Earticle

Pose estimation which is regard as the cross-technology of computer vision and pattern recognition, and an important prerequisite for human behavior understanding. Human pose estimation which use the probability theory, machine learning, pattern recognition, graph theory and other theories to get the position, the deflection angle of the various parts of the body. Then make the detection and estimation parameters for the human body pose. When the image has interference in the background, color and scale changed, human pose complex, self-occlusion and interpersonal interaction occlusion may make the precision and accuracy of pose estimation face great challenge. Thus, according to the above problems, this paper use the advanced model of the human body as contour model to descript the complex pose, in order to make the model more accurate and suitable for various human pose, we pre-clustering the human body pose of the training samples before we trained the model and in order to ensure the accuracy of the pose we use robust segmentation of multi-view with a novel shape prior. The experiment shows that the algorithm performs better than the classic algorithm on the public datasets.

There is a similar image retrieval method which uses a normalized histogram of the pixel values of the original image and that of the pixel values of the illegally copied search target image. In this method, when the correlation coefficient between the two histograms is greater than a predetermined threshold value, the target image is to determine to be similar to the original image. However, the contents of the image that has received ornament attack is different from the contents of the original image, this means that it is searching for images similar to the original image is very difficult. In general, ornament processing of the image is applied to the peripheral region of the image. Therefore, if only the pixels to limit in the ROI (region of interest) which are not affected by the ornament attack are used to create the normalized histogram, it may be effective. We, in order to extract only the pixels included in the ROI, have proposed a method of using a mask image. In this paper, it describes some detailed study about the element techniques including using the mask image and normalization histogram. Moreover, the computer simulations show the effectiveness of these element techniques.

3D face reconstruction and 3D face recognition belonged to different fields previously. In order to solve these two issues in a unified way, this paper designed a kind of regionalized implicit function feature (RIFF) algorithm frame. Firstly according to the biology characteristic, the face was segmented to five regions with different resolutions; secondly the control matrix which expressed the human face was solved for 3d face reconstruction; then dimensionality reduction was did for this control matrix; finally the low dimensional character description was computed completely for 3d face classification. The specific experiments for the RIFF method are performed and the results show that this method can acquire significant effect in the performance of reconstruction and recognition.

A kind of hot spots detection alarm device and method is proposed in the paper about infrared thermal imaging, including the hardware design, algorithm, software realization. The hardware mainly includes the infrared thermal imaging camera and intelligent analysis board; the algorithm uses connected component detection method based on run length to extract target rectangle coordinates. This design uses the simple character overlapping circuit and superimposes the alarm target rectangle on the composite video. It has high integration level, simple circuit, and can effectively deal with more than 12 bit pixel data. Because FPGA uses the parallel processing, greatly improving the graphics processing speed, it is suitable for embedded applications in high real-time.

Introduction of noise in image causes variation in brightness or colour information. Various noises such as Gaussian, Impulse, Speckle, Periodic andPoisson noise deteriorates the picture quality. Gaussian noise is the most popular one that incurs during acquisition process caused by poor illumination and/or high temperature. This paper proposes an improved modified adaptive Bilateral Filter to remove Gaussian noise from colour images. This technique is implemented in MATLAB-9 and various performance metrics taken into consideration are: Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE), Mean Absolute Error (MAE), and Normalized Colour Difference (NCD). The result shows that the proposed technique gives the best results than all other techniques in terms of all comparison parameters.

Moving shadow detection is a challenging task in computer vision applications, such as surveillance, video conference, visual tracking, object recognition, and many other important applications. In this paper, region-based moving shadow detection using affinity propagation (RMSDAP) is presented, which detects shadows in terms of texture similarity. Firstly, we divide foreground image into no overlapping blocks and extract color features from each block. Secondly, affinity propagation is utilized to cluster foreground blocks adaptively and sub regions are generated after coarse segmentation. Specially, each sub region has the characteristics of regional uniformity. Finally, we extract texture feature from irregular sub regions while calculate texture similarity and normalized correlation coefficient simultaneously for each sub region to classify moving shadows. Extensive experiments demonstrate that RMSDAP is superior to some well-known methods especially pixel-based methods. In particular, our method exhibits much better performance compared with fixed block method, which can maintain the texture consistency in one region adequately.

