最新刊期
卷 20 ， 期 7 ， 2015
- 摘要：Rate distortion optimization (RDO) techniques play an important role in high-efficiency video coding (HEVC) because these techniques are usually used to select optimal encoding parameters and implement the trade-off between bit-rate and distortion of reconstructed video. In HEVC, distortion is usually evaluated by mean squared error or sum of absolute difference because they are well-understood quantities and convenient. However, they do not clearly reflect human visual perception because video signals are ultimately received by human eyes. Thus, Lagrangian multiplier must be developed by considering human visual system characteristics. In this paper, a novel RDO algorithm is proposed based on human perception and is applied for inter prediction video coding. First, a human perception factor was defined by the features of human vision system about spatial motion region, texture region, temporal motion activity, and illumination in video. Multiplication of spatial motion region, texture region, temporal motion activity, and illumination factors were used to evaluate human perception. Then, Lagrangian multiplier was adaptively adjusted for each coding tree unit based on human perception factor. Finally, through derivation, we determined that variation of Lagrangian multiplier results in the change of quantization parameters. Hence, quantization parameters were further amended according to the relationship between Lagrangian multiplier and quantization parameters. Experimental results showed that the proposed method improves rate-distortion performance compared with HEVC reference software. For the same structural similarity (SSIM) index, the proposed approach saves bit by 3.1% for random access configuration and 4.9% on average for low delay encoding configuration. Maximal bit rate saving can reach 9.0%. The rate-distortion performance of the proposed approach is better than that of representative literature algorithm. For the same SSIM index, the proposed approach can save the bit by 0.7% for random access configuration and 2.2% for low delay encoding configuration on average. We conducted an experiment to explain the role of each perceptual factor. Experimental results showed that the factor of texture region is better than other factors and is not so good as the combination of four factors. In this paper, we proposed a novel RDO strategy that can adaptively adjust Lagrange multipliers based on different visual characteristics. Experimental results showed that the proposed strategy can improve HEVC coding performance and can achieve bit rate saving with the same reconstructed video quality.关键词：high efficiency video coding;human perception;rate distortion optimization;Lagrangian multiplier3446|308|0更新时间：2024-05-08
- 摘要：Side-scan sonar has been widely used in several tasks, such as underwater target detection, tracking of undersea pipeline, and marine investigations. Side-scan sonar image usually contains complex background and serious noise pollution. With regard to these features, an image segmentation method via fuzzy clustering with spatial contextual information is presented to improve segmentation accuracy and operation time. In this study, we selected the combination neighborhood median filter as spatial contextual information. The basic idea of the presented method is as follows. First, we selected a cross-shaped neighborhood and an oblique cross-shaped neighborhood, and medians were calculated in the neighborhood. By introducing a penalty term, we then chose small median to obtain gray level, which integrates gray information and spatial contextual information. Finally, we applied the fuzzy C-means (FCM) clustering method on the gray image that integrated two kinds of information. To prove that the presented method has higher segmentation accuracy and shorter operation time than other fuzzy clustering methods, we compared it with other five kinds of fuzzy clustering method, which include FCM, bias-corrected FCM, FCM_S1, FCM_S2, and FCM combined with traditional square neighborhood median filter. shows that the presented method has higher segmentation accuracy, shorter operation time, and stronger performance of keeping edge information than the other methods because of its combination with neighborhood median filter. By comparing with the traditional FCM and other improved FCMs, the presented method combined with mean filter and traditional square neighborhood median filter can segment side-scan sonar image quickly and effectively and has strong antinoise performance, robustness, and real-time performance.关键词：fuzzy clustering;side-scan sonar image;image segmentation;combination median filter3328|297|9更新时间：2024-05-08
- 摘要：Carotid atherosclerosis plaque is the primary cause of heart disease and stroke. Thus, accurate identification of plaque component is the main basis for plaque rupture prediction. However, multiple sequence MR images are hampered by noise and misalignment. To solve the first problem, an improved denoising method based on wavelet transformation is used. Then, a registration method based on shape context method is proposed to accurately align multiple MR image sequence. We replace the traditional threshold function used in wavelet transform denoising with an adaptive threshold value that changes according to noise distribution in high-and low-frequency subband to achieve effective MRI denoising capabilities. Multiple MRI registration is performed by first identifying edge points with matching shape context. Then, an iterative procedure is invoked to obtain deformation field, which determines B-spline interpolated output image. The proposed denoising method is proved to remove image noise effectively while preserving original information. The overlapping accuracy among multiple sequences of the same artery reaches 96%±0.8%. Results from the experiment show the effectiveness of the proposed method at enhancing image quality.3125|441|3更新时间：2024-05-08
