Discriminating crop, weeds and soil surface with a terrestrial LIDAR sensor

Area of Science:

• Agricultural Engineering
• Remote Sensing
• Precision Agriculture

Background:

• Previous research utilized only height profiles from LIDAR for vegetation analysis.
• Current study integrates LIDAR distance and reflection measurements for enhanced vegetation discrimination.

Purpose of the Study:

• To evaluate the accuracy and performance of a LIDAR sensor for detecting and discriminating maize plants and weeds from soil.
• To assess a combined approach using LIDAR height and reflection data.

Main Methods:

• Terrestrial LIDAR sensor deployed in a maize field (growth stage 12-14).
• Data acquisition included distances and reflection measurements.
• Plant heights estimated via digital images; correlation with LIDAR height (R² = 0.75).
• Binary logistic regression and Canonical Discrimination Analysis (CDA) used for validation.

Main Results:

• High correlation (R² = 0.75) between LIDAR-measured and actual plant heights.
• Up to 95% accuracy in discriminating vegetation from the ground.
• CDA effectively distinguished soil from vegetation, with limited crop-weed differentiation.

Conclusions:

• The LIDAR-based methodology shows promise for effective weed detection in agricultural settings.
• Integration with vision-based technologies can further enhance precision herbicide spraying.