最新刊期
卷 21 ， 期 1 ， 2016
- 摘要：The use of a shape is a popular way to define objects, and efficient shape coding is a key technique in object-based applications. Shape coding is also a hot research topic in the field of image and video signal processing, and many shape-coding techniques have been proposed. Among these methods, chain-coding is a popular technique that can be used for lossless shape coding. However, most existing chain-based shape-coding methods have not exploited the spatio-temporal redundancy contained within shape image sequences. Similar to the existence of strong spatio-temporal redundancy within and among video textures, a strong redundancy also exists within and between object contours. This redundancy can be exploited to improve coding efficiency. Hence, in this paper, a novel chain-based lossless shape-coding scheme is proposed by exploiting the spatio-temporal correlations among object contours to acquire high coding efficiency. First, for a given shape image sequence, the contours of visual objects are extracted, thinned to perfect single-pixel width, and transformed into chain-based representation frame by frame. Second, the activity of object contours in each frame is detected and evaluated. The shape frames are classified into two coding categories on the basis of this activity: intra-coding frames and inter-coding frames. If the contour activity in a frame is larger than a preset threshold, the activity will be encoded as an inter-coding frame; otherwise, it will be encoded as an intra-coding frame. For an intra-coding frame, the spatial correlations within object contours are exploited on the basis of chain-based spatial prediction and compensation. For an inter-coding frame, the temporal correlations among object contours are exploited on the basis of chain-based temporal prediction and compensation. Finally, a new method is introduced to efficiently encode the prediction residuals and motion displacements by analyzing the constraints among chain links. To evaluate the performance of the proposed scheme, experiments are conducted and a partial comparison is performed against some well-known existing methods, including the lossless coding scheme proposed by the Joint Bi-level Image Experts Group (JBIG) , the improved lossless coding scheme proposed by JBIG (JBIG2), the Context-based Arithmetic Encoding with Intra-mode (CAE Intra) of MPEG-4, the Context-based Arithmetic Encoding with Inter-mode (CAE Inter) of MPEG-4, the Digital Straight Line Segments-based Coding with Intra-mode ( DSLSC Intra) and the Digital Straight Line Segments-based Coding with Inter-mode (DSLSC Inter), are also presented., The experimental results show that the average code length of our scheme is only 28.4% of JBIG, 32.3% of JBIG2, 39.9% of CAE Intra, 78.1% of CAE Inter, 48.4% of DSLSC Intra, and 94.0% of DSLSC Inter. As a whole, the proposed scheme outperforms all existing techniques and is considerably more efficient than other methods. As far as we know, the DSLSC Inter is the most efficient lossless shape-coding approach. However, compared with the DSLSC Inter, the proposed scheme has an average code length that can be reduced by 6%. The proposed scheme has wide prospects in many object-based images and video applications, such as object-based coding, object-based editing, and object-based retrieval.关键词：visual objects;shape coding;predictive coding;high efficient coding3019|378|1更新时间：2024-05-07
- 摘要：This paper proposes a new interest point detection method by using contour shape and the direction derivative information entropy of surrounding pixels. The proposed method can improve localization accuracy and noise robustness. First, the multi-direction imaginary parts of Gabor filters are used to extract the gray variation information of input images to acquire the second small direction derivative. Scale multiplications are used as the measure to detect interest points. The proposed method is based on contour-based methods and is improved in two ways: the information of edge shape and the gray variation are combined to detect the interest points; the sensitivity of scale multiplication to local variation and noise on the edge contour is reduced. Second, the Canny edge detector is used to extract edge maps and fill gaps to obtain the edge contour. Third, the second small direction derivative information entropy of the edge contour pixel and its surrounding pixels is computed. The normalized entropy is used as new corner measure by comparing with the threshold. Finally, non-maximum suppression is applied to the result obtained in the third step. A local window (9×9) is used to slide through all the pixels of the edge contour. If the center pixel of the window is the local maxima within the window, the central pixel scale product is retained; otherwise, the center pixel will be set to zero. Compared with the gray method, which analyzes the contour shape, or the curvature-based method, the proposed method combines the idea of two algorithms that use the gradient direction entropy of the contour pixel and its neighbor pixels as the corner measure. Furthermore, unlike the gradient direction variation in homogeneous regions and edge lines, the information variation on the interest point presents an anisotropic feature. The information entropy according to the second small direction derivative (the first small direction derivative may be zero) is used as a new corner measure to improve localization accuracy. The proposed method has an average score of 1.625 in the performance index, which is significantly higher than the method of Harris (3.25) and He and Yung (2.625) and the CSS (2.5) interest point detection operator. Compared with the three state-of-art algorithms, our approach is competitive with respect to the repetitive rate of interest point, detection accuracy, and noise robustness.关键词：interest point detection;the imaginary parts of Gabor filters;gray variation information;second small gradient direction;information entropy2828|283|2更新时间：2024-05-07