- 摘要：An image inpainting algorithm based on combination of wavelet transform and texture synthesis is discussed to overcome the error repair of the boundary of structure and texture in traditional image inpainting algorithm. The discussed image inpainting algorithm utilizes characters of wavelet transform domain coefficients. Wavelet transform has been used as a good image representation analysis in addition to statistical properties. Multiresolution analysis of wavelet transform is helpful to predict coarse-to-fine image structure. In particular, texture and detailed patterns for natural images must be analyzed. Wavelet can treat these elements altogether. In view of the advantages of image decomposition algorithm, wavelet coefficient statistical properties, and visual effect of edge information of an image, we proposed an image inpainting algorithm based on combination of wavelet transform and texture synthesis. Our reconstruction modeling is based on classical image decomposition model. Some actions have been taken to improve reconstruction performance. An image can be seen as a combination of texture and structure. Thus, the image repair process should fully consider the texture and structural characteristics of an image. At first, the damaged image is decomposed into low-frequency sub-image and high-frequency sub-image with different resolutions via wavelet transformation. In cases where low-frequency component represents image structure, high-frequency component reflects edge changes of an image. Moreover, low-frequency component has a positional correspondence relationship with high-frequency component. Then, sub-images are reconstructed in accordance with their respective characteristics. The sub-image that reflects structural information of an image is reconstructed with fast multipole method, whereas the sub-image that reflects texture information of an image is filled in with texture synthesis based on the characteristics of wavelet coefficient in sub-images. We introduce edge factor in combination with the characters of the wavelet transform domain coefficients to update priority function in the process of reconstituting high-frequency sub-images. Finally, the recovered sub-images are reconstructed with wavelet. Simulation results show that this hierarchical classification method works well in edge damaged blocks. The power signal-to-noise ratio of the final result compared with the traditional algorithm has been improved by approximately 1 dB to 2 dB. The repair results are consistent with human visual perception. Image decomposition model is a widely used image inpainting method. However, fuzzy and mismatching can be generated easily during the repair process. Therefore, when high-frequency component is repaired, changes in factor coefficients of high-frequency components must be introduced to enable the repair process be in accordance with edge direction. In such case, repairing image edge and improving matching block search are top priorities to reduce mismatch error. The proposed method can eliminate point defects in the repair process. Compared with the related algorithms, our algorithm holds good integrated performance. It can effectively repair damaged image with strong edges and rich texture, particularly for the loss scenarios and natural images, to improve image inpainting quality and to be consistent with human visual effects.关键词：image inpainting;wavelet transform;texture synthesis;hierarchical classification2861|425|3更新时间：2024-05-08
- 摘要：The original image inpainting algorithm based on partition block of damaged region (IIPBDR) algorithm is not suitable to inpaint a damaged image with large damaged area that contains broken boundaries with large curvature because this algorithm uses a straight path to inpaint the matching broken boundaries in inpainting process. We aim to obtain a smooth curve path between any two matching broken boundaries. This smooth curve path is then used to improve IIPBDR algorithm, instead of a straight path. The following steps are performed in this study: 1) A smooth-fitting model that satisfies human visual connecting is proposed. 2) Least square is employed to search all coefficients of this model by using the prior information of the matching broken boundaries. Then, a smooth curve path between two matching broken boundaries is obtained. 3) The smooth curve path, instead of the straight path, is used in IIPBDR algorithm to inpaint two matching broken boundaries. “Improved IIPBDR algorithm based on smooth-fitting model” is proposed based on preceding studies. Experimental results indicate that compared with Bertalmio-Sapiro-Caselles-Ballester, Criminisi, IIPBDR algorithms, the improved method can increase peak signal-to-noise ratio value by approximately 2.4 dB, and the inpainted images have better visual effects. Thus, the improved method is more competent than the aforementioned algorithms in inpainting a damaged image that contains broken boundaries with large curvature. The proposed algorithm can solve the problem, in which IIPBDR algorithm is not suitable to inpaint a damaged image with large damaged area that contains broken boundaries with large curvature.关键词：image inpainting;smooth-fitting model;least square;smooth curve path2691|377|0更新时间：2024-05-08
