Research

From orchard perception to precise field action.

My dissertation develops and evaluates an AI-driven, vision-guided precision herbicide spraying system for sustainable weed management in apple orchards.

01

Orchard perception

Detecting, segmenting, tracking, and mapping weeds from a side-view camera in cluttered apple rows.

  • YOLOv7-CBAM segmentation
  • DeepSORT identity preservation
  • Pixel-grid density estimation
  • 0.82 MOTA · 88% IDF1
Open project page →
Orchard perception
02

Vision-guided spot spraying

Converting weed coordinates into nozzle-zone assignments, servo angles, timing compensation, and sequential spray commands.

  • Husky A200 platform
  • Three servo-steered nozzles
  • Ant-colony assisted scheduling
  • Dynamic hit evaluation
Open project page →
Vision-guided spot spraying
03

Weed morphology & dosing

Connecting plant size and growth habit to more appropriate treatment decisions through repeatable multiview imaging.

  • Dual-axis greenhouse gantry
  • Top and side ZED X Mini views
  • Marestail and lambsquarters
  • Size-dependent spray delivery
Open project page →
Weed morphology & dosing
04

Field integration

Bringing perception, navigation, target selection, and precision application together on a Farm-ng Amiga research platform.

  • ROS 2 and sensor fusion
  • ZED stereo vision and LiDAR
  • Orchard navigation
  • Orchard field testing
Open project page →
Field integration