International Journal of Pattern Recognition and Artificial Intelligence

Moving foreground objects detection in complex scenes is a tough job because it requires high recognition accuracy. Adaptive Gaussian mixture model (AGMM) can be used to extract the foreground objects and it shows good performance, however, the detection quality of the foreground objects under complex scenes is not excellent. In this paper, an AGMM and BP neural network hybrid method is proposed, which is used to extract the foreground objects in complex scenes such as, dynamic backgrounds, illumination changes and moving shadows. In this method, an improved BP neural network is used to post-process the images of the foreground objects that are extracted from the AGMM. The neural network has strong robustness by learning the statistical features of the images. Momentum term and adaptive learning rate are added in the BP neural network algorithm to improve the training speed and robustness of the network. The experimental results show that the proposed AGMM and BP neural network hybrid method can extract the complete foreground objects effectively when compared with some other moving objects detection algorithms.

The output power of wind turbine has great relation with its health state, and the health status assessment for wind turbines influences operational maintenance and economic benefit of wind farm. Aiming at the current problem that the health status for the whole machine in wind farm is hard to get accurately, in this paper, we propose a health status assessment method in order to assess and predict the health status of the whole wind turbine, which is based on the power prediction and Mahalanobis distance (MD). Firstly, on the basis of Bates theory, the scientific analysis for historical data from SCADA system in wind farm explains the relation between wind power and running states of wind turbines. Secondly, the active power prediction model is utilized to obtain the power forecasting value under the health status of wind turbines. And the difference between the forecasting value and actual value constructs the standard residual set which is seen as the benchmark of health status assessment for wind turbines. In the process of assessment, the test set residual is gained by network model. The MD is calculated by the test residual set and normal residual set and then normalized as the health status assessment value of wind turbines. This method innovatively constructs evaluation index which can reflect the electricity generating performance of wind turbines rapidly and precisely. So it effectively avoids the defect that the existing methods are generally and easily influenced by subjective consciousness. Finally, SCADA system data in one wind farm of Fujian province has been used to verify this method. The results indicate that this new method can make effective assessment for the health status variation trend of wind turbines and provide new means for fault warning of wind turbines.

Text-independent Chinese writer identification does not depend on the text content of the query and reference handwritings. In order to deal with the uncertainty of the text content, text-independent approaches usually give special attention to the global writing style of handwriting, rather than the properties of each individual character or word. Thanks to the existence of high-frequency characters, some characters probably appear in both the query and reference handwritings in most cases. If character images in the query handwriting are similar to those in the reference handwriting, this query handwriting and the corresponding reference handwriting are very likely to be written by the identical writer. In this paper, we exploit the above characteristic to improve the performance of Chinese writer identification. We first present an identification scheme using edge co-occurrence feature (ECF). Then, we detect the character pairs in the query and reference handwritings using a two-step framework and propose the displacement field-based similarity (DFS) to determine whether a character pair is written by the identical writer. The character pairs help to re-rank the candidate list obtained by text-independent ECF-based similarity and finally decide the writer of the query handwriting. The proposed method is evaluated on the HIT-MW and CASIA-2.1 datasets. Experimental results demonstrate that our proposed method outperforms the existing ones, and its Top-1 accuracy on the two datasets reaches 97.1% and 98.3%, respectively.

Recent research has shown that lumbar disease has become common in China. Since the structure of the lumbar spine is extremely complex, a finite element analysis method was used to perform biomechanical simulation and analysis of stress and strain on the L3–L4 lumbar segment to provide both a scientific and theoretical basis for clinical diagnosis and medical research. The MC volume-rendering 3D reconstruction method was the first step to accurately constructing the finite element model of the L3–L4 lumbar sacral segment, which was simulated prior to the addition of the ligaments, fibrous ring, and other major spinal tissue. The finite element model network was classified and the material properties of the corresponding parts were described. According to the normal model, careful simulation and deformation were performed, in addition to intervertebral disc degeneration in various cases. We have provided a detailed and professional analysis of the biomechanical properties, providing a powerful biomechanical basis for the diagnosis of intervertebral disc bulge and degeneration.

Saliency detection refers to the segmentation of all visually conspicuous objects from various backgrounds. The purpose is to produce an object-mask that overlaps the salient regions annotated by human vision. In this paper, we propose an efficient bottom-up saliency detection model based on wavelet generalized lifting. It requires no kernels with implicit assumptions and prior knowledge. Multiscale wavelet analysis is performed on broadly tuned color feature channels to include a wide range of spatial-frequency information. A nonlinear wavelet filter bank is designed to emphasize the wavelet coefficients, and then a saliency map is obtained through linear combination of the enhanced wavelet coefficients. This full-resolution saliency map uniformly highlights multiple salient objects of different sizes and shapes. An object-mask is constructed by the adaptive thresholding scheme on the saliency maps. Experimental results show that the proposed model outperforms the existing state-of-the-art competitors on two benchmark datasets.