Concerning of the impression of uncertainty and fuzziness that consumers have on product appearance, a synthetic evaluation model of music electronic equipment that meets the modern aesthetic needs is built in order to change the randomness and uncertainty of traditional product design. Through analyzing the three basic factors and several evaluation indexes in music player design, the weight of each evaluation index is determined by the application of Analytic Hierarchy Process (hereinafter referred to as AHP), and a two-level fuzzy synthetic evaluation model is built by using fuzzy mathematics. Finally, it is verified in experiment that the uncertainty in product design can be effectively improved, which at the same time increases the satisfaction rate and success rate of product design.

Early detection of cancer is the best method to increase the chances of survival. In early stage, cancer can be detected using mammography, fine needle aspirate, and surgical biopsy. In this study, we propose the use of Dual-Tree Complex Wavelet Transform (DTCWT) based feature with neural network classifier for mammography image analysis. The result is evaluated using specificity, sensitivity, and accuracy. Computational experiments show the proposed method is superior compare to DWT with 96.3% accuracy.

Research on ship photoelectric measuring instrument measurement error compensation technology, which can improve the accuracy of the measurement. According to the ship sailing in the influence of turbulence by the winds and waves, causing airlines direction deflection happen suddenly, and measurement data is not accurate with jumping up and down along the axis. In order to avoid these problems, this paper proposes a kind of ship turbulence measurement compensation algorithm, by adjusting the ship turbulence coefficient to predict the turbulence extend of the ship, and compensate the turbulence measurement error, use the ship turbulence coefficient correction method to revise the result of compensation, overcome turbulence effects impact. The experiment proves that the ship turbulence measurement compensation algorithm compensate measurement results, can improve the accuracy rate of measurement data and satisfactory results have been obtained.

Noise not only deteriorates image quality but also may result in loss of important information hidden in images (i.e. medical applications). Various types of noises are available in literature such as Gaussian noise, impulse noise, mixed noise etc. In order to remove these noises from images, bilateral filters and its variants are used. This paper surveys the impact of these techniques using several performance metrics such as Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE), Mean Absolute Error (MAE) and Time Complexity. For this purpose a simulator is designed in MATLAB to implement these techniques. The results show that Joint Bilateral Filter (JBF) technique is the best technique for removing Gaussian noise, Modified Double Bilateral Filter (MDBF) technique provides good results for removing impulse noise and Switching Bilateral Filter (SBF) technique work well for mixed noise as seen from results.

This paper introduces the DP-QPSK modulation coherent reception technology,the overall design and other key is sues for 100 Gbit/s DP-QPSK modulation. Using the technologies based on DSP, it realizes the long-range transmissions of 100 Gbit/s optical systems to achieve optical signal dispersion compensation, polarized solution reuse and phase estimation. The effect of this scheme is verified with OptiSystem and the simulation results indicate that this with the help of DSP module for processing of the received signal, the last constellation is ideal and scheme is implemented simply and has high reliability, and it also has reference significance for the optimization of coherent optical detection. The results show that the technical plan is easy and reliable, which has reference significance on optimization of coherent optical detecting hardware.

This paper presents a new hierarchical phoneme recognition system using the SVM classifier and different feature representations based on mellin transform. The proposed architecture uses different representations with each group of phonemes of the speech database TIMIT which are distributed in a way to reduce the confusions between phonemes having similar articulatory strcuture. The main idea of this new architecture is based on the principle that each group of phonemes has his own characteristics which requires us to choose the adequate representation for the given group. Experiments have proven the robustness of our new hierarchical phoneme recognition system (called MMP) and the use of conventional feature representations based on mellin transform explains its superior recognition performance.