- 摘要：Character is a basic element of graphics. Its anti-aliasing algorithm is one of the basic content of computer graphic research. To solve the jitter and bad real-time capability problems caused by inaccurate grayscale distribution in the traditional anti-aliasing algorithm for character rotation, an efficient anti-aliasing algorithm is proposed. First, the algorithm is based on the principle of area sampling and assigns grayscales according to the four pixels around the neighborhood.This feature increases the preciseness of the coordinate positioning of each pixel and solves jitter problems when the character is dynamically displayed. Second, the void phenomenon appearing after the character rotation is eliminated by the reverse coordinate transformation. Finally, the algorithm makes full use of the parallel computing ability of FPGA and is optimized in time and space. Experimental results show that the proposed algorithm solves the problem of character jitter well. The simulation speed obtained by the new algorithm is nearly six times more than that obtained by the traditional algorithm and is also faster than that of the Quincunx algorithm by 30%. Compared with other algorithms, the advantages of the new algorithm are reflected in two aspects. On one hand, the grayscale of characters has smooth transitions and better anti-aliasing effects; on the other, it is fast and operates in real time. The new algorithm is applicable to the display of dot matrix characters on various occasions in engineering, particularly in high real-time environments.关键词：anti-aliasing;area sampling;character rotation;grayscale distribution;field programmable gate array2895|233|0更新时间：2024-05-07
- 摘要：The detectionof maize seeds and other crops is a key problem in the field of agricultural informatization.In this paper, to enhance detection and improve accuracy rate, we present a detection algorithm for maize seed on the basis of extreme learning machine (ELM) and multi-scale feature fusion. In the first part of this paper, the feature of maize seed is described as the combination of local features and global features. Local features can be described as a multi-scale histogram of oriented gradient, and global features can be described as the color feature of HSV. In the second part, ELM will be used as the detection algorithm against long training period and slow detection speed, which are the characteristics of traditional BP neural network and SVM. Furthermore, the detection model uses the parallel algorithm to significantly decrease the time used in the training of each classifier. Furthermore, the high resolution of the original image can cause long detection periods and consume a large amount of memory. To address this problem, we propose a local means of an image compression algorithm. Finally, considering that the sliding windows of centralized scanning can produce a problem in creating multiple overlapping windows capturing the same object, we propose a local window fusion algorithm based on fuzzy clustering to address this problem. The simulation results show that the method proposed is able to accurately detect maize seeds. The accuracy of detection of maize seeds can reach 97.66% with an error of less than 0.1%. Compared with traditional methods, the no damage method proposed in this paper can improve the speed and accuracy of detection and has no strict hardware requirements.关键词：extreme learning machine;feature fusion;object detection;windows fusion;fuzzy clustering2867|301|2更新时间：2024-05-07
- 摘要：Infrared (IR) images are capable of showing hidden objects in an environment where image quality is low. Visible images have high resolution and good image quality but cannot show hidden objects. Image fusion attempts to combine information content from multiple images to obtain a fused image with IR object features from the IR image and retain the visual details provided by the image. Hence, the fused image has the advantages of visible and IR images and is suitable for subsequent processing tasks. A new image fusion method based on nonsubsampled contourlet contourlet transform( NSCT ) and compressed sensing is presented for visual image and IR image. First, the IR and visual images are transformed by NSCT to obtain a low frequency sub-band and a series of high sub-bands with diverse scales and directions. Second, high-frequency sub-bands are fused with the rule of the maximum of local energy. The low-frequency sub-band is measured by a scrambled matrix to produce an observed vector, and the