Agricultural and Biological Engineering

Research and development of a 3-axis robotic strawberry harvester in a greenhouse using image processing

2025-05-15T21:25:31+07:00

This research focuses on the development of a robotic strawberry harvester for greenhouse cultivation, aimed at reducing dependence on manual labor and improving harvesting precision. The robotic system features a three-dimensional Cartesian structure with dimensions of 150 × 200 × 140 cm (Width × Length × Height), driven by three stepper motors with a step angle of 1.8^o. The system is controlled using a single web camera with a resolution of 1280 × 720 pixels (720p/30fps), which captures images to detect the position and ripeness of strawberries. Image processing is performed using LabVIEW software, which guides the movement along the X, Y, and Z axes to position the gripper precisely at the harvesting location. The gripper is designed to grip and cut the stem of the strawberry, minimizing post-harvest damage to the fruit. Field tests conducted in a greenhouse demonstrated that the robot could accurately classify strawberry ripeness with 100% accuracy. At motor speeds of 1100, 1200, and 1300 rpm corresponding to linear speeds of 0.11, 0.12, and 0.13 m/s the robot achieved target positioning accuracy of 100, 96.66, and 100%, respectively. The average harvesting times per fruit were 53.9, 52.7, and 50.8 s. In addition to its technical performance, an engineering economic analysis showed that the robotic system offers a payback period of 9.6 months and a break-even point at 180.59 h/time/y. These results indicate that the robotic harvesting system is a cost-effective investment for medium- to large-scale greenhouse farming operations.

Hourly adjustment factor analysis using the Kalman Filter technique to reduce errors in Sattahip radar rainfall estimation

2025-06-23T13:23:12+07:00

Weather radar can continuously measure rainfall immediately as it occurs, covering wide areas and providing high-resolution rainfall both spatially and temporally. However, in radar rainfall estimation, even when using Z-R relationships that vary according to rainfall clusters based on rainfall intensity measured from ground-based automatic rain gauges to reduce estimation errors, there remain discrepancies in adjusting radar rainfall measurements above ground to match actual ground rainfall. Additionally, each rainfall event has different raindrop distribution characteristics, resulting in varying physical characteristics across different rainfall events. This study collected data from 510 rainfall events between February 2, 2018, and August 31, 2020, comprising hourly rainfall from 110 automatic rain gauges and radar reflectivity within a 240 km radius of the Sattahip radar. The aim was to analyze rainfall estimation methods using Z-R relationships that vary by rainfall cluster according to rainfall intensity, combined with appropriate hourly adjustment factors for the Sattahip radar. The study found that the rainfall estimation method using Z-R relationships that vary by rainfall cluster according to rainfall intensity, combined with hourly adjustment factors analyzed using the Kalman Filter technique, most effectively reduced the errors in adjusting radar rainfall above ground to match ground rainfall. Considering the RMSE values, the method can reduce the rainfall estimation error by 21.53%, 4.10% (21.87%,5.19%). Based on the MSE values, it can reduce the estimation error by 47.71%, 8.37% (48.50%, 10.63%). Similarly, based on the MAE values, it can reduce the estimation error by 29.05%, 4.55% (31.32%, 6.28%) for the rainfall events used for calibration (Verification), compared to the rainfall estimation methods based on the current Z-R relationship, and the Z-R relationship that varies according to rainfall type and intensity without adjustment.

Investigating the capability of near-infrared spectroscopy to measure cassava tuber deterioration levels

2025-06-20T21:17:31+07:00

This research aims to develop a model for assessing the deterioration level of cassava roots using near-infrared spectroscopy techniques. The study focuses on evaluating the capability of measuring the deterioration level of cassava roots from the Kasetsart 50 variety, harvested from a cassava field at ages of 9 and 12 months after planting, with a total of 42 roots. Additionally, samples were collected from a cassava field in Dunsat Subdistrict, Kranuan District, Khon Kaen Province, at the age of 10 months, totaling 21 roots. The model was constructed by scanning within the wavelength range of 500-1100 nm, using a detector installed on the side and connected to the Mini-NIR spectrometer. The data obtained were analyzed for physical and chemical properties over different storage periods. The analysis included brightness (L), color intensity, dry matter content (DMC), and starch content (SC) using ANOVA test. Significant differences were found at a confidence level of 95% for L, a*, b*, SC, and DMC, while the color intensity b* showed no significant difference. The spectroscopic measurements with the NIR Spectrometer indicated important changes in properties. In developing the K-Nearest Neighbors (KNN) model, it was found that using the raw spectrum yielded the highest accuracy at 69%, reflecting the ability to predict the deterioration of cassava roots in the future. This study not only contributes significantly to the knowledge of cassava root deterioration but also provides a recommendation for developing techniques to assess the quality and freshness of agricultural products in the future.