Search

Yield prediction for crops is essential information for food security. A high-throughput phenotyping platform (HTPP) generates the data of the complete life cycle of a plant. However, the data are rarely used for yield prediction because of the lack of quality image analysis methods, yield data associated with HTPP, and the time-series analysis method for yield prediction. To overcome limitations, this study employed multiple deep learning (DL) networks to extract high-quality HTTP data, establish an association between HTTP data and the yield performance of crops, and select essential time intervals using machine learning (ML). The images of Arabidopsis were taken 12 times under environmentally controlled HTPP over 23 days after sowing (DAS). First, the features from images were extracted using DL network U-Net with SE-ResXt101 encoder and divided into early (15–21 DAS) and late (∼21–23 DAS) pre-flowering developmental stages using the physiological characteristics of the Arabidopsis plant. Second, the late pre-flowering stage at 23 DAS can be predicted using the ML algorithm XGBoost, based only on a portion of the early pre-flowering stage (17–21 DAS). This was confirmed using an additional biological experiment (P < 0.01). Finally, the projected area (PA) was estimated into fresh weight (FW), and the correlation coefficient between FW and predicted FW was calculated as 0.85. This was the first study that analyzed time-series data to predict the FW of related but different developmental stages and predict the PA. The results of this study were informative and enabled the understanding of the FW of Arabidopsis or yield of leafy plants and total biomass consumed in vertical farming. Moreover, this study highlighted the reduction of time-series data for examining interesting traits and future application of time-series analysis in various HTPPs.

The real challenge for separating leaf pixels from background pixels in thermal images is associated with various factors such as the amount of emitted and reflected thermal radiation from the targeted plant, absorption of reflected radiation by the humidity of the greenhouse, and the outside environment. We proposed TheLNet270v1 (thermal leaf network with 270 layers version 1) to recover the leaf canopy from its background in real time with higher accuracy than previous systems. The proposed network had an accuracy of 91% (mean boundary F1 score or BF score) to distinguish canopy pixels from background pixels and then segment the image into two classes: leaf and background. We evaluated the classification (segment) performance by using more than 13,766 images and obtained 95.75% training and 95.23% validation accuracies without overfitting issues. This research aimed to develop a deep learning technique for the automatic segmentation of thermal images to continuously monitor the canopy surface temperature inside a greenhouse.

Silicon promotes nodule formation in legume roots which is crucial for nitrogen fixation. However, it is very time-consuming and laborious to count the number of nodules and to measure nodule size manually, which led nodule characterization not to be study as much as other agronomical characters. Thus, the current study incorporated various techniques including machine learning to determine the number and size of root nodules and identify various root phenotypes from root images that may be associated with nodule formation with and without silicon treatment. Among those techniques, the machine learning for characterizing nodule is the first attempt, which enabled us to find high correlations among root phenotypes including root length, number of forks, and average link angles, and nodule characters such as number of nodules and nodule size with silicon treatments. The methods here could greatly accelerate further investigation such as delineating the optimal concentration of silicon for nodule formation.

A high-throughput plant phenotyping system automatically observes and grows many plant samples. Many plant sample images are acquired by the system to determine the characteristics of the plants (populations). Stable image acquisition and processing is very important to accurately determine the characteristics. However, hardware for acquiring plant images rapidly and stably, while minimizing plant stress, is lacking. Moreover, most software cannot adequately handle large-scale plant imaging. To address these problems, we developed a new, automated, high-throughput plant phenotyping system using simple and robust hardware, and an automated plant-imaging-analysis pipeline consisting of machine-learning-based plant segmentation. Our hardware acquires images reliably and quickly and minimizes plant stress. Furthermore, the images are processed automatically. In particular, large-scale plant-image datasets can be segmented precisely using a classifier developed using a superpixel-based machine-learning algorithm (Random Forest), and variations in plant parameters (such as area) over time can be assessed using the segmented images. We performed comparative evaluations to identify an appropriate learning algorithm for our proposed system, and tested three robust learning algorithms. We developed not only an automatic analysis pipeline but also a convenient means of plant-growth analysis that provides a learning data interface and visualization of plant growth trends. Thus, our system allows end-users such as plant biologists to analyze plant growth via large-scale plant image data easily.

The natural user interface using hand gesture have been popular field in Human-Computer-Interaction(HCI). Many research papers have been proposed in this field. They proposed vision-based, glove-based and depth-based approach for hand gesture recognition. However, hand gesture itself is simple and not natural way to interact. In otherwise, hand gesture recognition using finger tracking and identification can be implemented more robust and subtle recognition. Recently, new horizons are open with the development of sensors and technology such as Kinect and Depth-Sense. This development has made possible robust recognition, like finger identification and hand gesture recognition in bad conditions such as dark light and rough background as well. In this paper, we proposed a new finger identification and hand gesture recognition techniques with kinect depth data. Our proposed finger identification and gesture recognition methods provide natural interactions and interface by using fingers. We implemented interfaces and designed hand gestures using this method. This paper explains finger identification method and hand gesture recognition in detail. We show the preliminary experiment for evaluating accuracy of finger identification and hand gesture recognition accuracy. Finally, we discuss the result of evaluation and our contributions.

We present a new natural user interaction technique using finger gesture recognition and finger identification with Kinect depth data. We developed a gesture version drawing, multi-touch and mapping on 3d space interactions. We implemented three type interfaces using their interaction such as air-drawing, image manipulation and video manipulation. In this paper, we explain finger gesture recognition method, finger identification method and natural user interactions in detail. We show the preliminary experiment for evaluating accuracy of finger identification and finger gesture recognition accuracy, evaluating user questionnaire for interaction satisfaction. Finally, we discuss the result of evaluation and our contributions.