- 摘要：Video target tracking algorithms based on Kalman filter require prior information, such as process noise and observation noise variance. However, we cannot determine the exact values of both noise variance in practical applications. Moreover, noise variance occasionally changes dynamically because of target randomness and background video scene complexity. If noise variance is inaccurate, tracking accuracy is degraded or tracking failure is brought out. In view of these problems, a new solution is proposed in this paper. In combination with the forgetting factor recursive least squares (EFRLS) method, a new algorithm that is applied to video target tracking without use of noise variance is presented in this article. First, mean shift is used to obtain a preliminary estimate of the target position. Then, the EFRLS method is used to estimate the position in the next frame. Experimental results show that the proposed algorithm is significantly better than traditional mean shift algorithm and is equivalent to Kalman tracking algorithm combined with mean shift. In addition, if severe occlusions exist in between targets, this algorithm is better than Kalman tracking algorithm combined with mean shift. The proposed algorithm also has good tracking performance. Setting parameters of noises is not necessary. Accurate tracking results can be achieved when serious occlusions exist or re-emerging occurs after occlusions. The robustness of the new algorithm is then enhanced, which can be used for some engineering applications.关键词：video target tracking;noise covariance unknown;Kalman filter;EFRLS;target occlusion2245|379|1更新时间：2024-05-08
- 摘要：Significant research has been conducted in the field of image haze removal locally and internationally. However, haze removal methods, which have good effects, often take a long time. Moreover, effects of fast methods are general to reach the requirements of applications in many occasions. This study aims to present a method that has fast processing speed and improved effect based on mainstream haze removal methods. The haze removal method via dark channel that is previously suggested by He is simple and effective. However, it keeps residual haze near depth edges after haze removal. Moreover, it leads to color distortion in sky area and large white areas, which do not previously meet dark channel. In addition, its processing speed is slow. This study aims to solve these problems and to present a fast and effective method based on the method suggested by He. The appearance of residual haze is due to block thought adoption via the method of He and the assumption that the transmission keeps unchanged in a local patch. We can abandon block thought, cancel the minimum filtering operation, and use per-pixel processing method to estimate transmission map. The appearance of color distortion is due to extremely low estimation by the method of He for the transmission in the sky area and large white areas, which leads to subtle differences among pixel RGB color channels in these regions, which are magnified nearly 10 times. Thus, we can increase transmission of these regions properly. Atmospheric light can be estimated with quadtree algorithm, which is efficient. Our method has solved the problems of residual haze and color distortion effectively. Moreover, haze removal speed is enhanced greatly because operations of minimum filtering and soft matting or guided filtering in the process of estimating the transmission map are abandoned and the atmospheric light is improved by solving efficiency. The speed of our method is about four times of that of the method by He. Experimental results show that our algorithm can greatly improve efficiency of haze removal and can save the time spend by haze removal under premise of good effect. Our method can keep good effect of haze removal for most haze images. However, if an image has deep scene depth, our method have a general manifestation on its distance scene. On account of the advantage in speed, our method is suitable for real-time demand higher occasions.关键词：haze removal;dark channel prior;fast haze removal;per-pixel processing;guided filter3979|426|7更新时间：2024-05-08
- 摘要：Human motion tracking from monocular image sequences is a challenging work in computer vision. It also has many penitential applications, such as human computer interface, computer animation, and intelligent video surveillance. Methodologies of example-based human motion tracking from monocular camera are explored in this study to meet timeliness, accuracy, and reliability requirements of human motion tracking for real applications. We focus on two main aspects: visual feature extraction and human motion modeling. Based on an example database that is constructed with motion capture data, an example-based approach for human pose estimation from monocular image sequences is proposed. First, we use a motion detection method to extract human region from images. Then, an edge-tracking method is used to detect human silhouette from human region. Second, shape context is used to describe the human silhouette detected from video frames, and candidate poses are obtained from the example database by silhouette matching. Third, we build probability and statistical model of human