As an important rice disease, rice bacterial leaf blight (RBLB, caused by the bacterium Xanthomonas oryzae pv.oryzae), has become widespread in east China in recent years. Significant losses in rice yield occurred as ...As an important rice disease, rice bacterial leaf blight (RBLB, caused by the bacterium Xanthomonas oryzae pv.oryzae), has become widespread in east China in recent years. Significant losses in rice yield occurred as a result ofthe disease’s epidemic, making it imperative to monitor RBLB at a large scale. With the development of remotesensing technology, the broad-band sensors equipped with red-edge channels over multiple spatial resolutionsoffer numerous available data for large-scale monitoring of rice diseases. However, RBLB is characterized by rapiddispersal under suitable conditions, making it difficult to track the disease at a regional scale with a single sensorin practice. Therefore, it is necessary to identify or construct features that are effective across different sensors formonitoring RBLB. To achieve this goal, the spectral response of RBLB was first analyzed based on the canopyhyperspectral data. Using the relative spectral response (RSR) functions of four representative satellite or UAVsensors (i.e., Sentinel-2, GF-6, Planet, and Rededge-M) and the hyperspectral data, the corresponding broad-bandspectral data was simulated. According to a thorough band combination and sensitivity analysis, two novel spectralindices for monitoring RBLB that can be effective across multiple sensors (i.e., RBBRI and RBBDI) weredeveloped. An optimal feature set that includes the two novel indices and a classical vegetation index was formed.The capability of such a feature set in monitoring RBLB was assessed via FLDA and SVM algorithms. The resultdemonstrated that both constructed novel indices exhibited high sensitivity to the disease across multiple sensors.Meanwhile, the feature set yielded an overall accuracy above 90% for all sensors, which indicates its cross-sensorgenerality in monitoring RBLB. The outcome of this research permits disease monitoring with different remotesensing data over a large scale.展开更多
Human Activity Recognition (HAR) is an important way for lower limb exoskeleton robots to implement human-computer collaboration with users. Most of the existing methods in this field focus on a simple scenario recogn...Human Activity Recognition (HAR) is an important way for lower limb exoskeleton robots to implement human-computer collaboration with users. Most of the existing methods in this field focus on a simple scenario recognizing activities for specific users, which does not consider the individual differences among users and cannot adapt to new users. In order to improve the generalization ability of HAR model, this paper proposes a novel method that combines the theories in transfer learning and active learning to mitigate the cross-subject issue, so that it can enable lower limb exoskeleton robots being used in more complex scenarios. First, a neural network based on convolutional neural networks (CNN) is designed, which can extract temporal and spatial features from sensor signals collected from different parts of human body. It can recognize human activities with high accuracy after trained by labeled data. Second, in order to improve the cross-subject adaptation ability of the pre-trained model, we design a cross-subject HAR algorithm based on sparse interrogation and label propagation. Through leave-one-subject-out validation on two widely-used public datasets with existing methods, our method achieves average accuracies of 91.77% on DSAD and 80.97% on PAMAP2, respectively. The experimental results demonstrate the potential of implementing cross-subject HAR for lower limb exoskeleton robots.展开更多
基金the Strategic Priority Research Program of the Chinese Academy of Sciences(Grant No.XDA28010500)National Natural Science Foundation of China(Grant Nos.42371385,42071420)Zhejiang Provincial Natural Science Foundation of China(Grant No.LTGN23D010002).
文摘As an important rice disease, rice bacterial leaf blight (RBLB, caused by the bacterium Xanthomonas oryzae pv.oryzae), has become widespread in east China in recent years. Significant losses in rice yield occurred as a result ofthe disease’s epidemic, making it imperative to monitor RBLB at a large scale. With the development of remotesensing technology, the broad-band sensors equipped with red-edge channels over multiple spatial resolutionsoffer numerous available data for large-scale monitoring of rice diseases. However, RBLB is characterized by rapiddispersal under suitable conditions, making it difficult to track the disease at a regional scale with a single sensorin practice. Therefore, it is necessary to identify or construct features that are effective across different sensors formonitoring RBLB. To achieve this goal, the spectral response of RBLB was first analyzed based on the canopyhyperspectral data. Using the relative spectral response (RSR) functions of four representative satellite or UAVsensors (i.e., Sentinel-2, GF-6, Planet, and Rededge-M) and the hyperspectral data, the corresponding broad-bandspectral data was simulated. According to a thorough band combination and sensitivity analysis, two novel spectralindices for monitoring RBLB that can be effective across multiple sensors (i.e., RBBRI and RBBDI) weredeveloped. An optimal feature set that includes the two novel indices and a classical vegetation index was formed.The capability of such a feature set in monitoring RBLB was assessed via FLDA and SVM algorithms. The resultdemonstrated that both constructed novel indices exhibited high sensitivity to the disease across multiple sensors.Meanwhile, the feature set yielded an overall accuracy above 90% for all sensors, which indicates its cross-sensorgenerality in monitoring RBLB. The outcome of this research permits disease monitoring with different remotesensing data over a large scale.
文摘Human Activity Recognition (HAR) is an important way for lower limb exoskeleton robots to implement human-computer collaboration with users. Most of the existing methods in this field focus on a simple scenario recognizing activities for specific users, which does not consider the individual differences among users and cannot adapt to new users. In order to improve the generalization ability of HAR model, this paper proposes a novel method that combines the theories in transfer learning and active learning to mitigate the cross-subject issue, so that it can enable lower limb exoskeleton robots being used in more complex scenarios. First, a neural network based on convolutional neural networks (CNN) is designed, which can extract temporal and spatial features from sensor signals collected from different parts of human body. It can recognize human activities with high accuracy after trained by labeled data. Second, in order to improve the cross-subject adaptation ability of the pre-trained model, we design a cross-subject HAR algorithm based on sparse interrogation and label propagation. Through leave-one-subject-out validation on two widely-used public datasets with existing methods, our method achieves average accuracies of 91.77% on DSAD and 80.97% on PAMAP2, respectively. The experimental results demonstrate the potential of implementing cross-subject HAR for lower limb exoskeleton robots.
