Algebra Based Human Skeleton Sequence Reduction and Action Recognition

Skeleton-based human action recognition is a research hotspot in recent years, but most of the research focuses on the spatio-temporal feature extraction by convolutional neural network. In order to improve the correct recognition rate of these models, this paper proposes three strategies: using algebraic method to reduce redundant video frames, adding auxiliary edges into the joint adjacency graph to improve the skeleton graph structure, and adding some virtual classes to disperse the error recognition rate. Experimental results on NTU-RGB-D60, NTU-RGB-D120 and Kinetics Skeleton 400 databases show that the proposed strategy can effectively improve the accuracy of the original algorithm.

Enhancing the Tracking of Seedling Growth Using RGB-Depth Fusion and Deep Learning

The use of high-throughput phenotyping with imaging and machine learning to monitor seedling growth is a tough yet intriguing subject in plant research. This has been recently addressed with low-cost RGB imaging sensors and deep learning during day time. RGB-Depth imaging devices are also accessible at low-cost and this opens opportunities to extend the monitoring of seedling during days and nights. In this article, we investigate the added value to fuse RGB imaging with depth imaging for this task of seedling growth stage monitoring. We propose a deep learning architecture along with RGB-Depth fusion to categorize the three first stages of seedling growth. Results show an average performance improvement of 5% correct recognition rate by comparison with the sole use of RGB images during the day. The best performances are obtained with the early fusion of RGB and Depth. Also, Depth is shown to enable the detection of growth stage in the absence of the light.

Recognition of Power Equipment Based on Multitask Sparse Representation

Image analysis of power equipment has important practical significance for power-line inspection and maintenance. This paper proposes an image recognition method for power equipment based on multitask sparse representation. In the feature extraction stage, based on the two-dimensional (2D) random projection algorithm, multiple projection matrices are constructed to obtain the multilevel features of the image. In the classification process, considering that the image acquisition process will inevitably be affected by factors such as light conditions and noise interference, the proposed method uses the multitask compressive sensing algorithm (MtCS) to jointly represent multiple feature vectors to improve the accuracy and robustness of reconstruction. In the experiment, the images of three types of typical power equipment of insulators, transformers, and circuit breakers are classified. The correct recognition rate of the proposed method reaches 94.32%. In addition, the proposed method can maintain strong robustness under the conditions of noise interference and partial occlusion, which further verifies its effectiveness.

Research and Application of Full-Link Management Intellectual Inspection Robot with Remote Pre-Control and Information Digitization

With the development of intelligent power grid maintenance, inspection robot has been widely used for its integrated perception and remote pre-control. Aiming at the problems of the weak information interaction capacity and the low correlation with digital requirements of inspection operation, a kind of Full-Link Management Intellectual Inspection Robot is proposed. The article introduces the system structure, function modules, engineering application and debugging process. It is with low response delay and strong anti-interference ability. The correct recognition rate of violations reached 88.2 %, meanwhile provided with the function of evaluation and early warning of workers’ unsafe behavior. The Full-Link Management Intellectual Inspection Robot has a broad application foreground in power grid maintenance.

Investigation of Channel Reduction Based on Brain Lobes in EEG-based Authentication System

Biometric authentication is recently used for verification someone’s identity according to their physiological and behavioural characteristics. The most popular biometric techniques are fingerprints, facial and voices recognition. However, these techniques have the disadvantage in which they can easily be imitated and mimicked by hackers to access a device or a system. Therefore, this study proposed electroencephalogram (EEG) as a biometric technique to encounter this problem. The wavelet packet decomposition is explored in this study for the feature extraction method. The wavelet packet decomposition feature is represented, root mean squared (RMS) wavelet features to extract a piece of meaningful information from the original EEG signal. These features were applied to classify between 15 subjects by using Support Vector Machine (SVM). The channel reduction was conducted to investigate the brain lobe effectiveness during the paradigms of familiar and unfamiliar EEG signals which the channel reduction is based on the brain lobes (temporal, occipital, parietal, and frontal). As a result, the above 14 channels obtained the best performance of the system which is 97.44% of correct recognition rate (CRR). The analysis of the paradigms among familiar only, unfamiliar only, and both familiar and unfamiliar was conducted to evaluate the contribution of the paradigms. The results show that 14 channels obtained the best familiar paradigms while the other contributed by unfamiliar. The result is promising because the CRR computed above 90%, however further analysis of channel reduction has to be work to obtain specific channel to develop the small number of channel for comfort and convenience biometric sensor which is suitable for future authentication.

Sequential subspace estimator for an efficient multibiometrics authentication and encryption.

