Search
A real-time smart sensing system for automatic localization and recognition of vegetable plants for weed control
Sources of information

Frontiers in Plant Science | Mar 27, 2023

Researchers at China Agricultural University have developed a smart weeder to combat weed threats in tomato production. Weeds pose a significant challenge, particularly during the early stages of tomato plant growth. To address this, the team created an integrated sensing system using cameras and color mark sensors. The system accurately locates tomato and pakchoi plants in real time, crucial for effective weed management.

Through experiments, the researchers determined that using the main stem of tomato plants as a reference yielded better results than pakchoi. Applying white markers on the lower main stem of tomato plants proved optimal. The system, equipped with six sensors, demonstrated high efficiency in detecting plant labels. A specially designed computer vision algorithm achieved an impressive overall accuracy of 95.19% in localizing tomato and pakchoi plants.

This sensor-based system offers a reliable and precise solution for automatic real-time localization of vegetable plants, enabling effective weed control in agriculture.

*
(A) Image of tomato plant and weeds captured by the system, (B) binary image after morphological operations, (C) binary image with plants contours (yellow), (D) position of the crop signal in the RGB image, (E) position of crop signal in the binary image, (F) crop signal tolerance bands in the binary image, (G) the contours of plants (yellow) and the crop signal tolerance bands (red) in the binary image, (H) the contours of tomato plant (red) and the contours of weeds (green), (I) pixel mapping of weeds (green) and crop plants (red).

 

TOP