motion and conducted pose estimation from these candidates. Experimental results on walking, running, and jumping videos demonstrate that shape context-based silhouette representation and matching method can effectively extract human visual feature from image. Compared with other methods, the proposed method can tackle orientation ambiguity problem effectively. Moreover, it is invariant to viewpoints. In this paper, we proposed an example-based method for human motion analysis. Shape context is used for visual feature extraction and matching. Probability and statistical model of motion are used for pose estimation. Experimental results on different types of motion demonstrate that the proposed method can analyze 3D pose from videos effectively, thereby increasing the efficiency and accuracy of human motion analysis. Moreover, the proposed method can solve the orientation ambiguity problem, and is invariant to viewpoints.关键词：motion analysis;example-based approach;shape context;statistical modeling2668|422|0更新时间：2024-05-08
- 摘要：Image-based modeling as a rapid means to reconstruct 3D models has become increasingly popular and gained significant attention. It has been widely used in virtual reality, 3D printing, and digital city construction. High-quality texture mapping is needed to improve visual effect and practical value of the reconstructed object to achieve high-quality 3D visual effects. For a texture mapping, generating a texture image fully automatically, eliminating color differences of an image, and achieving 3D seamless texture mapping are difficult because calibration parameters are inaccurate, images from different viewpoints have different illumination and colors, and 3D digital model has some minor differences. In addition, image pixels are not strictly aligned with model vertex, and one vertex corresponds to pixels of multiple images. Therefore, a seamless texture mapping algorithm based on multiparameter weighted fusion is proposed in this paper. Clustering and segmentation of the triangular grids are conducted according to image calibration information. Reconstruction models are clustered into several grid patches of different reference images, which are sequenced to create a texture image. Pixels of the texture patches are generated by weighted fusion of the reconstructed vertex normal angles, image viewpoints, and model errors. Finally, multiresolution decomposition and fusion are conducted for the texture image by eliminating patch gaps to achieve seamless texture mapping. Unlike traditional methods, all visible image sets are considered in the algorithm, and the influence of images that slightly contribute to the overall texture pixel is reduced through a weighting process to eliminate texture gaps while still guaranteeing texture clarity. We use different test data to verify the method. Experiments show that the proposed algorithm is effective and practical in obtaining a HI-FI 3D texture mapping effect relative to traditional methods. It maintains texture clarity while eliminating seams. The method can obtain satisfactory result even if grid accuracy and image calibration are limited by properly dealing with lighting differences in between images. The proposed algorithm also supports 3D texture mapping of large data. According to the demand for image 3D reconstruction, a seamless texture mapping algorithm based on multiparameter weighted fusion is proposed in this paper. The algorithm fully considered several factors, such as lighting differences, reconstruction accuracy, and limited calibration accuracy of the image. Moreover, texture gaps are eliminated by establishing a smooth weighting equation to fuse multisource information, thereby achieving seamless texture mapping while maintaining texture clarity. In addition, the algorithm can process large data and can be used in city-scale 3D reconstruction field.关键词：image-based 3D modeling;multi parameter weighted;texture mapping;clustering and segmentation;texture patches;multi-resolution decomposition and fusion3029|367|2更新时间：2024-05-08
- 摘要：Yunnan out-of-print woodcuts are special art form of woodcuts. They encompass Chinese traditional culture, and they are a typical representative of artistic characteristic of Yunnan local minorities, which has high artistic value. The digital synthesis of Yunnan out-of-print woodcuts not only solved “out-of-print” woodcuts, but also reserved and inherited this kind of Chinese traditional culture. According to the actual process of Yunnan out-of-print woodcuts, a technique of interactive digital synthesis of Yunnan out-of-print woodcuts is proposed. The preliminary work of the research was collecting nicks from real woodcuts images, and nick libraries were structured with them. In the simulation process of print, we initially used the color that was achieved via color clustering algorithm to layer the image. We then used a painting brush to render the nicks on the image to simulate engraving. After coloring each layer, which simulates printing, we rendered the abstract edge to enhance the contours. Compared with the results of Photoshop woodcut filter, our synthesized results are more realistic and satisfying. The method simulates the actual process of Yunnan out-of-print woodcuts and fills in the gaps of NPR(non-photo realistic rendering) in the area of Yunnan out-of-print woodcuts.关键词：Yunnan out-of-print woodcuts;interactive;digital synthesis;abstraction3047|384|2更新时间：2024-05-08
