A Quantitative Comparison of the Reliability of Animal Detection Systems and Recommended Requirements for System Reliability

Animal-vehicle collisions affect human safety, property, and wildlife, and the number of animal-vehicle collisions has been increasing in many regions across North America. For this project the authors investigated the reliability of nine different types of animal detection systems from five different manufacturers with regard to system reliability. These systems have the potential to improve human safety while not blocking or confining animal movements across the road. However, reliable warning signs are essential as the effectiveness of these systems depends on driver response. To investigate the reliability of the systems the authors constructed a controlled access test facility near Lewistown, Montana. The systems were installed to detect horses and llamas that roamed in an enclosure. The llamas and horses served as a model for wild ungulates. Data loggers recorded the date and time of each detection for each system. Animal movements were also recorded by six infrared cameras with a date and time stamp. By analyzing the images and the detection data, researchers were able to investigate the reliability for each system. The percentage of false positives (i.e., a detection is reported by a system but there is no large animal present in the detection zone) was relatively low for all systems (≤1%). The percentage of false negatives (i.e., an animal is present in the detection zone but a system failed to detect it) was highly variable (0-31%). The percentage of intrusions (i.e., animal intrusions in the detection area) that were detected varied between 73 and 100 percent. The results suggest that some animal detection systemsare quite reliable in detecting large mammals with few false positives and false negatives, whereas other systems have relatively many false negatives. The authors also surveyed three stakeholder groups—employees of transportation agencies,employees of natural resource management agencies, and the traveling public—with regard to their expectations on the reliability of animal detection systems. Based on the results from the survey, the researchers recommend the following performance requirements for the reliability of animal detection systems: (1) Animal detection systems should detect at least 91 percent of all large animals that approach the road; and (2) Animal detection systems should have fewer than 10 percent of all detections be false. The recommended reliability requirements of animal detection systems were compared to the results of the reliability tests. Five of the nine systems tested met the recommended requirements. The results of this study provide transportation and other agencies with the data to decide on minimum reliability requirements for animal detection systems. Furthermore, the data show that some of the systems tested are quite reliable and may be considered for implementation along a roadside where they can be investigated for their effectiveness in reducing collisions with large wild mammals. However, experiences with installation, operation and maintenance showed that the robustness of animal detection systems may have to be improved before the systems can be deployed on a large scale.