Stereo Vision System for Crop Height Measurement Using a Monocular Camera-Equipped Drone
Ryo Nakagawa, Yoshiki Yamaguchi, Hajime Nobuhara, Iman Firmansyah, Kai Sheng · 2024
This research presents a real-time crop height measurement system using monocular camera-equipped drones. The study defines the maximum disparity and frame rate prerequisites for stereo vision estimation for a flying drone. We achieve this by implementing image resizing and Semi-Global Matching using the Vitis Vision Library provided by AMD-Xilinx through high-level synthesis. Our custom-designed circuit was implemented on AMD-Xilinx’s SoC, an XCK26-SFVC784-2LV-C. Our experimentation with a 50k-pixel image measured the average computation latency for crop height measurement at 99.810 milliseconds. This latency comfortably meets our targeted frame rate of 10 frames per second.