In this paper, an effective method based on the color quaternion wavelet transform (CQWT) for image forensics is proposed. Compared to discrete wavelet transform (DWT), the CQWT provides more information, such as the quaternion’s magnitude and phase measures, to discriminate between computer generated (CG) and photographic (PG) images. Meanwhile, we extend the classic Markov features into the quaternion domain to develop the quaternion Markov statistical features for color images. Experimental results show that the proposed scheme can achieve the classification rate of 92.70%, which is 6.89% higher than the classic Markov features.

Precision and accuracy estimation is an important index used to reflect the measurement performance and quality of a measurement system. To reveal the significance and connotations of the precision and accuracy estimation index of a close-range photogrammetry system, several common precision and accuracy estimation methods are proposed. They are explained from a theoretical perspective, and the mechanism of the internal coincidence precision estimation and the external coincidence accuracy estimation are deduced, respectively. Through detailed experimental design and testing, the validity and reliability of the proposed precision and accuracy estimation methods are verified, which provides strong evidence for the quality control, optimization and evaluation of the measurement results from a close-range photogrammetry system. At the same time, it has significance for the further development of precision and accuracy estimation analysis of close-range photogrammetry systems.

Face recognition using sparse representation-based classification (SRC) is a new hot technique in recent years. However, the research indicates that it is the collaborative representation but not the $L_1$-norm sparsity that makes SRC powerful for face classification. Consequently, we propose a simple yet much more efficient face classification scheme, namely two-step collaborative representation-based classification (TSCRC) method. First, we exploit the symmetry of the face to generate new images of each test sample. Then, the original and new generated test samples are, respectively, used to perform TSCRC, which ultimately uses a small number of classes that are near to the test sample to represent and classify it. Finally, the score level fusion is taken to perform classification recognition. The experimental results clearly show that the proposed method has very competitive classification results.

Aiming at the problem of auditory negative enhancement of typical phase reconstruction method, an improved method of phase reconstruction and MMSE-LSA estimation is proposed. First, the geometric relationship between noisy speech and clean speech in unvoiced segment is used to estimate the phase of the clean speech; Second, considering the randomness of speech appearance in the actual noise environment, a modified MMSE-LSA amplitude estimation is proposed by using the binary hypothesis model. Finally, the new phase reconstruction in voiced and unvoiced speech is combined with the modified MMSE-LSA. The simulation results show that the performance of the algorithm proposed in this paper is better than typical phase reconstruction method in terms of the SegSNR and PESQ.

In the actual environment of security detection, many kinds of liquids often exist in the same detection background, and their dangerous levels are difficult to identify. Therefore, it is very important to research on identifying the dangerous levels of various liquids. The paper establishes the $S$-parameter database of tested samples under specific detection environment with free space method. In the actual detection, ultra-wide-band (UWB) centimeter wave is used to measure the $S$-parameters of several detected liquids first. Then the fast independent component analysis (FastICA) algorithm is used for unmixing the mixed signal by Newton’s iteration method and the negative entropy maximization search principle. The unmixed signal matches with the sample database adaptively, so the dangerous levels of the detected liquids are identified. Multiple experiments show that FastICA algorithm can reach a matching rate of 95% between water and 90${}^{\#}$ gasoline or alcohol and 90${}^{\#}$ gasoline, it also can reach a matching rate of around 73% between water and alcohol. This algorithm has a quick response and high reliability for identification of dangerous liquids. FastICA algorithm in this paper is applied for detecting the dangerous liquids for the first time, and it has high application value.

The goal of this study is to improve the accuracy of mobile node localization and to avoid the influence of moving direction-offsets introduced by the positioning system’s accuracy control. The proposed localization algorithm, which is based on an overlap self-adjustment method, and an anchor node selection mechanism which uses the Gaussian elimination method, is based on the error probability. The proposed algorithm has the advantages of requiring little prior information, and it reduces the power consumption. The simulation results show that the proposed algorithm is better than the self-adjustment localization (SAL) algorithm in terms of its localization accuracy.

This paper presents a new method for learning the structure of Bayesian Networks. Broadly speaking, we leverage the Branch and Bound (B&B) to derive the best Directed Acyclic Graphs (DAGs) that describes the structure of the network. Our contribution consists in introducing two main heuristics: the first one allows the selection of the graph that has the best score among those that contain less cycles, the second one eliminates the shortest cycle from the selected graph; it aims to reduce the number of explored nodes. Our experimental study asserts that the suggested proposal improves the results for multiple data sets. These facts are confirmed by the reduction of the computation time and the memory overhead.

Recently, a large number of studies have shown that Convolutional Neural Networks are effective for learning features automatically for steganalysis. This paper uses the transfer learning method to help the training of CNNs for steganalysis. First, a Gaussian high-pass filter is designed for pretreatment of the images, that can enhance the weak stego noise in the stegos. Then, the classical Inception-V3 model is improved, and the improved network is used for steganalysis through the method of transfer learning. In order to test the effectiveness of the developed model, two spatial domain content-adaptive steganographic algorithms WOW and S-UNIWARD are used. The results imply that the proposed CNN achieves a better performance at low embedding rates compared with the SRM with ensemble classifiers and the SPAM implemented with a Gaussian SVM on BOSSbase. Finally, a steganalysis system based on the trained model was designed. Through experiments, the generalization ability of the system was tested and discussed.