Earticle

Control performance is critical to a control system. To improve the performance of the steam generator level control system, the control system parameters need to be optimized. Traditional parameters tuning methods, such as trial and error and Design of Experiments etc., are usually experience-based, cumbersome and time-consuming. To address the above inefficiencies, in this paper, the simplex-search based Model-Free Optimization (MFO) has been proposed to search for the optimal control system parameters. The optimized parameters will be gained to maximize the system’s control performance. Rather than traditional controller parameter tuning method, this method optimizes the control system by directly using measurements of control performance. An example of the PID parameters tuning for the steam generator level control was illustrated. The efficiency and the effectiveness of the Simplex-search based Model-Free Optimization – based control parameters tuning methodology has been verified through simulation experiments.

Computer vision applications in the field of agriculture science are gaining importance. The paper presents a method for recognition of paddy varieties from bulk paddy grain image samples based on color texture features extracted from color co-occurrence matrices. The color texture features are obtained from H, S and I color planes and their combinations. The feature set is reduced based on contribution of features to the recognition accuracy. The reduced feature set of the HS plane includes Energy, Entropy and Correlation features from Hue plane and Energy, Entropy, Contrast, and Correlation features from Saturation plane. The paddy grain images are recognized using a multilayer feed-forward artificial neural network. The considered fifteen paddy varieties have given the recognition accuracy of 92.33%. The work is useful in developing a machine vision system for agriculture produce market and developing multimedia applications in agriculture sciences.

This paper proposes to estimate key speaker in meeting speech based on multiple features optimization. First, each feature is defined and their differences between key speaker and other speakers are analyzed. Then, a decision function of multiple feature weighting is generated for estimating key speaker in meeting speech, and the genetic algorithm is used to optimize these coefficients of feature weighting. The methods are evaluated on three different meeting speech datasets. Experimental results show that the proposed optimization method obtains average accuracy of 93.3% for estimating key speaker, and gains average accuracy improvement by 9.7% and 4.1% compared with the previous method and the feature weighting method without optimization, respectively.

Aiming at the drawback of the local search ability of weakness of adaptive chaos particle swarm optimization (ACPSO), which is based on the variance of population’s fitness, this paper presents that introduces the chaos mutation and chaos search to the ACPSO. By using the An chaos mapping to proceed the chaos mutation for some particles and take chaos search for the global optimal particle, it proposes a rule which takes into account the positions of particles and adaptive mutative scale of optimizing space. The results of numerical simulation show that the convergence, and the global and local search ability of the new method are improved, and can effectively avoid premature convergence.

Speech enhancement is a long standing problem with various applications like telephone conversation and speech recognition. The corruption of speech due to presence of additive background noise causes severe difficulties in various communication environments. If the background noise is evolving more slowly than the speech, then the estimation of the noise during speech pauses is easier as compared to non stationary noise. If in case the Noise is varying rapidly then estimation is more difficult. This paper focuses on the class of single-channel noise reduction methods that are performed with frequency domain using short-time Fourier Transform (STFT). There are number of publications and implementations on noise reduction systems. But, there are still some issues in non- stationary noisy systems. This single-channel approach is more dominant and effective approach for practical systems. From last few years, algorithms have been proposed for this problem but most of them are worked on noisy signal in current frame. So in this paper we are trying to propose the new model using Wiener filter by using the concept of multi-frame approach with different window sizes and overlaps. The proposed method shows the results with its superiority.

When a space camera works on the track, it presents in vacuum environment, under the condition of low temperatures, the camera can only maintain its temperature level by absorption of radiation. If stayed in this environment for a long time, mechanical structure of the space camera may become brittle, lose strength, and an exfoliation phenomenon may appear, in addition, abnormal operation and malfunction of electronic equipments may also occur, these may affect imaging quality of the camera. In this paper, a heating control method is designed for a space flight off-axis camera. Firstly, heating control working platform of the camera is introduced, a design method of heating control is determined. After that, heating control treatment procedure is introduced. Finally, the ground temperature monitor is used to validate correctness of the method designed. The comparison diagram before and after thermal analysis and temperature data obtained from experiments are presented, results show that, the heating control method is simple and effective, it has wide applicability and strong generalization performance.