Face recognition (FR) system can automatically identify or check face image from a digital camera or image generation equipment, in order to do this, to extract facial features from images obtained, and compared with face data in the database. At present, almost all of the FR face barriers associated with facial Angle, including the lack of light and the low resolution, these problems has greatly reduce the recognition rate. In order to solve this problem, this paper proposes a face recognition framework based on the sub pattern under the condition of illumination change, first of all, the framework using minimize the total variation image of discrete cosine transform (DTV) and the Gabor filter, and combined the sub-mode analysis (SMP) and distinguish the accumulative feature transformation (DAFT), can effectively solve the face recognition problem of light conditions big change, secondly by extracting the texture characteristics of local model is not sensitive to illumination changes, using Distance transformation measures (Distance Conversion Metrics, DCM) and k-means (K-Mean) algorithm, the recognition rate of face recognition is improved effectively. The effectiveness of the method verified respectively on the two face library: ATR - Jaffe and Yale, the experimental results show that compared with other state-of-the-art methods, the proposed method in dealing with the face recognition problem under the condition of the unrestricting obtained better recognition effect.

The feature description of image and the index mechanism of the feature are the keys to realize the content-based image retrieval, aiming at the problem of massive image data and “dimension curse”, this paper proposes the BOF-based image retrieval improved algorithm, and combines with VLAD and soft assignment it generates the soft assignment local aggregation descriptor (SA-VLAD) which has a better ability to resist the dimension reduction and a higher recognition rate. When the index mechanism IVFADC is at query time, to ensure the recall ratio and precision rate of the result, the candidates inverted index chain are increased, which leads to the problems of distance calculation and the query time’s increasing. For this point, in the index phase, the scattered distribution is carried out aiming at the database vector, which reduces the burden of distance calculation, and improves the quality of the query results at the same time. The experimental results show that the algorithm in this paper obtains a good effect in the content-based massive image database retrieval.

Vehicle detection and tracking applications play an important role for military and civilian applications such as in highway traffic surveillance control management and traffic planning.This paper presents a review on the various techniques of On-Road Vehicle detection systems that are based on motion model. In this paper a literature Survey of previous and recent works is presented on vision-based vehicle detection using sensors. Detecting the objects in the video and tracking their motion to identify their characteristics has been emerging as a demanding research area in the domain of Image Processing and Computer Vision. The traffic image analysis comprises of three parts: (1) Traffic Analysis (2) Motion Vehicle Detection and Segmentation Approaches and (3) Vehicle Tracking Approaches. In this survey, we have classified these methods into various groups, and these groups are providing a detailed description of various representation methods and find out their positive and negative aspects.

This paper addresses Chinese discourse segmentation based on punctuation mark. Particularly, we propose various kinds of lexical, syntactic, position and punctuation features to train classifiers for Chinese discourse segmentation. Experimental results on CDTB (Chinese Discourse Treebank) show that our method based on punctuation mark is appropriate for Chinese discourse segmentation with 89.2% in accuracy.

There are many shortcomings in traditional JPEG2000 image compression transmission algorithms in the wireless networks, such as, large energy consumption of resources in the network, large stress and energy consumption of the compressed image acquisition node and long transmission time, the image compression algorithm based on singular value decomposition is proposed. Block adaptive compression algorithm based on singular value decomposition is used in image compression processing, through the analysis of singular value decomposition and network energy consumption model. Meanwhile, the task of the network data processing and long-distance data transmission is assigned to different roles node to complete to balance the distribution of the network energy consumption. Finally, detailed experimental simulation have done, and the experimental results show that compared with the traditional compression algorithm (JPEG2000 image compression), the image compression algorithm based on singular value decomposition greatly alleviates the network camera key node energy consumption, and improves the speed of multimedia data transmission in wireless multimedia sensor network, and quality.

A new direction-of-arrival estimation method for uncorrelated wideband sources is proposed. Firstly, the covariance matrices of different frequency bins are transformed into higher dimensional vectors through Khatri-Rao product to produce the new models of the array outputs. Secondly, the higher dimensional noiseless signal subspace at each frequency is focused to a certain reference frequency and a single correlation matrix is constructed, where the focusing matrices are designed by interpolating the virtual steering matrices of the new model. Finally, the DOA estimates are acquired through using narrowband MUSIC method. The proposed method is able to increase the degrees of freedom of a uniform linear array and enable us to handle more sources than sensors. Numerical results demonstrate the performance of the proposed method.

