最新刊期
卷 5 ， 期 2 ， 2000
- 摘要：Skeleton is a kind of important geometric characteristic of image. How to fast acquire the non-distorted skeleton of image with different shape and size is the precondition for those applications such as shape analysis, characteristic extraction or pattern recognition. Morphologic thinning algorithms based on mathematical morphology are effective thinning approaches to obtaining skeletons of image, which use some pre-defined structure elements with definite shape to operate on the image, only some shifting and logical operations are needed. But the traditional sequential morphological thinning algorithm has the disadvantage that its convergence speed is too slow and using only one structure element to carry out the thinning in every interation may produce distorted skeletons. A new thinning algorithm named fast shape preserving morphological thinning algorithm is advanced in this paper. The algorithm uses four types of structure elements of different direction simultaneously to operate on the image with the hitting operation. The union set of the hitting result is taken as the foundation of whether the outer pixels are exfoliated, thus it can cause the image to converge to its skeleton in every direction fast and symmetrically. The test result of an image of Chinese characters with various fonts enunciates that this algorithm can obtain symmetrical, regular and non-distorted skeletons of the Chinese characters, and its convergent speed is four times more than the speed of sequental morphological thinning algorithm.2348|235|0更新时间：2024-05-08
- 摘要：In recent years the artificial neural network has been developed and applied to remotely sensed data classification problem. Most modal of them are error back-propagation(BP), BP learning algorithm based multi-layer perceptron. Compared to the conventional statistical classifier, BPNN RS image classifier are non-parametric and may have the capacity of more robust proximity especially when distributions are strongly non-Gaussian, but its main shortcoming is its slow training speed, local minimum and even being unable to converge. The Radial Basis Functions Neural Network (RBFNN) modal, integrating the parametric statistic distribution modal and non-parametric single layer perceptron modal, trains faster and more stable than BPNN while keeping the complicated proximity. In this article, the survey and analysis of the RBFNN for the classification of remotely-sensed multi-spectral image is presented, and the RBF RS image classification modal, detailed algorithms and realization procedures is intially raised. The framework which fuses Geo-Knowledge into RBFNN by RBF functions and hierarchical clustering means with optimization evolution theory also are introduced. Finally, the case of practical application of remote sensing land cover classification in Hong Kong region is presented. After the procedure of RBFNN and BPNN approaches are synthetically analyzed, experimental results show that RBFNN approach has more advantages in train time, network structure, knowledge fusion, etc.关键词：artificial neural networks;Radial basis function;Map theory;remote sensing image classification2815|148|0更新时间：2024-05-08
- 摘要：Considering the shortcomings of low accuracies in manual selecting registration control points (RCP) and difficulties in extrcting regions and edges because of SAR speckle noise in the registration for multi-spectrum and SAR remote sensing images, a new registration method for multi-spectrum and SAR images is presented in this paper. With the images of Landsat TM and JERS-1 SAR, this method extracts the RCP from the correlation of the binary classified images, and the RCP accuracies are improved with surface spline interpolation. Because the texture characters of SAR speckle noise are employed in SAR image classification, the speckel noise is no longer obstacles in the registration, and all these efforts make the registration errors of experiment images lower than a pixel.关键词：Remote sensing image registration;classification;Correlation;Surface spline interpretation;Regis-tration control point2373|135|0更新时间：2024-05-08
- 摘要：Matching between two images is often needed in automated visual inspection. Template matching, which is the most principle approach for shape match, is time consuming in case of variation in position and rotation. In this paper, an improved algorithm for 2D shape matching based on Hausdorff Distance is proposed. Hausdorff Distance is used to measure the degree of similarity between two objects to make matching more efficiently. A high dimensional, non-diferentiable, and multi-modal objective function can be derived based on Hausdorff Distance. Although Genetic Algorithm is a powerful and attractive procedure for function optimization, the solution generated by the procedure do not guarantee to be the global optimal. A follow-up optimization scheme such as the line search method is applied, which is capable of finding the minimum value of a unimodal function over a finite search interval. Initially the non-differentiable function is solved using multi-point stochastic search, and the solution is further improved by executing a sequence of successive line searches that approach the optimal to a pre-determined precision. The experimental results show that the proposed method is capable of matching 2D shape with higher speed and precision.关键词：shape matching;genetic algorithm;Line search2398|172|0更新时间：2024-05-08