A data adaptive approach to spectral analysis of audio signals is implemented in this paper. The audio signals are non-stationary as well as non-linear in nature and the traditional Fourier based spectral representation is not effective. The Hilbert spectral analysis implemented by noise assisted bivariate empirical mode decomposition (NA-BEMD) is introduced here as an efficient spectral representation scheme of audio signals. In BEMD, the fractional Gaussian noise (fGn) and analyzing speech signal are used as two separate variables. Both signals are decomposed together yielding a finite set of intrinsic mode functions (IMFs) for individual variables (signals). The use of fGn implements BEMD with dyadic filterbank characteristics. The instantaneous frequencies of individual IMFs are computed by applying Hilbert transform and then the time-frequency representation is achieved by arranging the energy with respect to time and frequency simultaneously. Such representation is called Hilbert spectrum (HS) which is analogous to spectrogram. The marginal HS derived from HS corresponds the total energy at each frequency component. The experimental results show that the Hilbert spectral analysis provides better representation of audio signal contents compared to the Fourier based approach.

Recent advances in 3D have increased the importance of stereoscopic content creation and processing. Therefore, converting existing 2D contents into 3D contents is very important for growing 3D market. The most difficult task in 2D-to-3D conversion is estimating depth map from a single-view image. Thus, in this paper, we propose a novel algorithm to estimate the map by simulating haze as a global image feature. Besides, the visual artifacts of the synthesized left- and right-views can also be effectively eliminated by recovering the separation and loss of foreground objects in the proposed algorithm. Experimental results show that our algorithm can produce a good 3D stereoscopic effect and prevent the separation and loss artifacts with low computational complexity.

Individual identification at a distance using gait features has newly gained growing interest from biometrics researchers. Most of the researchers have been shown that different covariate factors can affect different parts of the human body. In this paper, we propose a new approach that minimizes these difficulties, especially for carrying objects by combining static, dynamic, and part-based features. The Gait Features of the walking sequences are extracted by selecting only four sub bands of the Discrete Wavelet Transform (DWT) of the individual images. Moreover, Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT) are implemented to extract lowest and middle frequency components that are used to create robust gait feature images (RGFIs). Then we select effective parts of the body from the Robust Gait Feature Images. After that, these parts of the body are trained using Principal Component Analysis (PCA) and Multiple Discriminant Analysis (MDA) to identify individuals. Experimental result shows promising performance in comparison with other methods.

A novel automatic salient object detection algorithm, which integrates context-based saliency with location computation based on the boundary priors, is proposed. Input image is expressed as a close-loop graph with superpixels as nodes and salient object of image has a well-defined graph-based manifold ranking location. The saliency of the image elements is defined based on their relevances to the given seeds or queries. Saliency object location is carried out in a two-stage scheme to extract background regions and foreground salient objects efficiently. We introduce a location weight to measure the relationship of superpixels and the centroid of the detected salient regions to eliminate the background. Saliency map is computed through context analysis and location computing based on multi-scale superpixels. Experimental results on three public benchmark datasets demonstrate that our approach performs well compared to existing state-of-the-art methods.

A robust and accurate real time eye tracking system has been a challenging task for many computer vision applications. Different researchers working world wide have tried various approaches to solve this problem. Although many different algorithms exist to perform eye detection, each has its own weaknesses and strengths. But so far no system / technique exists which has shown satisfactory results in all circumstances. This research work is a comparative study on the performances of algorithms – Template Matching, Skin Segmentation, Artificial Neural Network and Haar Cascade Classifier for eye recognition. All the algorithms are developed on OpenCV platform and tested on images from Mathworks Video, GTAV, Face Expression and VITS database in the laboratory. The comparison is done based on the success rate i.e. total number of images with eyes detected to the total number of input images. The comparison results show that Haar Cascade Classifier has satisfactory results on images under different conditions such as tilted head position, closed eyes, occluded face, etc., .The purpose of this research work is to develop a Non-intrusive Driver's Drowsiness detection system based on eye blink rate for preventing accidents on road.

