Automated Profiling Cutting of Salix Psammophila: A Vision Servo System Integrating YOLOv5n and U-Net for Stubble-Cutting Machine

Salix psammophila is pivotal for ecological protection in arid areas, capable of regeneration through stubble growth. Yet, in China, the automation of stubble equipment for this species lags, notably in profiling cutting tasks, leading to inconsistent stubble heights and adversely affecting sprouting rates. Addressing this, this paper presents a novel system integrating Raspberry Pi and visual imaging technologies, leveraging OpenCV, YOLOv5n, and a U-Net segmentation network, to automate profiling cutting operations via visual servo control. This system enhances the precision of mechanical stubble cutting, particularly vital during the optimal pre-stubble growth phase. Utilizing a dataset of images from Inner Mongolia's Kubuqi Desert, the system achieves <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 5. 4 \%}$</tex> precision in recognizing Salix psammophila and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 1. 8 \%}$</tex> MPA in segmenting root contour with UNet, significantly advancing the automation of stubble machinery and supporting sustainable ecological management in arid regions.