Computer Vision for Driverless Tractors

An intelligent computer vision system is one of the most important sensing subsystems of an autonomous farming vehicle, which first of all allows it to respond to unexpected changes in the field.

Automation of agriculture vehicles promises to improve productivity and safety and, at the same time, reduce operational costs by minimizing the labor and maintenance needed. Like other autonomous vehicles, tractors observe their position using several digital cameras, and decide speed and adjust direction without a driver in the cabin. Our expertise in computer vision and deep-learning technologies allowed us to build a robust solution, capable of segmenting the scene and detecting obstacles in real time.

The visual control system includes:

• Stereo vision subsystem, which calculates distances
• Scene segmentation, which recognizes visual patterns
• Controller, which calculates actual course

There are two groups of positioning systems that control the vehicle: absolute positioning, such as GPS, magnetic compass, etc, and relative positioning, which detects local cues, crop lines, and obstacles. BitRefine group offers solutions for relative positioning system for farming vehicles. The base technology is automatic scene segmentation, which detects rows and crop lines to perform accurate navigation. The computer vision system also has object detection and a classification subsystem, which helps the tractor to avoid collision with other vehicles, humans, or animals. Apart from navigation technologies, BitRefine group offers solutions for automatic weed detection and plant disease detection, which can bring further cost reduction and efficiency improvement.