Vol. 85, No. 4
Abstract
Deer Behavior Recognition with Deep Learning
── Discussion of the Analysis Method for Camera Trapping Movie Using Body Part Trajectories ──
[Keywords: wildlife damage, image recognition, neural network, behavior analysis, YOLOv5]
Camera trapping is a method for analyzing wild animal behaviors and for wildlife damage control and ecosystem studies. Although this method can provide detailed information about animal behaviors, it still requires manual data confirmation, thereby making its analysis labor intensive. To automatically analyze camera trapping data, we created“BoxFlow images”in this study. These images depict loca- tions and trajectories of deer parts in movies. In addition, these images were input into a convolutional neural network, from which we obtained outputs of three categories, namely,“Moving”，“Eating”， and“Distance.”A cross-validation-like approach was employed for the classification evaluation. When deer parts were successfully detected, mean accuracies of the behavior classification were 0.789 for Moving, 0.862 for Eating, and 0.872 for Distance.
Detection of Lodging Rice Areas in Rice Fields by Semantic Segmentation via a Fisheye-lens Camera
[Keywords: agricultural robot, artificial intelligence, deep learning, rice lodging, semantic segmentation]
To detect lodging rice areas with a fisheye-lens camera, two semantic segmentation models based on deep neural networks were developed and tested with fisheye images and converted images. The two models could detect the lodging rice areas and unharvested rice areas in the front of the combine harvester. The two models were trained to detect the following seven classes, 1) unharvested rice ar- eas, 2) harvested rice areas, 3) lodging rice areas, 4) ridges, 5) humans, 6) other combine harvesters, and
7) background. Unharvested and lodging rice areas are the main focus of this study. Intersection over union (IoU) of these two classes was 0.8446 and 0.8524 respectively of the best model FCNResNet50, trained with both fisheye images and converted images.
Development and Performance Evaluation of Handheld Electrostatic Pollen Duster for Artificial Pollination of Orchard
[Keywords: orchard, artificial pollination, electrostatic, charge to mass ratio, adhesion]
In this study, we developed a handheld electrostatic pollen duster intended to decrease pollen consumption and save labor during artificial pollination of orchards.
The structure of the developed prototype was based on an existing pollen duster. We added a needle-shaped corona charge electrode near the nozzle, and connected variable high-voltage DC power supply.
The results showed that the charge to mass ratio of pollen and the number of charged pollen particles adhered to the artificial target increased linearly with the increase in electrode voltage. The number of charged pollen particles adhered to a pistil of‘Sumomoume’（Prunus salicina Lindl. × P. mume Sieb. et Zucc.）at an electrode voltage of －15 kV was eight times larger than that of non-charged.