observed vectors are fused by the rule of the maximum of standard deviation. Finally, the fusion image is obtained by the inverse NSCT. Different fusion rules based on NSCT are simulated to compare with the new fusion method. Quantitative analysis is conducted for the fused image under parameters such as entropy, spatial frequency, standard deviation, and structural similarity. Experimental results show that the new proposed method can provide better fusion than other fusion methods in terms of subjective visual quality and objective fusion metric values. The proposed method can also preserve the details of the visible light image and the legible target of the IR image. The fused image enables the extraction of the target location for easy observation and provides information for the further processing of tasks. This image can be extensively used in many fields such as target detection, target localization, remote sensing, and robotic vision.关键词：image fusion;nonsubsampled contourlet transform;compressed sensing;infrared image;visual image3651|294|18更新时间：2024-05-07
- 摘要：Space resection is one of the fundamental issues in the field of digital photogrammetry and computer vision. Traditional methods based on control points limit the robust automation of the space resection procedure. The primary objective of space resection is to effectively and reliably describe the relationship between object points, the corresponding image points, and the projection center. One of the classical solutions to space resection is based on collinearity equations. A solution based on collinearity equations can represent the rotation matrix in many ways, such as Euler angle, Rodriguesmatrix, quaternion, and dual quaternion. Existing space resection methods are mainly based on point primitives, and do not fully consider the geometric topological relations of space lines; the geometric constraint conditions of the collinearity equation are weak. In this paper, a space resection method based on the Plücker line is presented. Plücker coordinates, a representation of space lines consisting of six parameters,has the advantages of clear geometric meaning,concise formation, and cost-effective computation.In this method, the imaging ray is representedin Plücker coordinates to obtain the Plücker line.Thereafter, on the basis of the screw movement of Plücker lines in space, the object point is obtained on the line and is identified by the image point and projection center. Finally, a collinearity equation based on the imaging ray of Plücker line is established. Therefore, the relationship between the control line in object and their corresponding line in image is established. By using the Plücker coordinates of some control lines in the object and corresponding image lines, the exterior orientation elements can be anti-calculated. The tests performed on aerial images and close-range images with small inclination indicate that the maximum error of the line elements of the exterior orientation elements decreases from 0.88 m to 0.53 m and from 0.06 m to 0.01 m; the maximum error of the angle elements decreases from 0.016 2 radian to 0.002 9 radian and from 0.087 radian to 0.066 radian compared with the traditional Euler method. For images with large inclinations, the space resection method can be used for correct calculations, where as the Euler method cannot. The Euler method has strong dependence on the initial values of attitude elements and line elements. When the initial values are poor, such as a large attitude, the iterative calculation will slowly converge or even not converge. The Plücker line method uses dual quaternion to represent the rotation matrix, which overcomes the aforementioned issues by uniformly representing the attitude and line elements. The Plücker line method avoids the complicated process of solving the attitude and position elements separately, thus significantly improving efficiency.Given the enhanced intensity of the geometry of the model, the method obtains high positioning accuracy and can be used for the positioning of the exterior orientation elements of images with large inclination.关键词：line photogrammetry;space resection;Plücker;collinearity condition;screw displacement3057|282|0更新时间：2024-05-07
- 摘要：Saliency detection has become a highly active research field in recent years. Considering that many traditional methods suffer from insufficient feature learning and bad robust detection, this study proposes a novel saliency detection model based on deep convolutional neural networks. First, a pixel with similar characteristics is clustered by using superpixels and the target edge is extracted by imitating the human visual cortex cell to obtain the region and edge features. Thereafter, image regions and edge features are identified by convolutional neural networks to obtain the corresponding target-detection decision confidence images. Finally, we introduce the output of the deep-convolution neural network confidence coefficient into the conditional random field to calculate energy minimization. The discrimination of saliency and non-saliency is realized to complete the saliency detection task. Compared with the state-of-the-art method, the detection accuracy of our algorithm increases by approximately 1.5% in the MSAR database and 5% in the Berkeley database. Furthermore, our detection algorithm produces the best results whether in natural/artificial construction scenarios or large/small objects. Our detection algorithm can avoid the uncertainty brought by manual features and has high robustness and universality. Experimental results show the superiority of our proposed algorithm to the method using shallow artificial features. Both subjective visual pleasure and objective detection accuracy attest the effectiveness of the proposed algorithm.关键词：saliency detection;superpixel;convolutional neural network;conditional random field6911|448|28更新时间：2024-05-07
