Vol. 84, No. 3
Abstract
Real-Time Object Detection in Rice Field Images by Semantic Segmentation for Robotic Combine Harvester
── Pixel-wise Level Detection of Rice Lodging Existence ──
[Keywords: agricultural robot, artificial intelligence, deep learning, rice lodging, image cascade network]
To ensure safe and efficient operations using a robot combine harvester, seven semantic segmentation models were developed for pixel-wise level detection of the objects in a rice field. These models were trained and tested on four datasets. The results showed that all models performed well on the detection of rice field images. The pixel accuracy, class mean accuracy, mean intersection over union (IoU), IoU of lodging area ,and detection accuracy of loging existence of the best model were 0.9719, 0.8801, 0.8449, 0.6933, and 0.9448, respectively. The frame rate of the best model reached 14.04 frames per second (FPS), with an image size of 640×480 pixels on an embedded processor (Jetson TX2), which was fast enough to detect the objects in the rice field images.
Fish Volume Estimation and Accuracy of Multiple-Neck Underwater Helmholtz Resonators
[Keywords: Helmholtz resonator, multiple neck, fish, volume, inland aquaculture, feeding, density]
Inland aquaculture is spreading worldwide. As fish are a relatively inexpensive protein source and their demand is increasing, highly efficient fish production is necessary and thus, knowing fish volume inside aquaculture tanks is essential to prevent overstocking and overfeeding. In this experiment, a double cavity resonator with openings in resonator walls were used to explore the possibility of non-invasive and in-situ fish volume estimation. As reference and live samples, air to validate derived theoretical formula and Japanese dace fish (Tribolon hakonensis) for proof of concept were used. A high correlation between the actual and estimated volume was obtained with R-squared (R 2) value of 0.994. Finally, a fish volume estimation model was established that had a high correlation with R 2 value of 0.985.
Dynamic Instability Assessment of Nonlinear Tractor Dynamics based on Lyapunov Exponents
[Keywords: tractor, nonlinear dynamics, bouncing, Lyapunov exponents, dynamic instability]
Nonlinear phenomena in tractor dynamics, such as impact and sliding, can cause dynamic instability of tractors and lead to overturning accidents. Tractor static stability has been evaluated using static stability indicators, such as the static rollover angle. However, tractor dynamic stability assessments have not yet been established. Nonlinear time series analyses generally use Lyapunov exponents as dynamic stability indicators. We used Lyapunov exponents calculated from the vertical acceleration generated by a nonlinear tractor model to assess tractor dynamic stability. The parametric study results indicated that the Lyapunov exponent could detect nonlinear resonance regions and be used to assess tractor dynamic stability.
Construction of Experimental Track for Reproducing Accident Conditions Using Actual Tractor and Measurement System on Tractor Behavior （Part 1）
[Keywords: tractor, track, motion capture, slope, dropping accident, side-overturning]
The object of this study is to develop an experimental track for reproducing accident conditions using an actual tractor and measurement system on tractor behavior employing the optical motion capture system. The platform can reproduce two accident situations: 1） side-overturning under where the upper side wheel travels toward an obstacle on a slope. 2） Side-overturning from deviating from an operating road and dropping. The results of an overturning tests using a remote controlled 7.7
Construction of Experimental Track for Reproducing Accident Conditions Using Actual Tractor and Measurement System on Tractor Behavior （Part 2）
[Keywords: tractor, track, motion capture, slope, dropping accident, side-overturning]
This study establishes a method for measuring the dynamic behavior of a tractor using the experimental track and measuring system described in Part 1 of this report. On the motion capture software, a stable measurement of the behavior of the center of gravity can be ensured by placing the virtual point at an intermediate position between reflective markers, which were fitted on the right and left side of the center of gravity of the tractor. The motion capture has measurement values with precision equivalent to that of an inertial measurement unit about vertical acceleration and 3-angle angular velocity. It is possible to measure the dynamic behavior and torsion of the front and rear axles and the center of gravity.