Delivered live furrow detection and visual guidance, distinguishing working rows from tire tracks in field imagery
Delivered System
Find the working row in field imagery
We delivered a field vision system that detects agricultural furrows and distinguishes working rows from tire tracks in live video. The pipeline segments the usable field region, then combines LAB luminance, Gaussian smoothing, adaptive thresholding, and perspective-aware linear regression to estimate row direction. A yellow guide trajectory overlays the camera view, making the estimate clear to the operator. Protected camera installation supports image acquisition in the field. Testing covered changes in soil appearance, lighting, and moisture, alongside distracting tire tracks, with checks of row detection and temporal stability throughout the visual guidance workflow.
What we did on it...
Requirements
Defined working-row detection
We analyzed field imagery to distinguish usable agricultural furrows from distracting tire tracks. Reference frames guided the definition of row direction and usable image regions, establishing the visual information needed for a clear operator guidance overlay.
Structure
Implemented live row-direction analysis
We implemented field-region segmentation and processed LAB luminance with Gaussian smoothing and adaptive thresholding. Perspective-aware linear regression estimates furrow direction, and the pipeline processes continuous video to carry the detected row through successive frames of the camera view.
Interface
Delivered an interpretable guide trajectory
We added a yellow guide trajectory to the live camera view so operators can read the estimated row direction directly. The protected camera installation supports field acquisition, connecting the image analysis pipeline with an accessible visual guidance interface.
Validation
Tested guidance under field variations
We collected field imagery across changes in soil appearance, lighting, and moisture, including visible tire tracks. Testing checked row detection and temporal stability in live video, then reviewed the guide trajectory's readability through the completed camera and display workflow.
Other Projects
Metal Surface Inspection
Implemented visual inspection for metal surface defects, with synchronized part tracking and PLC rejection
Packaging Closure Inspection
Implemented multicamera inspection for open flaps and closure defects, with automatic diversion for rework
TA Quality Check
Delivered laser inspection for screw and pin presence, with flexible positioning and verified reject removal