- 摘要：A new coding method for sequential images called fractal-like coding is presented in this paper. The approach that expresses the self-similarity of an image in fractal coding is used to the description of the inter-similarity between two neighbored images of an image sequence. The new method, compared to the traditional fractal coding, dose not need to satisfy those specific conditions for the later, nor iteration during decoding. Therefor, the coding speed and quality are both increased to an extent. Experimental results on some standard sequential images indicate that the presented method achieves high compression ratio at a reasonable coding quality for sequential images.关键词：Image compression and coding;fractal;affine transform2083|139|0更新时间：2024-05-08
- 摘要：In PACS,CT and MRI scanners provided lots of 3D images every day. For this reason, JPEG ,as a 2D image coding technology is not compatible. In this paper, a three-dimensional medical images compression method based on 3D wavelet transform is proposed. At first, various filter banks in the slice direction are used and the results of rebuild images are compared to find the best filter bank for 3D transform. Then using MSE as performance index, the author presents a new 3D best quantization method independently. And DPCM is replaced by S+P transform in low-resolution subband for enhance coding efficiency. Furthermore, we test various entropy coding methods, based on which a synthesis entropy coding method is found. We think it best fit for wavelet transform coding. At last, an integrity 3D wavelet transform coding method is proposed. We test out the method on lots of 3D images, the results show that this method can elevate compression ratio region from 20% to 47%, compared with JPEG or 2D wavelet transform. This 3D coding method can offset the failure of 2D coding method in PACS perfectly.关键词：3D wavelet transform;Frequency subband;Best bite allocation;Math coding2367|154|0更新时间：2024-05-08
- 摘要：The requirement of GISs' data management demands a universal organization model instead of the conventional spatial and attribute departed data management model. The OODB organization technique is far from mature. Under this condition, a research on how to manage the attribute and spatial data in a universal way in theRDB was conducted. The organization of the spatial database is illustrated in four parts: the total database level table used to maintain the whole spatial data; the atlas table used to maintain the layers' information in a project; the map table used to maintain the layer information; the spatial index table used to store the spatial index information among a layer. The storage process of spatial data is illustrated through spatial data storage and spatial data calculation. A Geographical Relational Database Model was constructed and implemented in WebGIS.关键词：GIS;spatial database;data model2668|178|0更新时间：2024-05-08
- 摘要：In this paper, a new method is proposed based on environment texture mapping to achieve real-time fake Phong shading effect. Using predefined shading texture map and affine texture mapping method, the method is very effective, and can produce pictures with high quality. So the method can do real-time fake Phong shading effect on low-end systems.关键词：texture;texture mapping;Environment texture mapping;Shading2153|145|0更新时间：2024-05-08
- 摘要：Splitting is an important step in solid modeling, and boundary problems are main factors to influence the efficiency and the robustness of splitting algorithms. Quadric surfaces are commonly used in engineering, and their intersecting curves are often degenerated or self-intersected which bringing difficulties for the topological representation and solid' s reconstruction. In addition, boundary overlapping is another problem in solid splitting. Based on geometric representations of quadric surfaces, In this paper, we develop a reasonable strategy for solid splitting, present an effective splitting algorithm considering their intersection curve' s characteristics, and boundary overlapping problems are also be successfully resolved.关键词：Solid modeling;Splitting algorithm;Quadric surfaces2112|142|0更新时间：2024-05-08
- 摘要：The research background of personal identification based on biometrics is introduced in this paper in more detail. The palm character space, which is unique and can be easily classified, is set up through analysis of palm' s physiological characters and detecting edges of palm prints. And the personal identification system with faster recognition speed and higher reliability is established using ART2 neural network.关键词：Biometrics;Palm prints;pattern recognition;Personal identification2163|173|0更新时间：2024-05-08