Identifying the interest points in an image is a key step in image processing and computer vision tasks. Every corner of the images represents a lot of information. Extracting the true corners is the main object to image processing, which can reduce much of the time and calculations. Many algorithms have been suggested in the image processing to detect the true corners, based on the robust statistics. In this paper the corner detection algorithms SIFT and FAST have been studied in image processing under the various image formats. Also, it can provide a direction to the researchers to use the algorithm for the suitable image format and to develop a new algorithm which can detect the exact corners of an image/blurred image. The FAST corner detection method compared with the results of SIFT corner detection method. Experimental results show that the FAST corner detection gives better results compared to SIFT method. All the experiments are carried out MATLAB software.

In this paper, the disappearing daur culture and its color of traditional craftwork are surveyed and classified. By reviewing the relevant literature and video data, the RGB color module of single utensil and visual recognition image are selected to represent the pattern of this craftwork of daur. Specially, color quantization is achieved by feature extracting for the building of color database based on which a further color message classification is accomplished. Finally, fuzzy clustering is conducted to compute the complexity of color which also verifies its high precision.

To enhance the security of the digital image information transmitted through Internet, this paper proposes an image scrambling algorithm based on spiral filling of bits. This algorithm first reads the pixel values of all the odd-number lines in the images successively from top to bottom and then does the same to the even-number lines. All results will be saved in an one-dimensional array. All the elements in the array are transformed into binary digits and saved in a one-dimensional array that is 8 times longer than the previous one. In this new array every element only saves one binary digit. The bit sequences corresponding to the image odd-number lines in this array are rearranged in an inverted order. The arrays disposed into an inverted order successively fill a matrix whose line number is the same as the old image and column number is 8 times of the old image in a direction of heliciform. Then a new matrix is generated. Lastly every 8 elements in this new matrix are considered as 8 binary digits and are transformed into decimal digits. That is how image scrambling is conducted. This algorithm is simple and easy to be realized. Abundant experimental results have shown that the algorithm has well scrambling effect and is able to recover image without distortion.

A new Sliding Window Based Corner Detection Algorithm has been developed for image mosaicing. Using this algorithm, a new architecture suitable for VLSI implementation has also been designed. The proposed design incorporates a high degree of pipelining and parallelism and hence offers high throughputs. The color image mosaicing is achieved by first convolving an original image using a 3x3 sliding window followed by smoothening the image using median filters. The design has been coded in Verilog as per RTL coding guidelines. The algorithm has also been coded in Matlab in order to validate the hardware results. The Verilog design of the proposed architecture for image mosaicing has been implemented on Xilinx Spartan 6 xc6slx45-3fgg676 FPGA device. The design utilizes about 144,858 gates and the operating frequency is about 100 MHz. The design is capable of processing high resolution color pictures of sizes of up to 1600×1200 pixels in real time.

In this paper, a novel image segmentation algorithm based on fuzzy clustering and entropy analysis using space information for optical images is proposed. We adopt the general properties of Hopfield neural network (HNN) and multi-synapse neural network (MSNN) to gain the center of the clusters and the fuzzy membership degrees for solving the optimization problems. As far as the noise influence is concerned, we introduce a novel window to improve the robustness of the proposed algorithm. In the experimental analysis part, we compare our method with some state-of-the-art methodologies and adopt the well-known test image databases to conduct the experiment. The result indicates that compared to FCM and some other clustering methods, our entropy and neural network based algorithm performs better. Our approach is less time-consuming and more robust to noise.

