CPISNet: Delving into Consistent Proposals of Instance Segmentation Network for High-Resolution Aerial Images

Instance segmentation of high-resolution aerial images is challenging when compared to object detection and semantic segmentation in remote sensing applications. It adopts boundary-aware mask predictions, instead of traditional bounding boxes, to locate the objects-of-interest in pixel-wise. Meanwhile, instance segmentation can distinguish the densely distributed objects within a certain category by a different color, which is unavailable in semantic segmentation. Despite the distinct advantages, there are rare methods which are dedicated to the high-quality instance segmentation for high-resolution aerial images. In this paper, a novel instance segmentation method, termed consistent proposals of instance segmentation network (CPISNet), for high-resolution aerial images is proposed. Following top-down instance segmentation formula, it adopts the adaptive feature extraction network (AFEN) to extract the multi-level bottom-up augmented feature maps in design space level. Then, elaborated RoI extractor (ERoIE) is designed to extract the mask RoIs via the refined bounding boxes from proposal consistent cascaded (PCC) architecture and multi-level features from AFEN. Finally, the convolution block with shortcut connection is responsible for generating the binary mask for instance segmentation. Experimental conclusions can be drawn on the iSAID and NWPU VHR-10 instance segmentation dataset: (1) Each individual module in CPISNet acts on the whole instance segmentation utility; (2) CPISNet* exceeds vanilla Mask R-CNN 3.4%/3.8% AP on iSAID validation/test set and 9.2% AP on NWPU VHR-10 instance segmentation dataset; (3) The aliasing masks, missing segmentations, false alarms, and poorly segmented masks can be avoided to some extent for CPISNet; (4) CPISNet receives high precision of instance segmentation for aerial images and interprets the objects with fitting boundary.

Application of Adaptive MOMEDA with Iterative Autocorrelation to Enhance Weak Features of Hoist Bearings

Low-speed hoist bearings are characterized by fault features that are weak and difficult to extract. Multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) is an effective method for extracting periodic pulses in a signal. However, the decomposition effect of MOMEDA largely depends on the selected pulse period and filter length. To address these drawbacks of MOMEDA and accurately extract features from the vibration signal of a hoist bearing, an adaptive feature extraction method is proposed based on iterative autocorrelation (IAC) and MOMEDA. To automatically identify the pulse period, a new evaluation index named autocorrelation kurtosis entropy (AKE) was constructed to select the optimal IAC. To eliminate the influence of the filter length on the decomposition effect, an iterative MOMEDA strategy was designed to gradually enhance signal impulse features. The Case Western Reserve University bearing dataset and bearing data from a self-made hoisting test setup were used to verify the effectiveness of IAC-MOMEDA in extracting weak features. Moreover, the capability of IAC-MOMEDA for features extraction of normal bearing vibration signal was further confirmed by field test data.

Adaptive feature extraction and fault diagnosis for three-phase inverter based on hybrid-CNN models under variable operating conditions

The increasing reliability and availability requirements of power electronic systems have drawn great concern in many industrial applications. Aiming at the difficulty in fault characteristics extraction and fault modes classification of the three-phase full-bridge inverter (TFI) that used as the drive module of brushless DC motor (BLDCM). A hybrid convolutional neural network (HCNN) model consists of one-dimensional CNN (1D-CNN) and two-dimensional CNN (2D-CNN) is proposed in this paper, which can tap more effective spatial feature for TFI fault diagnosis. The frequency spectrum from the three-phase current signal preprocess are applied as the input for 1D-CNN and 2D-CNN to conduct feature extraction, respectively. Then, the feature layers information are combined in the fully connected layer of HCNN. Finally, the performance status of TFI could be identified by the softmax classifier with Adam optimizer. Several groups of experiments have been studied when the BLDCM under different operating conditions. The results show that the fusion features can get a higher degree of discrimination so as to the presented network model also obtains better classification accuracy, which verify the feasibility and superiority to the other networks.

Rolling Bearing Feature Extraction Method Based on Improved Intrinsic Time-Scale Decomposition and Mathematical Morphological Analysis

To overcome the difficulty of extracting the feature frequency of early bearing faults, this paper proposes an adaptive feature extraction scheme. First, the improved intrinsic time-scale decomposition, proposed in this paper, is used as a noise reduction method. Then, we use the adaptive composite quantum morphology analysis method, also proposed in this paper, to perform an adaptive demodulation analysis on the signal, and finally, extract the fault characteristics in the envelope spectrum. The experimental results show that the scheme performs well in the early fault feature extraction of rolling bearings.

A Parameter-Optimized DBN Using GOA and Its Application in Fault Diagnosis of Gearbox

Aiming at the problems of poor self-adaptive ability in traditional feature extraction methods and weak generalization ability in single classifier under big data, an internal parameter-optimized Deep Belief Network (DBN) method based on grasshopper optimization algorithm (GOA) is proposed. First, the minimum Root Mean Square Error (RMSE) in the network training is taken as the fitness function, in which GOA is used to search for the optimal parameter combination of DBN. After that the learning rate and the number of batch learning in DBN which have great influence on the training error would be properly selected. At the same time, the optimal structure distribution of DBN is given through comparison. Then, FFT and linear normalization are introduced to process the original vibration signal of the gearbox, preprocess the data from multiple sensors and construct the input samples for DBN. Finally, combining with deep learning featured by powerful self-adaptive feature extraction and nonlinear mapping capabilities, the obtained samples are input into DBN for training, and the fault diagnosis model for gearbox based on DBN would be established. After several tests with the remaining samples, the diagnosis rate of the model could reach over 99.5%, which is far better than the traditional fault diagnosis method based on feature extraction and pattern recognition. The experimental results show that this method could effectively improve the self-adaptive feature extraction ability of the model as well as its accuracy of fault diagnosis, which has better generalization performance.