- 摘要：Facial features localization is very important in face recognition by machine. In this paper a novel method which performs main facial features localization in color information is put forward. It utilizes only red component of the color image to simplify processing cost and improve SNR, segments high gray-level region and low gray-level region in order to localize facial features in segmented regions, first localizes the vertical position of chin, mouth, nose, and eyes with gray-level projection, then localizes the horizontal position of them in associated vertical positions. In this method, nose is first localized to aid other features' localization. The output of this method is positions of eyes, position of nose tip and width of nose, position of mouth and width of mouth, and position of chin. Computer simulation shows that this algorithm has high accuracy and good robustness.关键词：edge detection;Vertical gray-level projection;Horizontal gray-level projection;Valid peak;Houghtransform2248|180|0更新时间：2024-05-08
- 摘要：Computer vision and image-processing techniques have been found increasingly useful for the fruit automatic quality inspection and defect sorting operation. However, real-time fruit surface defect inspection and recognition is still a challenging project due to its complexity. In this paper, a fast approach for box-dimension estimation based on a dual-pyramid data structure is developed. Utilizing traditional fractal dimension and 4 oriented fractal dimensions as input values, a BP neural network is designed for identifying fruit defect area and stem, calyx concave area. The results of experiment show that the approach is effective for real-time defect identification and is accurate. The rate of correct classification is 93% and the executing time of microcomputer for recognition of one undefined blob on the surface of apple is 4~7ms.关键词：fractal;Box dimension;computer vision;Defect;Fruit;Image processing2747|212|0更新时间：2024-05-08
- 摘要：Neural Network is a non-linear dynamics kinetics system, which can do a lot of complex jobs, especially in information science. Inputting image information into computers is an inevitable and troublesome job in many practical systems such as CAD and GIS, etc. An automatic method is needed badly. After analysing fundamental theories of Hopfield model and multi-layered network, a new method to solve the problem realized by a BP algorithm is proposed. During experimental simulations,this algorithm really showed fine performance.关键词：Map recognition;Multilayer network;Intersecting lines recognition;BP algorithm2413|163|0更新时间：2024-05-08
- 摘要：In this paper, a restriction-based virtual hand model that uses a“father and son”link with the joints as nodes has been established based on the analysis of hand structure and joint motions. And then a virtual hand has also been drawn. By analyzing and designing the digital glove and interface, a digital glove-based 3D interactive operation environment used for gesture synthesis, collision detection and collision treatment is realized according to the motion type restriction condition and graphic transformation. In this environment, the virtual hand can conduct different motions and operations such as moving, grasping and releasing etc.关键词：virtual reality;Hand model;Digital glove;Interaction2654|150|0更新时间：2024-05-08
- 摘要：In this paper, a microcomputer image processing system using a PCVISIONplus Frame Grabber is built, and the principle for writing the programs is explained. Because it takes much time to exchange the data with hard disk, so an algorithm based on the properties of DFT is described in order to resolve this problem. The FFT and the image frequency domain processing can be carried out in the conventional memory of a microcomputer using the algorithm without exchanging the data with hard disk. The speed of the image processing was speeded up. A lot of X-radiology have been processed with the algorithm in the microcomputer image processing system. The result shows that the proposed algorithm is efficient and fast.2296|147|0更新时间：2024-05-08
- 摘要：An algorithm for estimating mean square error of the pixel position of scanned and thinned image is put forward in this paper. Based on some experiments, the conclusion is obtained that mean square error of pixel position is generally composed of both constant error and ratio error, the effect of different thinning approach on the accuracy is too small and the scanner may exist regular error etc. Thinned and scanned image is the basis of tracing vectorization and its accuracy is important in such areas as translating it into datum used by GIS. So it is necessary to study pixel accuracy of scanned and thinned image.关键词：Scanned image;Pixel position accuracy;Method of estimation2162|163|0更新时间：2024-05-08
- 摘要：Based on straight property of longitudinal and latitudinal lines in partical scope, contour lines are separated with them automatically, by taking advantage of property that average freeman chain code of straight line is equal to it' s separate one, so as to realize automatic vectorization of contour lines.关键词：contour line;Longitudinal and Latitudinal lines;Chain code of straight line;Separation2459|158|0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