- 摘要：The Delaunay triangulation is a fundamental problem in the field of computational geometry. It is widely used because of its excellent characteristics. To construct a Delaunay triangulation network efficiently and accurately on a large-scale point set, an improved Delaunay triangulation algorithm based on priority point is presented in this paper. An initial base edge is selected from the convex hull edges in anticlockwise order. A Delaunay triangle is constructed by base edge and the third point in the anticlockwise order, which maximizes the angle opposite the base edge. Whether the generated two edges need to be expanded is determined by the array of the needed expanded edges. The first-in-first-out strategy is used to extract the base edge from the array of the needed expanded edges. The local Delaunay triangulation is constructed around the priority point and accelerates the priority point to become a closed-point. Then, the closed-point is removed. The running time ratio of improved algorithm to classical algorithm is less than a third when using the same point set, and the ratio gradually decreases with increasing point set scale. Compared with the classical algorithm, the improved algorithm has significant enhancement in time efficiency. The improved algorithm has better adaptability for point set scale and high efficiency for constructing Delaunay triangulation. It can also be used for large-scale point set Delaunay triangulation.关键词：computation geometry;Delaunay triangulation;growth algorithm;first in first out;closed-point;priority point3908|284|6更新时间：2024-05-07
- 摘要：To overcome existing problems in generating digitallg reconstructed radiongrapl (DRR) and to measure the similarity needed for large computations in the process of 2D/3D medical image registration, a hybrid registration algorithm was proposed. This algorithm combines pattern intensity with gradient on the basis of the Bresenham line generation algorithm. First, the space coordinate system, in which the position of the virtual point light source, 3D volume data, and projection panel can be determined, is established. Second, virtual point light source emits virtual light and the light passes through volume data. The DRR image can be obtained by computing the gray value of points projected by every virtual ray to the image panel via the ray-casting algorithm, which is an enhancement of the Bresenham line generation algorithm. Third, the registration components, including interpolator, space transformation component, optimizer, and pyramid filter, are instantiated. The pyramid filter is initialized such that the number of layers and shrinkage factor of each layer are set. Fourth, the reference image and floating image are processed separately into different resolution sequence images and image pyramids. The image resolution increases from top to bottom in the two pyramids. The arranged images in a pyramid correspond to those of another pyramid. The transformation parameters are calculated by the improved mode intensity measure, and the optimal transformation parameters are obtained by the improved Powell algorithm. The above process will be executed circularly until the image registration is complete. The experimental results show that compared with an algorithm based on correlation coefficient, mutual information, and pattern intensity, the proposed method shows a substantial increase in registration efficiency. The completion time of the simulated registration experiment and clinical registration experiment are 76.2 s and 64.9 s, respectively. The proposed method increases efficiency by 3~6 times compared with the traditional registration algorithm. The hybrid algorithm not only has strong robustness and high clinical requirement accuracy but also significantly reduces the time required for registration.关键词：2D/3D registration;improved Ray-Casting;Bresenham line generation algorithm;improved pattern intensity (PI);multi-resolution;improved Powell algorithm3959|347|8更新时间：2024-05-07
