IENE 2016

Roads have multiple effects on wildlife, from animal mortality, habitat and population fragmentation, to modification of animal reproductive behaviour. Amphibians in particular, due to their activity patterns, population structure, and preferred habitats, are strongly affected by traffic intensity and road density. Monitoring road-kills is expensive and time consuming, and depend mainly on volunteers. Thus, cheap, easy to implement, and automatic methods for detecting road-kills over larger areas and along time are necessary. We present results from the research project Roadkills, a cheap and efficient system for detecting amphibians road-kills using computer vision techniques from robotics. We propose two different solutions: 1) a Mobile Mapping System to detect automatically amphibians' road-kills in roads, and 2) a Fixed Detection System to monitor automatically road-kills in a particular road place during a long time. The first system detects and locates road-kills through the automatic classification of road surface images taken from a linear camera installed in a trailer with a standalone power generation, an imaging recording computer, an accurate GPS receiver, and a linear standardised lighting. The linear camera has an optical resolution between 250 µm/pixel at 35 km/h, 500 µm/pixel at 70 km/h and 1000 µm/pixel at 140 km/h. The camera acquire sequential images of 4096 pixels width and 1 pixel length (approx. 1.0 m x 0.25 mm). The fixed system consists on a standalone tower which captures a section of a road and allows close-up snapshots of instances moving along that section of the road. The tower is easy to setup and it is energetically autonomous to provide long runs without human intervention or mains powering. Both Fixed and Mobile system use similar software, based on robotic computer vision techniques such as object recognition or structure from motion. The algorithms are trained with existing data from pictures of roadkilled amphibians. We have tested several solutions for classifying the images: SIFT (Scale-Invariant Feature Transform), SURF (Speeded Up Robust Features), BRIEF (Binary Robust Independent Elementary Features), ORB (Oriented and Rotated BRIEF), and Haar Cascade, which provided the best results. All algorithms are implemented in the OpenCV library. Both systems were developed using low cost components with the idea of saving funds, time and personal resources for wildlife preservation.