AN INSECT SWARM SEGMENTATION ALGORITHM BASED ON MASK R-CNN

Radar has become an important tool for continuously observing insect migration, measuring and recording the biological and behavioural information of aerial insects crossing beams at different heights. Insects usually actively choose their migration time and height according to the environment, gather in swarms at specific time and height ranges, and exhibit some similar behavioural characteristics. Extracting the insect swarm and combining them with information of individual insects in the swarms provided by the high-resolution insect radar to study insect migration is becoming a new research path for radar entomology. Therefore, this article proposes an intelligent insect swarm segmentation algorithm based on the difference of insect spatiotemporal distribution. Firstly, the spatiotemporal distributions of insects measured by high-resolution insect radar are presented in the form of heat maps. Then, the Mask R-CNN network is used to segment the image based on the aggregation degree of insect individuals to identify the aggregated insect swarm. Finally, the aggregated swarms are extracted by matching the original data based on the pixel coordinates of the segmentation results. The good performance of the proposed method was verified based on simulation and experimental data.