the only alternative approach to species recognition of untagged animals is using image processing, e.g., [11], [12], however this is usually performed o↵-line, using static images, taken as snapshots and it is error-prone.Lastly, in terms of mathematical analysis we studied the usage and e↵ect of pattern classification, pattern recognition [13] and clustering algorithms [14] to our final proposed system.…”