- 摘要：Polygon offsetting is a fundamental problem in the field of computer graphics, computational geometry, and computer-aided geometry design, and it has many applications. A pixel-based domain subdivision algorithm is proposed in this paper to deal with various types of polygon offsetting effectively. The given domain is subdivided with quadtree data structure, and the sets of all those pixels that are on polygon offsets are calculated with interval arithmetic. For polygons that are composed of line segments only, the shortest distance operator is used between point and line to speed up calculations. Pixel-based domain subdivision algorithm for polygon offsetting can handle different types of polygon offsetting problems, and it is compared with traditional pixel-based polygon offsetting expansion algorithm. The experimental results show that the proposed algorithm can deal with all kinds of polygon offsetting problems effectively. The performance of the proposed algorithm is better on the processing of vertices than that of traditional expansion algorithm. The proposed algorithm is also less time consuming than traditional expansion algorithm. Compared with traditional edge offsetting algorithm, the proposed algorithm has more extensive applicable scope, and it can effectively deal with various types of polygon offsetting problems that traditional edge offsetting algorithm fails. A new pixel-based domain subdivision algorithm for polygon offsetting is proposed in this paper. The proposed algorithm does not need to consider intersection and connection problems, and it can handle various types of polygon offsetting problems that many other conventional algorithms fail, such as when polygon contains arcs and islands.关键词：polygon offsetting;domain subdivision algorithm;interval arithmetic;quadtree3273|330|0更新时间：2024-05-08
- 摘要：Traditional virtual fitness systems employ mouse-or keyboard-based interaction operations and simple message delivery mechanisms. Most traditional fitness systems not only have problems in the user immersion and playability, but also lack scientific guidance of fitness strategies. In this paper, a personalized virtual fitness system design is presented based on natural interaction, highly personalized contents, and user-central feedback regulation model, which can improve the entertainment value of the fitness system and the effectiveness and scientificalness of fitness exercise. The fitness system design in this paper contains three fundamental elements, namely, semantic-based natural interface, highly personalized strategy, and user-central feedback regulation. The interface takes advantage of Kinect interaction equipment to strengthen the connection between a user and a computer with gesture semantics. The interface does not require any input of mouse or keyboard. Deeply customized strategy adequately considers the need of users and gives them personalization configurations with an interactive tool. Game rules and scene presentations can be freely changed by users. Feedback regulation uses a loop control among physiological signals, game show, and exercise strategy. The system monitors user condition and subtly visualizes it into the game show to gain a beautiful scene that users want. Therefore, users positively change exercise strategy. On the basis of system design, we further implement a virtual bicycle fitness system, which uses a fitness bicycle, a Kinect, a heart rate sensor, and a personal computer. This system collects wheel turning of bicycle, gesture information, and user heart rate in real time, and these signals are mapped into virtual 3D scene. The system generates a feedback regulation model among system scene, user heart rate, and exercise strategy, and a scientific fitness guidance is provided in real time. Then, users can adjust exercise strategy on their own, and the system scene can be personalized with gesture-based interactions without mouse and keyboard. Experimental results show that most volunteers are interested in the system, and the proposed design of virtual fitness system can improve the entertainment value of the fitness system and the effectiveness of fitness exercise for users. In this paper, a personalized 3D virtual fitness system design is presented based on natural interface, highly personalized strategy, and feedback regulation. Implementation of the system and results show that this system design has good effect on the entertainment value of fitness system and improves the effectiveness and scientificalness of fitness exercise.关键词：virtual reality;fitness systems;natural interface;serious games;feedback regulation3144|296|2更新时间：2024-05-08
- 摘要：Since the launching of the LANDSAT-8 satellite, the U.S. Geological Survey has released LANDSAT-8 standard scene products around the globe based on worldwide reference system. However, the coverage areas of a standard scene product are too small to satisfy regional remote sensing application, which needs long strip satellite products. The conventional method to obtain long strip satellite product is process-and-mosaic, which brings several problems, such as seam line problem and low processing efficiency. A novel LANDSAT-8 long strip satellite data processing method is proposed to solve the shortages of the conventional method. The proposed method was divided into three steps. First, in the preprocessing stage, auxiliary data were checked, and the imagery and the auxiliary data of LANDSAT-8 long strip satellite data were matched according to the uniqueness of data acquisition time. Second, in