Root Location and Root Diameter Estimation of Trees Based on Deep Learning and Ground-Penetrating Radar

A full understanding of the growth and distribution tree roots is conducive to guiding precision irrigation, fertilization, other agricultural work during production. Detecting with a ground-penetrating radar repeatable detection method that does no harm earth surface roots. In this research, rapid accurate automatic was conducted on hyperbolic waveforms formed by root targets in B-scan images based YOLOv5s. Following this, regions interest containing target hyperbolas were generated. Three or more coordinate points hyperbola selected according three-point fixed circle (TPFC) locate system estimate diameter. The results show accuracy using YOLOv5s 96.7%, recall rate 86.6%, time single image only 13 ms. simulation image, TPFC used diameter through three different frequency antennas (500 MHz, 750 1000 MHz). result obtained when antenna average distance error positioning being 3.17 cm, slope R2 linear fitting between estimated actual one 1.029 0.987, respectively. Verified pre-buried test wilderness field test, both localization estimation our research proved gain good conform rules found experiments. Therefore, we believe can quickly accurately detect system, diameter, provide new perspective for non-destructive three-dimensional reconstruction system.