Advancing Sika deer detection and distance estimation through comprehensive camera calibration and distortion analysis

Abstract

Commercial camera traps are widely used in global wildlife monitoring, but their effectiveness is often compromised by oversights in specifications and technical considerations. This study evaluates the performance of three camera trap models, including the Solar Powered 4K-trail, HC-801A-Pro and HC-801A, by analysing their resolution limits and lens distortion correction capabilities. To determine the resolution limits, A4-sized templates with red circles of varying diameters were placed within the cameras’ field of view at various distances, with measurements taken using a measuring tape and verified with GPS. Using the cv2.TM_CCOEFF_NORMED template matching algorithm with templates scaled from 0.01 to 2.0 and a confidence threshold of 0.6, the resolution threshold was defined as the distance at which the observed circle size deviated from the expected size. Among the models, the Solar Powered 4K Trail camera had the highest resolution threshold at 17.29 m, while the HC 801A had the lowest at 15.3 m. Lens distortion coefficients were derived by analysing checkerboard pattern images taken at different distances and angles. All three camera models exhibited lens distortion. The Solar Powered 4K-trail and HC-801A-Pro exhibited barrel distortion, while the HC-801A exhibited pincushion distortion. The calculated coefficients successfully corrected these distortions, improving image accuracy. The derived coefficients effectively corrected these distortions, improving image accuracy. This practical and reproducible calibration method, which does not require expensive optical equipment, offers significant improvements in camera trap optimisation, enabling conservationists and researchers to obtain more reliable ecological data.