Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Accurate crop phenotyping can help growers adjust their management with in-depth analysis. A sensor attached to trucks and tractors while working on the farm can generate recurring analyses, and can also be attached to smart machinery to both control chemical applications and data collection. A low-cost smart sensing system was designed using citrus as a case study. The prototype comprised a LiDAR, machine vision, GPS, flow meters, sensor fusion, and AI to scan trees for tree height, tree classification, and fruit counting. Using this information, the system controls an airblast tree crop sprayer, adjusting the spraying pattern based on the crop needs. The system was used and tested on commercial groves to further understand the requirements for use on the field.

This study evaluates the resolution of the lidar sensor between a low-end and high-end sensor on the capabilities to measure tree height and canopy density. Using the higher resolution lidar obtained slightly higher precision on height measurements (94% on low resolution, 95% on high resolution). The field-of-view (FOV) of the camera sensors is evaluated on the classification models capabilities for tree health condition. A convolutional neural network (CNN) was used to perform tree classification with an average accuracy of 84% in classifying the collected imagery into mature, young, dead, and non-tree objects. A larger FOV provided a similar accuracy for the models, but a higher reliability on different grove conditions, as it guarantees that the system can see the whole tree regardless of crop size and grove tree spacing.