82 results about "Cost aggregation" patented technology

The invention relates to the technical field of stereo vision, in particular to a stereo matching method. The method solves the problem that the accuracy of disparity correction optimization of the existing stereo matching method is insufficient. The stereo matching method based on disparity map pixel classification correction optimization comprises the following steps that (I) cost aggregation is conducted by taking a left view and a right view as references and based on a method combining a gray scale difference with a gradient, and a left disparity map and a right disparity map are obtained and subjected to left and right consistency detection to generate an initial reliable disparity map; (II) correlation credibility detection and weak texture area detection are conducted, and a pixel is classified into stable matching pixel points, unstable matching pixel points, occlusion area pixel points and weak texture area pixel points; (III) the unstable matching points are corrected by an adaptive weight algorithm based on improvement, and the occlusion area points and the weak texture area points are corrected by a mismatching pixel correction method; and (IV) the corrected disparity maps are optimized by an algorithm based on division, and dense disparity maps are obtained.

The invention discloses a binocular stereo matching method based on a convolutional neural network. For matching cost calculation, context information is integrated by using dense blocks on the basisof initial features. For matching cost aggregation, a regularization cost amount of a small coding and decoding structure is provided. And for parallax calculation, a differentiable soft argmin operation is executed on the parallax dimension of the cost quantity to obtain an initial parallax. For parallax refinement, the initial parallax is guided to be refined by taking a residual block as a mainpart and taking similarity measurement as an auxiliary part. According to the method, four stages of a stereo matching algorithm are strictly followed, and four steps are integrated into one network,so that the network can be trained end to end. According to the stereo matching method, contextual information is integrated in the feature extraction process, mismatching of pixel points in an ill-conditioned area is effectively relieved, memory occupation and running time in the network training / speculation period are remarkably reduced through small and medium coding and decoding structures inthe regularization process, and parallax prediction precision is improved.

The invention discloses a distance measuring method based on a binocular camera. The method comprises the steps that a left camera and a right camera are calibrated, and the calibration of the binocular camera is achieved by calculating the rotation and translation matrix between the left camera and the right camera; distortion correction and three-dimensional correction are conducted on the radial distortion and the tangential distortion of the image; cost calculation, cost aggregation, parallax calculation and parallax optimization are carried out on images formed by the binocular camera insequence; depth calculation is carried out. According to the three-dimensional matching algorithm of a self-adaptive region, the self-adaptive region is constructed to carry out cost polymerization, so that the error and the time are reduced, and the error is reduced by means of the method for optimizing the scanning lines in the four directions (up and down, left and right) by means of binocularimage correction, the parallax precision is improved. A voting method and other methods are adopted in the parallax optimization process, the shielding points and the mismatching points which are matched with the parallax are eliminated, and the parallax matching precision is further improved, the depth of the to-be-measured point is finally calculated, the precision is high, and the operation time is short.

The invention discloses a nonlocal stereo matching algorithm combined matching cost algorithm and parallax joint filling algorithm. According to the algorithm, a nonlocal stereo matching algorithm of cost aggregation based on a minimum spanning tree is used, and noise of an original image is removed in the preprocessing stage by using the algorithm based on the improved minimum spanning tree so that texture information can be enhanced and the enhanced image can be acquired. Then information of the original image is used in the matching cost stage and information of the enhanced image is also used. Therefore, local information of the image is acquired and global information is also acquired so as to have certain effect in later cost aggregation. The joint weight filling algorithm is put forward for the problem of parallax discontinuity and mismatching in the final parallax refinement part. The experiment proves that the algorithm has low mismatching rate and high robustness on the edge area and the repeat texture and occlusion areas in comparison with other nonlocal stereo matching algorithms. The experiment result proves the effectiveness of the algorithm.

A computer-implemented depth estimation method based on non-parametric Census transform with adaptive window patterns and semi-global optimization. A modified cross-based cost aggregation technique adaptively creates the shape of the cross for each pixel distinctly. In addition, a depth refinement algorithm fills holes within the estimated depth map using the surrounding background depth pixels and sharpens the object boundaries by exerting a trilateral filter to the generated depth map. The trilateral filter uses the curvature of pixels as well as texture and depth information to sharpen the edges.

An image matching apparatus includes a bilateral filter that filters a left image and a right image to output a second left image and a second right image; a census cost calculation unit performing census transform on a window based on a first pixel of the second left image and a window based on a second pixel of the second right image to calculate a census cost corresponding to a pair of pixels of the first and second pixels; a support weight calculation unit obtaining support weights of the left and right images or the second left and second right images; a cost aggregation unit obtaining energy values of nodes corresponding to the pair of pixels of the first and second pixels using the census cost and the support weights; and a dynamic programming unit performing image matching using dynamic programming by the energy values of each node obtained.