In recent years, automatic image annotation (AIA) has been applied to cross-media retrieval usually due to its advantage of mining correlations of images and annotation texts efficiently. However, some AIA methods just annotate images as a unit and the accuracy of annotation may not be acceptable. In this paper, we propose a kind of probabilistic model which may assign keywords to an un-annotated image automatically based on a training dataset of images. Images in the training dataset are segmented into regions and a kind of vocabulary called blob is used to represent these image regions. Blobs are generated by using K-Means algorithm to cluster these image regions. Through this model, we can predict the probability of assigning a keyword into a blob. After the accomplishment of annotation, a keyword corresponds to one image region. Furthermore, the feature vectors of text documents are generated by TF.IDF method and images’ automatic annotation information is used to retrieve relevant text documents. Experiments on the IAPR TC-12 dataset and 500 Wikipedia webpages about landscape show the usefulness of applying probabilistic model of AIA to the cross-media retrieval.

Hand written Character Recognition is area of research since many years. Automation of existing manual system is need of most industries as well as government areas. Recognition of hand written characters is a demand for many fields. In this paper we have discussed our approach for hand written character segmentation. This paper discusses various methodologies to segment a text based image at various levels of segmentation. This paper serves as a guide for people working on the text based image segmentation area of Computer Vision. First, the need for segmentation is justified in the context of text based information retrieval. Then, the various factors affecting the segmentation process are discussed. Followed by the levels of text segmentation are explored. Also, the available techniques with their advantages and weaknesses are reviewed, along with directions for quick referral are suggested. At last, we have given our approach to text segmentation in brief.

As a new representation technique, collaboration representation is used to represent the test sample with all training samples from all classes. Plurality voting is one of the most widely used combination strategies in pattern recognition. This paper presents a new and an efficient face recognition approach using plurality voting and collaboration representation based on the binary bit-plane images. First, face gray images are equalized and decomposed. Next，employing collaboration representation and the corresponding five bit-plane images which have more discrimination information，five identities about the same test image can be obtained. Finally, these five identities vote to the true identity of the test image. If the plurality voting fails, the true identity of the test image will be decided by virtual weight sum face images constructed by 8 bit-plane images with collaboration representation. Weight vector, which is important for the virtual images, is determined by the recognition rate and the order of bit planes. The extensive experiments demonstrate the proposed approach has the higher right recognition accuracy and its speed is faster than the SRC prominently.

In this study, the algorithm that applies Wavelet and multi-scale analysis to remote sensing images is proposed for region variation detection on Markov random field. First of all, the Wavelet transform is adopted to decompose an original image into several sub-images, then the Mahalanobis distance decision function is used to detect the changes in different scale images, and finally the Markov random field is applied to fuse the change detection results at different scales. Since the Markov random field fusion method takes full account of the correlation between the adjacent pixels and the links of the change detection results at different scales, the fusion results are accurate and practical. The testing results prove that the studied algorithm is effective and robust.

In this paper, we propose a novel game theory based framework for synthetic aperture radar image de-noising and segmentation based change detection. We find out the balance of the two aspects. The Nash game theory helps us find out the balance of segmentation accuracy and overall de-noising performance. In the de-noising part, we adopt the multi-diagonal matrix filter based algorithm to undertake the de-noising mission. Segmentation and change detection are finalized by the state-of-the-art methodologies in which the segmentation procedure transfers the difference map into the change map. As far as time-consuming is concerned, we compare the different methods for generating difference map. Fusion map is selected to be our difference map for image segmentation using fuzzy clustering. The experimental analysis shows the effectiveness and robustness of our propose framework with the comparison of other well-known change detection algorithms under the outer environment of noisy and noise-free. Finally, some potential optimization methods are discussed for future research.

Image segmentation is a fundamental and challenging problem in image processing and often a vital step for high level analysis. Due to the active contour model (ACM) need to choose the initial contour for the following evolution, it limited the utilities of this kind of segmentation to a large extent. For the purpose of avoiding the processing of human choosing initial contour, in this paper, we proposed an automatic initial contour choosing algorithm of the input image information. Based on the chosen initial contour, the iterative efficient and the accuracy of segmentation have been improved when the initial contour is incorporated into the local based segmentation framework. Extensive experiments on synthetic and real images are provided to evaluate our method, showing significant improvements on the segmentation accuracy and stability, comparing to the human chosen initial contour, such as LBF and LGIF.