- 摘要：In the interpolation process,biomedical images can be converted to isotropic discretedata. Thus, biomedical images are convenient for manipulation and analysis. Inter-slice interpolation is often required for the 3D reconstruction of medical images. Although many interpolation methods have been proposed in published literature,most of these methods do not consider the gray levels and objective shape variation of images. Moreover, the calculation processes of existing methods are relatively complicated. Therefore, an interpolation algorithm based on the combination of the wavelet and Lagrange polynomials is proposed in the current study. First, images were decomposed by using wavelet analysis to obtainthe positions of wavelet coefficients that belong to the edges. Second, the Lagrange polynomial was applied to interpolate the intensities and positions between the corresponding wavelet coefficients of the cross-sectional images. In this work, the proposed method used three sets of patients' head magnetic resonance images from clinical settings compared with the linear gray-level and cubic interpolation methods.One slice in these data sets is estimated by each interpolationmethod and compared with the original slice by using two measures: number of points of disagreement and mean-squared difference. By using the proposed algorithm, the number of points of disagreement decreased by 10% to 50%, and the mean square error decreased by an average of 3%.The interpolation image smoothed the gray levels and shapes between the original cross-sectional images, thus satisfying the requirements of medical image interpolation. Compared with the linear interpolation and Cubic interpolation methods, the proposed algorithm is able to extract the image shape transform with wavelet transform.Therefore, the proposed method has certain advantages in dealing with the rapid changes of sequential images. In this study, the proposed method can ensure high image quality and reduce calculation errors. The interpolated images can be used to efficiently perform 3D reconstruction for the object tissues of medical images.关键词：cross-sectional images;inter-slice interpolation;wavelet transform;Lagrange polynomial3288|415|2更新时间：2024-05-07
- 摘要：The speckle appearing on SAR images is a natural phenomenon generated by the coherent processing of radar back scattering. The presence of speckle not only reduces the interpreter's ability to resolve fine details but also hinders the automatic segmentation of such images. In remote sensing image analysis, the spatial structure hidden in an image may contain even more information than the pixel itself, particularly for SAR images, where the large variability owing to the speckle noise makes the single pixel value unreliable. In particular, Markov random field (MRF) presents many interesting properties to model the relationship among neighbor pixels, which not only allows the creation of segmentation techniques that are able to improve segmented results in a statistically optimal manner but also provide an efficient regularization method, which is necessary because image segmentation is an inverse problem that is generally ill-posed from a mathematical point of view. However, MRF-based spatial dependency models lead to an extremely large amount of computations for the estimation of model parameters with a standard expectation-maximization (EM) algorithm. Moreover, the EM-based model parameter estimation is performed over the whole image and does not reflect local image properties. To improve MRF-EM image segmentation, this paper proposes the identification of the global MRF segmentation model by distributing a set of EM procedures within a multi-agent framework. The image is segmented into situated agents, which estimate the parameters of an MRF model. Local features are changed by a cooperation agent with genetic algorithms (GA). To address a statistical approach to SAR intensity image, each pixel in the image is assigned a label to indicate the homogeneous region to which it belongs depending on the statistical properties of the pixel and its neighbors. The individual labels form a so-called label field, in which labels are modeled with a prior Potts model where labels of neighboring pixels will tend to be similar. In each homogeneous region, the pixels are modeled with Gamma distribution with parameters that are dependent on the type of region. By Bayesian theory, the segmentation model can be built by producing the image model and label model. To simulate the segmentation model, an approach combining EM algorithm and GA within a multi-agent system (MAS) framework is proposed. The MAS is composed of a series of segmentation agents and a cooperation agent. Each segmentation agent initializes a global segmentation by EM algorithm, and the cooperation agent employs GA to implement global optimal segmentation. The parameters of the model are estimated by E-and M-steps, and global segmentation is also completed simultaneously. GA attempts to obtain an optimal segmentation by genetically breeding a population of individuals obtained by the segmentation agents. Each individual is encoded by the label field as a chromosome, and each chromosome defines a measure of fitness built by the segmentation model. Fitness can be applied to determine the ability of the chromosome to survive and to produce an off-spring. To obtain optimal segmentation, the selection operator directly inherits individuals from parents by roulette, and the crossover operator exchanges some pixels from pairs of parents to produce new off-springs.Thereafter, the mutation operator changes certain site labels, which are randomly selected and replaced with another category depending on its neighborhoods. Experiments are conducted with simulated and real SAR intensity images by the proposed method and EM algorithm, respectively. Results show that the proposed methodhas