the radiometric correction stage, temperature sensitivity correction factor was introduced to eliminate the disproportion in brightness in LANDSAT-8 long strip satellite imagery, and a smoothing algorithm was proposed to avoid the seam line problem inside LANDSAT-8 long strip satellite imagery. Third, in the geometric correction stage, achieving high-precision georeferencing from LANDSAT-8 long strip products was needed for provision of good quality exterior orientation elements. Several algorithms were adopted to refine exterior orientation elements (mainly include orbit data and attitude data). Data verification algorithm were employed to identify and eliminate failure exterior orientation elements. Smooth filtering and Lagrange interpolation algorithm were used to keep continuity of orbits data. Orbital motion affections in gory data were removed, and Kalman filtering algorithm was introduced to obtain high accuracy of attitude data by combining it with the original attitude data. Experiments are designed and conducted on three sets of LANSDSAT-8 long strip satellite data, which were obtained in different time intervals and had different sizes. The conventional process-and-mosaic method were applied for comparison. Experimental results showed that the long strip satellite products produced by the proposed method obtain geometric accuracy same as those of the products produced by conventional method. Moreover, the long strip satellite products produced by the proposed method have no seam line problems. In addition, the efficiency comparison of two methods showed that the processing efficiency of proposed method increased by 45.9% on average. The proposed method achieves high-quality long strip satellite imagery with a high efficiency than the conventional method, thereby greatly satisfying the needs of regional remote sensing applications.关键词：Landsat-8;long strip products;radiometric correction;geometric correction3678|280|1更新时间：2024-05-08
- 摘要：With the development of imaging spectrometer, the spectral resolution and space resolution have been enhanced dramatically which makes a challenge to hyperspectral unmixing processing. So a new distributedhybrid parallel model has been proposed to accelerate hyperspectral unmixing processing. In order to reduce the computational complexity of endmember extraction algorithm, the original serial method has been redesigned for parallel computation and a fast implementation of improved method has been proposed based on partitioned determinant operations. At the same time, the Jama and JCuda components have been used to accelerate the computation in distributed cluster environment.Result The proposed distributed hybrid parallel method plays a large role in accelerating hyperspectral unmixing based on maximum simplex volume algorithm. The improved MapReduce model method is near ten times more rapid than the original method for the hyperspectral image which size is 400×400×224.And the more computational load, the more speed up.Conclusion In this paper, the proposed distributedhybrid parallel method can increase the hyperspectral unmixing processing speed dramatically. At the same time, the partitioned determinant solving method can reduce the complexity of MSVA algorithm.The experimental results indicate that the proposed method can achieve great speedup to the algorithms which have characters of parallel executive tasks、lower data transmission between main node and sub nodes and massive calculations.关键词：spectral unmixing;distributed computation;endmember extraction;partitioned determinant3205|519|1更新时间：2024-05-08
- 摘要：Dengue fever is a global health problem. Probing its spatio-temporal characteristics and building corresponding spatio-temporal models from geographical analyses may constitute a novel approach to its effective prevention and control. Spatio-temporal data mining and spatio-temporal process model-building were used to examine spatial relationships between dengue fever and environmental, climate, geographical, and population factors, as well as spatial autocorrelation of dengue fever cases, aiming to identify influencing factors of dengue fever. A Genetic Algorithm (GA) was introduced to overcome the weakness of Back Propagation (BP) neural network models, which is typically subject to local optima. The improved model was applied to spatio-temporal simulation of dengue fever. the spatio-temporal diffusion of dengue fever was significantly associated with temperature, humidity, residential areas, traffic, and population density. A remarkable spatial autocorrelation was found amongst the dengue fever cases. Nonlinear relationships were revealed between the occurrence and diffusion of dengue fever and environmental, climate, geographical, and population factors. The improved GA-BP neural network model may enhance the accuracy of simulation of spatio-temporal diffusion of dengue fever (RMSE=0.081). The occurrence and diffusion of dengue fever were affected by multiple factors. The GA-BP neural network model can effectively simulate spatio-temporal processes of dengue fever. The improved model could likewise apply to simulations of other aedes-related diseases.关键词：dengue fever;spatio-temporal data mining;genetic neural network model;spatio-temporal simulation;influencing factor3126|265|3更新时间：2024-05-08
- Postal code：100190
- Tel：010-58887035/58887030/58887418 Email：jig@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备05080539号-4
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0