Stereo camera trap for wildlife in situ observations and measurements.

This paper proposes a stereo camera trap to expand field of view (FOV) of the traditional camera trap and to measure wildlife sizes with a centimeter-scaled accuracy within the detection distance of 10 m. In the method, FOVs of the two cameras are partly overlapped with a 30-cm-long baseline and a posture angle of 100°. Typically only targets in the public FOV can be measured; in contrast, when only parts of targets appear in the public FOV they are difficult to measure. To solve the problem, a part-matching algorithm is provided. In the proposed camera trap, a central process unit is realized by a micro control unit, an advanced reduced-instruction-set-computing machine, and a field-programmable gate array, and then motion sensors trigger the cameras to capture stereo images when animals pass by. In addition, the camera trap has daytime mode and nighttime mode switched by a photosensitive sensor by perceiving ambient lights. Finally, the stereo camera trap data is transmitted by a long-term-evolution module at a scheduled time. Experimental results show that the proposed stereo camera trap can broaden the FOV of a monocular camera by up to 77% at 5 m and estimate feature sizes of targets with centimeter-scaled accuracy.