better image segmentation performance than the EM method. The problems of edge blur and segmentation mistake in homogeneous regions are solved by combining the distributed EM procedures and genetic operators. The selection operator diversifies the segmentation results, and the crossover process performs global exchanges.Furthermore, the mutation process ensures that the agents can escape from a local point. Therefore, the global optimal solution can be obtained. To quantitatively analyze this method, the overall accuracy and Kappa coefficient of the producers and users are calculated from the confusion matrix and are compared with those of the EM algorithm. Result shows that the presented approach is more accurate than the EM algorithm. All experiments demonstrate the robustness and veracity of the proposed approach. This paper presents a distributed segmentation approach that combines the EM algorithm and GA within the MAS framework. The experiment results indicate that the proposed approach is effective and promising. Although this method produces excellent results, efficiency still needs improvement.Furthermore, MAS has many functions that can be used for segmentation. In the future, our research will focus on the definition of each agent and the interaction among agents (e.g., communication).关键词：multi-agent system;Markov random field model;expectation maximization;genetic algorithms;SAR image segmentation2913|266|5更新时间：2024-05-07
- 摘要：This paper presents our approach on registering aerial stereo imagery to LiDAR point cloud through iterative parallax mutual information computation. Proposed approach takes photogrammetry collinear equation as strict mathematic mode and three stages are involved: First, dense disparity map is generated from aerial stereo imagery using SGM algorithm; Second, LiDAR depth image with same size and approximate spatial resolution, geometrical distortion to the unregistered aerial image is generated, based on the perspective projection of LiDAR point cloud using initial image orientation parameters. In this step, supposed 2D rigid transformation between LiDAR depth image and the disparity map from aerial stereo imagery to be registered is estimated through maximizing their Mutual Information. As a result, by taking LiDAR depth image as a "bridge", visible LiDAR points are roughly mapped to aerial image pixels based on the estimated geometry transform parameters and the known projection relations between Lidar point clouds and its depth image. Third, with all mapped aerial image pixels as observed values and their parallax mutual information as weight, photogrammetry space resection algorithm is implemented to obtain improved image orientation parameters. Keeping repeating above two stages until the given iteration calculation condition is met and the registration of LiDAR point clouds with aerial stereo image is realized. Selected aerial stereo imagery with size 7 216×5 428 pixel, 60% overlap and spatial resolution 0.5 m are automatically geo-registered to LIDAR point cloud (averaged point distance 1.5 m, horizontal accuracy 25 cm) and approximate to 1 pixel registration precision is obtained. The experimental results show that proposed algorithm has moderate registration precision and is in high degree of automation. Theoretically, with known interior of parameters (IOP) and initial exterior of parameters (EOP), proposed algorithm is suitable for various aerial stereo imagery taken for different types of scene and thus, has good value in application.关键词：image geo-registration;LiDAR;parallax mutual information;iterative registration3168|260|2更新时间：2024-05-07
- 摘要：Noise in magnetic resonance images(MRI) lowers its quality, affect the visual effect and computer aided diagnosis. This paper designs an effective MRI denoising algorithm aim to remove Rician noise in MRI . In the complex wavelet domain through dual-tree complex wavelet transform(DT-CWT), combined with Bilateral Filter (BF) and NeighShrink based on Stein's unbiased risk estimation(NeighShrinkSURE)、BivariateShrink, this paper designs an effective MRI denoising algorithm which fully consider the noise distribution characteristic in MRI and wavelet coefficient's inter and intra-scale dependencies. The performance of this method mainly depends on the estimation precision of the noise standard deviation in the coefficients of square MRI after DT-CWT transform, then relate to the parameters of BF and two shrink methods' weight factors. In order to make the cooperative method show the best performance, this paper takes mean square error (MSE) 、peak signal-to-noise ratio(PSNR)、structural similarity index (SSIM) as the image quality evaluation indexes to correct traditional noise standard deviation estimation method, determine the parameters in BF and the weight factor between two shrink methods. This paper designs an effective algorithm that combins three methods in the dual tree complex wavelet domain. The experimental results of image denoising show that in the aspect of visual quality, indicators PSNR and SSIM and elspsed time, the proposed method's comprehensive performance is superior to several traditional MRI denoising algorithms ,the PSNR ratio has improved by approximately 0.51~1dB,the SSIM ratio has improved approximately 5%~10%. Denoising through DT-CWT transform is superior to the basic wavelet transform, the filtering accounts for inter-scale dependency and neighboring similarities , the use of bilateral filter enhances the low frequency part of image, aiming at removing Rician noise in MRI, the proposed algorithm has better noise reduction while preserve image's margin and detail.关键词：MRI denoising;DT-CWT;neighShrink;bivariate shrink;bilateral filter3428|291|5更新时间：2024-05-07