In this paper, a novel blind color image watermarking technique using Quaternion and Joint QSVD Blocks is proposed to protect the intellectual property rights of color images. The proposed method tries to insert the watermark in the joint blocks .In this method, the color image is considered as an array of pure quaternion numbers. Then the array of pure quaternion is divided into non-overlapping blocks and performs QSVD to the blocks. The watermarking is inserted into the S blocks by changing the 2 of S of two adjacent blocks. Besides, in the procedure of watermark insertion and extraction, ensuring higher fidelity and robustness and resilience to several possible image attacks have been considered. The experimental results showed that the proposed method performance created watermarked images with better PSNRs and more robustness versus several attacks such as JPEG, Salt &Pepper noise, Brightness adjustment, Sharpen and Blurring.

As the state-of-art denoising method, BM3D is capable of achieving good denoising performance by exploiting both the non-local characteristics and sparsity prior knowledge of images. Nevertheless, experimental results show that the dissimilarity measurement defined in BM3D sometimes results in grouping patches with distinct structure. Inspired by the fact about the different impact of noise on patches with various structures, we propose a structure-adaptive image denoising method with 3D collaborative filtering by optimizing the block matching procedure. In our method, the similarity in the variance between patches is incorporated in block matching procedure. Besides, based on the prior knowledge of correlation among patches in the same neighborhood, the spatial distance between the reference patch and the candidate is also taken into account when measuring patches’ dissimilarity. Several numerical experiments demonstrate that the proposed approach achieve better results in PSNR and visual effect than original BM3D.

Tracking articulated hand motion from visual observations is challenging mainly due to the high dimensionality of the state space. Dense sampling is difficult to be performed in such high-dimensional space, so the traditional particle filtering can’t track articulated motion well. In this paper, we propose a new algorithm by combining differential evolution with a particle filter, to track the articulated motion of a hand from single depth images captured by a Kinect sensor. Through the optimization procedure of differential evolution, the particles are moved to the regions with a high likelihood. Only single depth information is used as the input, so our method is immune to illumination and background changes. The tracking system is developed with OpenSceneGraph (OSG). Experiments based on both synthetic and real image sequences demonstrate that the proposed method is capable of tracking articulated hand motion accurately and robustly.

Augmented Reality (AR) is a new computer application and human-machine interaction derived from virtual reality. It has been widely applied in social each domain. With the improvement on the performance and the quality of built-in camera of smartphones, it is possible to have the augmented reality system to come true. And with people entertainment needs growing, mobile augmented reality based on face detection has become a research hotspot. This paper discusses the development and typical technology of face detection technology and mobile augmented reality. A development framework of the face detection technique based on OpenCV in mobile augmented reality application is proposed. And this essay takes the AR face book show as an example analyzes the key technology in the process of development and proves its feasibility and effectiveness.

In recent years, l1 norm is usually considered as the regularization term in the field of sparse representation. However, the non-zero entries obtained by the l1 regularization term always neglect the correlations with each other. In fact, different relationships or structures among non-zero entries are necessary in many applications. K-support norm is firstly proposed in the field of sparse prediction. The most important property of the k-support norm is grouping feature of the largest entries in the obtained solution. In this paper, we present a new image processing model by introducing the k-support norm to image gradient domain. The proposed model can be applied to image denoising and edge detection simultaneously. Some examples demonstrate the effectiveness of the novel model and its improvements.

Sparse representation (compressive sampling) has achieved impressive results in object tracking by looking for the best candidate with minimum reconstruction error using the target template. However, it may fail in some circumstances such as illumination changes, scale changes, the object color is similar with the surrounding region, and occlusion etc., in addition, high computational cost is required due to numerous calculations for solving an l1 norm related minimization problems. In order to resolve above problems, a novel method is introduced by exploiting an accelerated proximal gradient approach which aims to make the tracker runs in real time; moreover, both classic principal component analysis algorithm and sparse representation schemes are adapted for learning effective observation model and reduces the influence of appearance change. Both qualitative and quantitative evaluation demonstrate that the proposed tracking algorithm has favorably better performance than several state-of-the-art trackers using challenging benchmark image sequences, and significantly reduces the computing cost.

