A road mobile mapping device for supervised classification of amphibians on roads
We present the classification results of a supervised algorithm of road images containing amphibians. We used a prototype of a mobile mapping system composed of a scanning system attached to a traction vehicle capable of recording road surface images at speed up to 30 km/h. We tested the algorithm in three test situations (two control and one real): with plastic models of amphibians; with dead specimens of amphibians; and with real specimens of amphibians in a road survey. The classification results of the algorithm changed among tests, but in any case, it was able to detect more than 80% of the amphibians (more than 90% in control tests). Unfortunately, the algorithm presented as well a high rate of false-positive detections, varying from 80% in the real test to 14% in the control test with dead specimens. The Mobile Mapping Systems (MMS) is ideal for passive surveys and can work by day or night. This is the first study presenting an automatic solution to detect amphibians on roads. The classification algorithm can be adapted to any animal group. Robotics and computer vision are opening new horizons for wildlife conservation.
KeywordsRobotics Computer vision Conservation Road ecology Mobile Mapping System
This work is funded by the Life LINES project LIFE14 NAT/PT/001081. Previous work was financed by FEDER Funds, through the Operational Programme for Competitiveness Factors—COMPETE, and by National Funds through FCT—Foundation for Science and Technology of Portugal, under the project PTDC/BIA/BIC/4296/2012: Roadkills – Intelligent systems for mapping amphibian mortality on Portuguese roads. NS is supported by an IF contract by FCT (IF/01526/2013). HR and MF are supported by research grants by Life LINES. CS and MF were supported by research grants by FCT (UMINHO/BI/172/2013 and UMINHO/BI/175/2013 respectively).
- Bossler J, Goad C, Johnson P, Novak K (1991) GPS and GIS map of the national highways. GeoInfo Systems Magazine, pp 26–37Google Scholar
- Bradski GR, Pisarevsky V (2000) Intel’s computer vision library: applications in calibration, stereo, segmentation, tracking, gesture, face and object recognition. IEEE Computer Vision and Pattern Recognition (CVPR) II, pp 796–797Google Scholar
- Carretero MA, Rosell C (2000) Incidencia del atropello de anfibios, reptiles y otros vertebrados en un tramo de carretera de construcción reciente. Bol Asoc Herpetol Esp 1:39–43Google Scholar
- El-Sheimy N (1996) A mobile multi-sensor system for GIS applications in urban centers. Int Arch Photogramm Remote Sens 31:95–100Google Scholar
- Franch M, Silva C, Lopes G, Ribeiro F, Trigueiros P, Seco L, Sillero N (2015) Where to look when identifying roadkilled amphibians? Acta Herpetol 10:103–110Google Scholar
- Freund Y, Schapire RE (1995) A decision-theoretic generalization of on-line learning and an application to boosting. European conference on computational learning theory. Springer, Berlin, pp 23–37Google Scholar
- Matos C, Sillero N, Argaña E (2013) Spatial analysis of amphibian road mortality levels in northern Portugal country roads. FrogLog 469:2013Google Scholar
- Petrie G (2010) An introduction to the technology mobile mapping systems. GEOInformatics Magazine 13:32–43Google Scholar
- Sialat M, Khlifat N, Bremond F, Hamrouni K (2009) People detection in complex scene using a cascade of boosted classifiers based on Haar-like Features. In: Proc. IEEE Int. Symposium on Intelligent Vehicles, pp 83–87Google Scholar
- Spellerberg IF (1998) Ecological effects of roads and traffic: a literature review. Glob Ecol Biogeogr:317–333Google Scholar
- Trigueiros P, Ribeiro F, Lopes G (2011) Vision-based hand segmentation techniques for human-robot interaction for real-time applications. VIPIMAGE, III Eccomas thematic conference on computational vision and medical image processingGoogle Scholar
- Viola P, Jones M (2001) Rapid object detection using a boosted cascade of simple features. In Computer vision and pattern recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE computer society conference 1: I–IGoogle Scholar