Detection and Classification of Rice Infestation with Rice Leaf Folder (Cnaphalocrocis medinalis) Using Hyperspectral Imaging Techniques

The detection of rice leaf folder (RLF) infestation usually depends on manual monitoring, and early infestations cannot be detected visually. To improve detection accuracy and reduce human error, we use push-broom hyperspectral sensors to scan rice images and use machine learning and deep neural learning methods to detect RLF-infested rice leaves. Different from traditional image processing methods, hyperspectral imaging data analysis is based on pixel-based classification and target recognition. Since the spectral information itself is a feature and can be considered a vector, deep learning neural networks do not need to use convolutional neural networks to extract features. To correctly detect the spectral image of rice leaves infested by RLF, we use the constrained energy minimization (CEM) method to suppress the background noise of the spectral image. A band selection method was utilized to reduce the computational energy consumption of using the full-band process, and six bands were selected as candidate bands. The following method is the band expansion process (BEP) method, which is utilized to expand the vector length to improve the problem of compressed spectral information for band selection. We use CEM and deep neural networks to detect defects in the spectral images of infected rice leaves and compare the performance of each in the full frequency band, frequency band selection, and frequency BEP. A total of 339 hyperspectral images were collected in this study; the results showed that six bands were sufficient for detecting early infestations of RLF, with a detection accuracy of 98% and a Dice similarity coefficient of 0.8, which provides advantages of commercialization of this field.

Automated band selection for Bayesian FFT modal identification Based on RetinaNet

Bayesian FFT algorithm is a popular method to identify modal parameters, e.g., modal frequencies, damping ratios, and mode shapes, of civil structures under operational conditions. It is efficient and provides the identification uncertainty in terms of posterior distribution. However, in utilizing the Bayesian FFT algorithm, it is tedious to manually select frequency bands and initial frequencies. This step requires professional knowledge and costs most of time, which prevents the automation of Bayesian FFT algorithm. Regarding the band selection as an object detection problem, we design a band selection network based on the RetinaNet to automatically select frequency bands and a peak prediction network to predict the initial frequencies. The designed networks are trained using the singular value spectrum of measured ambient vibration data and verified by various data sets. It can achieve the human accuracy with much less operation time, and thus provides a corner stone for the automation of Bayesian FFT algorithm.

Novel Bearing Fault Diagnosis Using Gaussian Mixture Model-Based Fault Band Selection

This paper proposes a Gaussian mixture model-based (GMM) bearing fault band selection (GMM-WBBS) method for signal processing. The proposed method benefits reliable feature extraction using fault frequency oriented Gaussian mixture model (GMM) window series. Selecting exclusively bearing fault frequency harmonics, it eliminates the interference of bearing normal vibrations in the lower frequencies, bearing natural frequencies, and the higher frequency contents that prove to be useful only for anomaly detection but do not provide any insight into the bearing fault location. The features are extracted from time- and frequency- domain signals that exclusively contain the bearing fault frequency harmonics. Classification is done using the Weighted KNN algorithm. The experiments performed with the data containing the vibrations recorded from artificially damaged bearings show the positive effect of utilizing the proposed GMM-WBBS signal processing to filter out the discriminative data of uncertain origin. All comparison methods retrofitted with the proposed method demonstrated classification performance improvements when provided with vibration data with suppressed bearing natural frequencies and higher frequency contents.