The purpose of this study is developing a technology to support honeybee management in the beekeeping industry. We have been studying a method for detecting cells and classifying their states (larva, pupa, pollen, etc.) from honeycomb images using SSD. However, there were some problems, such as, misclassification of cells having similar features and detection miss of smaller larva. To solve these problems, we used Feature fused SSD (FSSD) that is able to use attributes of small objects by combining feature maps in our system. We evaluated its performance for the honeycomb cell classification task in this paper.

View full abstract

Synecoculture ™ is a method of farming that produces useful plants while making multifaceted use of the self organizing ability of the ecosystem by growing a wide variety of plants densely mixed in the same farmland. As a technology to support Synecoculture , robotics are being developed to automate major management tasks Still, the complexity of recognition and operation is imposing a heavy burden against automation compared with conventional farming that is based on a uniform operation of a single plant. On Synecoculture it is essential to grow plants with high diversity , but the dominance of some plants over other s may change the species composition and occupancy in the ecosystem which might result in reduce d diversity Pruning these excessively dominant plants is needed to maintain the balance of species composition in the vegetation of Synecoculture . In this study, we aim to detect such overly propagating plants that m ight reduce the diversity of the vegetation community (dominant plants). The proposed method detects the dominant plants using the Chopped Picture Method (CPM), a Convolutional Neural Network CNN learning method for segmenting RGB images. In this study, we treat Mentha suaveolens ( as one of the dominant plants to be detected and trained the CNN with three labels: “mint,”　“plants other than mint” and “others.” As a result, we obtained high accuracy segmentation in detecting the dominant plants, especially in distinguishing the plant group from the non plant group.

View full abstract

In order to monitor the health of cows from images taken by cameras installed on the ceiling of barns (barn images), we proposed a cow identification method using 3D models. Specifically, the method identifies cows by matching images generated from 3D models and barn images. However, the method is less accurate due to the following problems: (i) images generated by 3D models and barn images are from different domains, (ii) the method does not fully focus on spot pattern features, which is an important clue to identify cows. To consider domain differences and focus on the spot pattern features, we propose a method that combines domain adaptation with channel attention mechanism.

View full abstract

Synecoculture™ is a method of farming that produces useful plants while making multifaceted use of the self-organizing ability of the ecosystem by growing a wide variety of plants densely mixed in the same farmland. As a technology to support Synecoculture, robotics are being developed to automate major management tasks. Still, the complexity of recognition and operation is imposing a heavy burden against automation compared with conventional farming that is based on a uniform operation of a single plant. In Synecoculture, it is essential to cover the topsoil with vegetation. If the topsoil is exposed, it is necessary to introduce seeds and seedlings to fill the gap with vegetation. In this study, we aim to recognize the area of the bare soil surface with pixel-wise precision. In the proposed method, each pixel segments into two classes: “vegetation” or “no vegetation.” by applying semantic segmentation to RGB images with the Focal Loss function. By comparing accuracy with different values of parameters for the semantic segmentation, our approach showed that this method could achieve high accuracy with a relatively small number of images for training.

View full abstract

When tracking multiple dairy cows from sequential images of barns taken by ceiling-mounted cameras (barn images), it is difficult to obtain accurate motion trajectories due to lens distortion and changes in the angle of the camera. In this study, we aim to improve the accuracy of dairy cow tracking by applying distortion correction to barn images. Experiments confirmed that distortion correction is effective for improving the accuracy of dairy cow tracking and identification.

View full abstract

In this presentation I will introduce the latest NICT activities and future research plans. Description of the newly developed optical bench for the ground station and preliminary test results during the international experiments with the German Aerospace Center (DLR) OSIRISv1 payload is explained. Also, part of our future research topics in the field of site diversity and space optical communications with cubesats will be presented.

View full abstract

The ever-increasing demand for higher data rates communication services, has led to the innovation of new advanced technologies and techniques over communication systems with the ultimate goal to fulfil the requirements of the next generation networks (e.g. 5G and beyond). To address these requirements, Free space optics (FSO) systems have been recognized as a promising wireless interconnecting technology for high-capacity communication networks, ensuring data rates similar to those offered by optical fiber systems but at a fraction of its deployment cost. They combine both the advantages of high transmission capacity enabled by optical device technologies and the ease of deployment and mobility of wireless links. In this talk, we discuss the system requirements for a reliable and resilient wireless link, including the transceiver structures design, as well as the practical challenging issues facing the real deployment of these systems and to address them. We also introduce our next generation all-optical FSO terminal based on an intelligent optical beam stabilizer (OBS), utilizing a cost-effective voice-coil motors (VCM). Our proposed system can enable wider transceiver field of view (FOV), as well as reliable and high-capacity communication system, with lower cost and complexity.

View full abstract

This talk will provide an overview of the seamless access networks consisting of various types of transmission media with direct signal conversion for beyond 5G networks.

View full abstract