Comparative evaluation of different sensors for foliage detection to assist in intra-row weeding applications

Abstract

Accurate detection of plants foliage is essential criteria for the development of efficient intra-row weeding mechanism to prevent crop damage. This study evaluated the detection performance of three sensors-ultrasonic, infrared and camera, to identify the most suitable sensor to assist intra-row weeding applications. The sensors were tested under laboratory conditions at varying distances (30, 60 and 90 cm), sensor heights (10, 20, 30 and 40 cm) and angular positions (45°, 60°, 75° and 90°). Plants were classified as large, medium and small based on foliage density and height to assess sensor response under different canopy conditions. A custom sensor-holding setup was fabricated, and embedded programs were developed using an Arduino platform for ultrasonic and infrared sensors, while image acquisition and processing for the camera were carried out using OpenCV. Sensor performance was evaluated based on the number of successful detections across all test combinations. The camera sensor recorded the highest number of detections (84 out of 144), followed by the ultrasonic sensor (68) and the infrared sensor (47). The camera consistently detected plants across varying foliage densities due to its rich visual information and precise localization capability. Ultrasonic and infrared sensors showed reduced performance, particularly for small plants, due to weak signal reflection and sensitivity to environmental factors. The results indicate that camera-based sensing is the most reliable approach for foliage detection and is well suited for integration into automated intra-row weeding mechanisms.