This experiment studies the feasibility of tuber yield prediction in cassava fields using multispectral imagery based on unmanned arial vehicle. The imageries of a cassava field were taken monthly, four times. The cassava’s height, normalized difference vegetation index (NDVI), simple ratio vegetation index (RVI), and chlorophyll vegetation index (CIRedEdge) were calculated. Yield models were developed using Simple linear regression with vegetation indices (VIs), canopy area, and average height from 3 methods: excluded soil pixels (1), zero soil pixels (2), and included soil pixels (3). The results show the average height and canopy area from method (1) provides the highest R2 0.87 and 0.65. VIs values from method (3) gives R2 0.58, 0.57, and 0.50 for NDVI, CIRedEdge, and RVI.
In Japan’s tractor accidents, overturning is the main cause, which need to be prevented. We proposed a tractor behaviour model considering the effect of driving force on pitch angle and compared the simulation results using a conventional model. The accident prevention effect by driving force control was verified by a simulation using topographic information from a tractor overturn accident site. Simulation results indicated that the proposed model is suitable in analysing steep slopes, and driving force of the control model was smaller than that of the non-control model. These results clarified the need for a proposed model to assess pitching overturning risk and suggested that the attitude angle can be suppressed by controlling the driving force for preventing overturning accidents.
In this study, we aimed to develop a simple measurement method for estimating the firmness of foods, based on spatially resolved diffuse reflectance measurement. Apples stored for different periods were measured, and the gradient of the intensity profiles was shown to gradually decrease with storage. To estimate apple firmness from the intensity profiles, the coefficients of the theoretical formula and the gradients of the intensity profiles calculated at 1.0 mm intervals were used. As a result, the firmness could be estimated at a practical level with RMSE less than 1.5 N, which was more precise than sensory evaluation. Compared to other non-destructive methods used for texture estimation, the method proposed in this study is advantageous in terms of easy implementation.
Adulterating extra-virgin olive oils (EVOO) with lower grade olive oils, like virgin olive oil (VOO), and selling it as EVOO to unsuspecting consumers has sparked concern in the recent years. Developing inexpensive and quick adulteration detection methods to unravel such acts will promote trust in the industry. This study focused on the quality degradation of EVOO when adulterated by different proportions of VOO. Excitation emission matrices (EEMs) and fluorescence images were taken for analysis. Partial least square regression (PLSR), support vector machine (SVM), decision tree and convolutional neural network (CNN) models were used to explore both the EEMs and fluorescence images of adulterated oils, which indicate the extent of adulteration of extra virgin olive oils can be detected.