- 摘要：The traditional augmented reality algorithms for natural image detect and extract the natural features of a template image for multi scales. At the tracking stage the algorithms match the screen features with all of the features extracted from the template image, which produces a large amount of useless computation and reduces the efficiency of feature matching. To solve this problem, this article presents an algorithm which can locate the scale and area of the template image fast and accelerate the speed of 3D tracking. In the preprocess stage, the image in each scale is partitioned into small areas, and the features of the template image are managed according to the scale and area. In the real-time tracking stage, the method locates the scale and area of current camera image quickly using some computation geometry algorithms so that the scope of feature matching is decreased. The search field of feature matching, cuts down the time spent by feature matching, compared with the traditional method, the time spent by feature matching is cut short by about 10 times and the time spent by tracking one frame is cut short by 1.82 times when the resolution of template image is high. The frame rate of the 3d tracking algorithm keeps at 15 frames per second on average. Our method can be used to the tracking of natural image on mobile platform, especially when the resolution of image is high, the method can narrow the search field of image scale and area and promote the efficiency of real-time 3d tracking algorithm.关键词：augmented reality;three dimensional tracking;feature matching;mobile platform;fast locating;templateimage3126|266|2更新时间：2024-05-07
- 摘要：As a popular research direction in computer vision, the development of local invariant feature algorithms has become more and more mature and stable. But now, almost all the feature point extraction algorithms cannot give the accuracy index of feature points. In fact,the precision of feature points'position is requested in many areas,such as device calibration and visual inspection. To solve this problem, this paper proposed a feature point precision index-feature point's range. Since there are always disturbances in the process of acquiring images,it is difficult to obtain absolutely perfect images and feature points. This paper defines the feature point's range as the range of a feature point's fluctuation in different conditions which is stable in different disturbances. This paper choses the most popular algorithms of local invariant feature named SIFT(Scale-invariant feature transform)as an example. To make the result more intuitive and clear,the experiment selects those pictures whose backgrounds are simple and objectives are clear. This paper analyses the fluctuation of feature points in noise conditions,fuzzy conditions ,light transform conditions and all those disturbances which are the common interferences in actual operations to achieve the fluctuation range of different feature points. First,this paper establishes the experimental galleries. We make not only a noise gallery, light treatment gallery and fuzzy diagram gallery but also a gallery which contains the three disturbances. It is worth noting that considering the randomness of noise,we generate 100 pictures for one noise intensity. Secondly, the stable feature points which can be detected in all disturbances are found by matching. We get 16 points in this experiment which has its own point cloud. Once more, because different points fluctuate differently in different situations, we use a circle to fit every point's fluctuation in different conditions. It means we use a circle to fit every point's cloud. The radius of those circles can characterize those feature points' ranges. Last but not least,this paper uses histograms which is very intuitive to describe every point's fluctuation. In addition,those points' coordinates are provided. In this experiment we gain 16 stable points. This experiment shows that the fluctuation ranges of different feature points are different,but there are still a part of feature points whose precisions are higher. The points whose fluctuations are smaller in the case of the presence of interference can be considered as better and more accurate points. Selection of feature points is based on the following work. If the requirement of precision is higher,a lower threshold should be designated. Therefore,fluctuation range can characterize the precision of different points very well. This can underpin the related work. Although this paper choses SIFT feature point as an example,other local invariant feature points' have similar properties. This paper provides an idea and method to study feature points' properties to be helpful to the related work.3070|240|0更新时间：2024-05-07
- 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