The principal challenge in biometric authentication is to mitigate the effects of any noise while extracting biometric features for biometric template generation. Most biometric systems are developed under the assumption that the extracted biometrics and the nature of their associated interferences are linear, stationary, and homogeneous. When these assumptions are violated due to nonlinear, nonstationary, and heterogeneous noise, the authentication performance deteriorates. As well, demands for biometric templates are on the rise in the field of information technology, leading to an increase in the vulnerability of stored and dynamic information. Thus, the development of a sophisticated authentication and encryption method is necessary to address these challenges. This dissertation proposes a new Sequential Subspace Estimator (SSE) algorithm for biometric authentication. In the proposed method, a sequential estimator is being designed in the image subspace that addresses challenges arising from nonlinear, nonstationary, and heterogeneous noise. The proposed method includes a subspace technique that overcomes the computational complexity associated with the sequential estimator. In addition, it includes a novel MultiBiometrics encryption algorithm that protects the biometric templates against security, privacy, and unlinkability attacks. Unlike current biometric encryption, this method uses cryptographic keys in conjunction with extracted MultiBiometrics to create cryptographic bonds, called “BioCryptoBond”. To further enhance system security and improve authentication accuracy, the development of a biometric database management system is also being considered. The proposed method is being tested on images from three public databases: the “Put Face Database”, the “Indian Face Database”, and the “CASIA Fingerprint Image Database Version 5.1”. The performance of the proposed solution has been evaluated using the Equal Error Rate (EER) and Correct Recognition Rate (CRR). The experimental results demonstrate the superiority of the proposed method in comparison to its counterparts.

Sequential subspace estimator for an efficient multibiometrics authentication and encryption.

The principal challenge in biometric authentication is to mitigate the effects of any noise while extracting biometric features for biometric template generation. Most biometric systems are developed under the assumption that the extracted biometrics and the nature of their associated interferences are linear, stationary, and homogeneous. When these assumptions are violated due to nonlinear, nonstationary, and heterogeneous noise, the authentication performance deteriorates. As well, demands for biometric templates are on the rise in the field of information technology, leading to an increase in the vulnerability of stored and dynamic information. Thus, the development of a sophisticated authentication and encryption method is necessary to address these challenges. This dissertation proposes a new Sequential Subspace Estimator (SSE) algorithm for biometric authentication. In the proposed method, a sequential estimator is being designed in the image subspace that addresses challenges arising from nonlinear, nonstationary, and heterogeneous noise. The proposed method includes a subspace technique that overcomes the computational complexity associated with the sequential estimator. In addition, it includes a novel MultiBiometrics encryption algorithm that protects the biometric templates against security, privacy, and unlinkability attacks. Unlike current biometric encryption, this method uses cryptographic keys in conjunction with extracted MultiBiometrics to create cryptographic bonds, called “BioCryptoBond”. To further enhance system security and improve authentication accuracy, the development of a biometric database management system is also being considered. The proposed method is being tested on images from three public databases: the “Put Face Database”, the “Indian Face Database”, and the “CASIA Fingerprint Image Database Version 5.1”. The performance of the proposed solution has been evaluated using the Equal Error Rate (EER) and Correct Recognition Rate (CRR). The experimental results demonstrate the superiority of the proposed method in comparison to its counterparts.

Study on face recognition under unconstrained conditions based on LBP and deep learning

At present, the correct recognition rate of face recognition algorithm is limited under unconstrained conditions. To solve this problem, a face recognition algorithm based on deep learning under unconstrained conditions is proposed in this paper. The algorithm takes LBP texture feature as the input data of deep network, and trains the network layer by layer greedily to obtain optimized parameters of network, and then uses the trained network to predict the test samples. Experimental results on the face database LFW show that the proposed algorithm has higher correct recognition rate than some traditional algorithms under unconstrained conditions. In order to further verify its effectiveness and universality, this algorithm was also tested in YALE and YALE-B, and achieved a high correct recognition rate as well, which indicated that the deep learning method using LBP texture feature as input data is effective and robust to face recognition.

A Method for Identifying Fatigue State of Driver's Face Based on Improved AAM Algorithm

The change of lighting conditions and facial pose often affects the driver's face's video registration greatly, which affects the recognition accuracy of the driver's fatigue state. In this paper, the authors first analyze the reasons for the failure of the driver's face registration in the light conditions and the changes of facial gestures, and propose an adaptive AAM (Active Appearance Model) algorithm of adaptive illumination and attitude change. Then, the SURF (speeded up robust feature) feature extraction is performed on the registered driver's face video images, and finally the authors input the extracted SURF feature into the designed artificial neural network to realize the recognition of driver's fatigue state. The experimental results show that the improved AAM method can better adapt to the driver's face under the illumination and attitude changes, and the driver's facial image's SURF feature is more obvious. The average correct recognition rate of the driver's fatigue states is 92.43%.