Digital image correlation (DIC) is an effective displacement field measurement method featuring non-contact and full-field, which has been successfully applied in lots of fields, especially in the field of experimental mechanics. Unfortunately, the traditional DIC technique (TDIC) depends on the output values of image sensor heavily that usually noised in practical imaging system. So, TDIC cannot deal with bias error introduced by the image noise effectively. In this paper, a kernel based DIC is proposed to help reduce the bias error by establishing and optimizing a weighted correlation function. Compared with the methods of low-pass filtering, KDIC preserves the high frequency information well but reduces the bias error caused by image noise effectively. To demonstrate, two kinds of kernel function are analyzed. One is Epanechnikov kernel, with which KDIC is equivalent to TDIC. The other is Gaussian kernel, which can be used to improve the anti-noise performance of TDIC and get more accurate sub-pixel displacements. Both simulation analyses and experimental results validate the effectiveness of this new method.

This paper proposes a color encoding structural light approach for the three-dimensional measurement of turbine blade surface. It adopts RGB color mode design coding scheme, by using the six basic colors, red, green, blue, yellow, magenta, and cyan for color-coded stripes. Every three stripes are grouped together, while the color combinations for each group are different. The coding pattern based on these color stripes is projected to the blade. A CCD camera then captures the image modulated by the blade. We then decode this image to obtain the three-dimensional coordinates of the measured points on blade surface. According to the structure light measurement principle, we designed and determined the parameters of the three-dimensional measurement. This paper established a mathematical model of a turbine blade through simulation by using 3ds max software. We designed a three-dimensional measurement system for the simulation, and obtained the three-dimensional coordinates of the blade surface by measuring the turbine blade. Finally, the three-dimensional blade model reconstructed by the measured data is established.

Sky detection is an important problem in computer vision. Generally, the color and gradient of the pixel in sky border change greatly. So, sky in the scene can be detected by measuring the sky border points. In this paper, a new method for sky detection based on border points is proposed, where the border points in each column of image can be detected to complete complex sky region identification in image. Different to the existing sky detection methods, the proposed algorithm is efficient in detecting sky in scenes of complex structure, for example, sky which are separated by buildings or flags. Furthermore, the paper introduced the concept of sky border for the sky detection, which made the proposed scheme effective. The proposed method is able to handle even more complex sky conditions, including a cloudy sky and a sunny sky. Experimental results have proven that the effectiveness of this method.

This paper presents a Uyghur text-line separation algorithm based on classified connected components of Uyghur Handwritten scripts. In order to get the location of main text-lines, this paper proposes an adaptive painting and thinning algorithm. To insure the efficiency of text-lines segmentation, the post-processing of text-line detection procedures are introduced. The experimental results show that the algorithm is strongly robust for segmentation of text-lines with having some skewness, touching and overlapping, and small strokes remained.

As the important case of conformal arrays, conformal arrays on cylindrical surfaces have been widely used in radar and communication systems. This paper explores the chances of designing conformal reflectarray on cylindrical platform. An extended array theory method is proposed in this paper for analyzing the radiation patterns of cylindrical conformal reflectarray. The radiation characteristics of conformal reflectarray on convex and concave cylindrical surfaces are studied, and the performances of the two cases are compared with planar cases. It is shown that the cylindrical conformal reflectarray can be a good candidate of high gain antennas for applications where a slightly curved conformal platform is available.

Based on the basic theory, instrumental effect, main methods and knowledge database on TRIZ, this paper explores the application principles, effectiveness and innovative design along with the experimental methods of TRIZ in field of family exercise equipment. And through the innovative design example of leg stretching equipment, the experimental methods are verified for its effectiveness, which provides an important reference for the application of TRIZ in exercise equipment field.

The pre-processed remote sensing images are often polluted by Gaussian and salt-pepper noises. In order to solve this problem, a Sobel-TV based hybrid model is proposed to de-noise the pre-processed remote sensing images. It uses TV model to de-noise and uses Sobel algorithm to control smoothness of the image’ edge. This proposed method will not only efficiently remove image noise but also simultaneously reserves detail information such as edge and texture. Experimental results show the proposed algorithm achieves better SNR and SSIM compared with other methods. In terms of visual quality, the proposed algorithm can remove the noise of the images and preserve more details, which is important value to preprocess remote sensing image.