The term content-based image retrieval to describe his experiments into automatic retrieval of images from a database by color and shape feature. The term has since been widely used to describe the process of retrieving desired images from a large collection on the basis of features that can be automatically extracted from the images themselves. The features used for retrieval can be either primitive or semantic, but the extraction process must be predominantly automatic. Retrieval of images by manually-assigned keywords is definitely not CBIR as the term is generally understood – even if the keywords describe image content. And this paper deals, retrieve the images into three basic categories of color called RGB. After retrieving the image with their component the transformation can be applied. The HOG method is used to retrieve the feature of image vectors and others. In this paper, the HOG method is fully analyzed and proves its accuracy and efficiency of image retrieval with reduced number of steps.

Incorporating prior knowledge (PK) into learning methods is an effective means to improve learning performance. On the bases of requirements of engineering practice and the characteristics of knowledge representation of extension neural network (ENN), with the purpose of further improving the performance of ENN in engineering practice, a prior-knowledge-based ENN (PKENN) recognition method is proposed and applied in the application of safety status pattern recognition of coal mines in this paper. The PKENN recognition method effectively combines domain knowledge with training data set. The prior knowledge can provide additional information about the classical domain of characteristic vector that may compensate for the low quality of training data in a complex application environment. This method can set the initial weights of ENN, guide the learning of ENN and alleviate the learning burden. To demonstrate the validity and effectiveness of the proposed method, a real-world application on the geological safety status pattern recognition of coal mines is tested. Comparative experiments with existing methods and other ANN-based methods are conducted. The experimental results show that the proposed PKENN recognition method has a better performance.

With the development of spectral imaging technology, it makes hyperspectral imagery widely used. According to the features of multiple bands and the strong mutual correlation among these bands, this paper presents a band selection method for hyperspectral imagery classification based on improved PSO (Particle Swarm Optimization). First of all, we use information divergence to describe the correlation of the bands, then build the information divergence matrix to make the classification of subspaces. Secondly, we construct the fitness function of the algorithm with the band information and categories of the Bhattacharyya distance (B distance) to improve the inertia weight updating method in PSO. Finally, based on the AVIRIS hyperspectral imagery and compared with existing method to conduct experiments, the average classification accuracy of the proposed method is 81.36%, which is distinctly improved 0.91% compared with the existed method. Meanwhile, the proposed method has a significantly faster convergence speed during the process of the band selection. Therefore, the experimental results verify the effectiveness of the proposed method in this paper.

Digital watermarking is an application associated with copyright protection. Any digital object can be used as a carrier to carry information. If the information is related to object then it is known as a watermark which can be visible or invisible. In the era of digital information, there are multiple danger zones like copyright and integrity violation of digital object. In case of any dispute during violation, content creator can prove ownership by recovering the watermark. In this paper, a comparative study of two most recent digital watermarking techniques namely DWT and DWT-SVD over RGB images is presented. In case of DWT (Discrete Wavelet Transform) watermarking technique, decomposition of the original image is done to embed the watermark and in case of DWT-SVD watermarking technique, firstly original image is decomposed according to DWT and then watermark is embedded in singular values obtained by applying SVD (Singular Value Decomposition). The performances of the proposed techniques are compared on the basis of PSNR values.

Aiming at the inaccurately modeling and some uncertain existing in servo system seriously affected the control quality and the instability problem, sliding mode control algorithm with improved reaching law is proposed in this paper. The improved reaching law is used to weaken the chattering problem existing in the sliding mode control. Also the kalman filter is used to inhibit the interference, which makes the servo system have strong anti-interference ability and the ability of weakening the chattering problem existing in the sliding mode control. The method of sliding mode control is the basis of a number of patents and patents pending. The simulation results show that the algorithm can effectively inhibit the external disturbance and noise existing in the system, and make the system have strong anti-interference ability. At the same time, the chattering also is obviously inhibited, and the method makes the system stability and control quality been further improved.