The invention belongs to the field of stereo matching of images, and aims at disclosing a new local stereo matching framework so as to achieve the aims of rapidly and correctly obtaining stereo image disparity maps. The technical scheme of the method is a similarity measure function-based improved local stereo matching method which comprises the following steps of: constructing a new similarity measure function by combining a Census transform method according to space distance information among a color, a gradient and pixels of a to-be-matched image, and calculating a matching cost of each pixel; calculating the sum of the matching cost of each point, namely, carrying out cost aggregation, and in the cost aggregation process, obtaining an initial disparity map through a guide map filter method by adoption of a cross frame-based self-adaptive window; and carrying out image segmentation on the initial disparity map by utilizing a mean shift algorithm, and optimizing mistaken matching points in the disparity map through a surface interpolation method by combining a credibility distribution map of images, so as to obtain a final parallax result. The method is mainly applied to the stereo matching occasions of images.

A method and system for stereo correspondence. The method for stereo correspondence includes a matching cost computation step, a cost aggregation step, a disparity computation step, and a disparity optimization step. The matching cost computation step acquires a left disparity space image and a right disparity space image by using horizontal gradients and vertical gradients of intensities of all component channels of every pixel in a left image and a right image. Utilizing the invention, accurate disparity maps may be acquired quickly.

The invention relates to the field of computer vision and the technical field of image processing, in particular to a census transformation-based binocular stereo matching method. The method comprisesthe steps of initialization, matching cost calculation, cost aggregation, parallax calculation and parallax optimization. An existing matching method based on Census transform cannot form a disparitymap with higher precision, and particularly, more accurate disparity can be obtained in a shielding area and an edge. The binocular stereo matching method is improved on the basis of the prior art. The matching precision of the binocular stereo matching method is improved. A good foundation is laid for the subsequent three-dimensional reconstruction process.

The invention discloses a stereo matching method and system. The method comprises steps that a reference image and a target image are acquired, and a parallax area of an aggregation area is determined; a partitioning transformation algorithm is utilized to convert the reference image into an aggregation reference image and convert the target image into an aggregation target image; CENSUS characteristics of the images are respectively calculated to acquire a matching cost value of the aggregation reference image and the aggregation target image; area cost aggregation of each reference partitioning image and the target partitioning image at the corresponding identical partitioning position is respectively calculated; if the matching cost value is greater than area cost aggregation, the reference partitioning images and the target partitioning images in an area correspondence mode are re-segmented, and area cost aggregation after re-segmentation is calculated till the matching cost value is smaller than each area cost aggregation; a first cost aggregation parallax image of each pixel of each reference partitioning image and the corresponding target partitioning image is calculated. The method is advantaged in that a problem of false matching in the weak texture area can be effectively solved, and low computational complexity and good timeliness are realized.

A weak texture detection-based cross-scale cost aggregation stereo matching method of the present invention belongs to the computer vision field, in particular relates to a stereo matching method of the weak texture images, and comprises the following steps of inputting two color images which are separately a left image and a right image, utilizing the gradient information of the left image to carry out the weak texture detection and segmentation on a picture; calculating the matching cost according to the gradient information and the color information of the left image and the right image; taking the above weak texture detection and segmentation results as the references to carry out the Gaussian filter-based internal scale and cross-scale cost aggregation; adopting a winner taking strategy to calculate the parallax; adopting the left and right consistency detection and an adaptive weight-based method to refine the parallax, and outputting a parallax image. The method of the present invention achieves the technical purposes of improving the weak texture area matching correct rate under the premise of guaranteeing the texture area matching correct rate, and obtaining the better parallax image.

The invention relates to a stereo matching method based on a self-adaptive weight and local entropy, belonging to the technical field of binocular stereo vision. The method includes redefining the calculation of the window weight according to the gray similarity and the spatial similarity under a self-adaptive weight theory framework; introducing a data smoothing item into the cost function, and determining the penalty coefficient of the data smoothing item by applying the image local entropy; and performing parallax refinement by applying left-right consistency check and weighted median filtering on the initial parallax map obtained by cost aggregation to obtain a high-precision parallax image. The accurate parallax map can be obtained, and the parallax accuracy is improved.

The invention relates to a three-dimensional image matching method based on a convolutional neural network, which comprises the following steps: acquiring a plurality of three-dimensional matching image pairs and corresponding real disparity, and taking the three-dimensional matching image pairs and the corresponding real disparity as a data set; Constructing a convolutional neural network, and selecting a linear correction unit (RELU) function for activation; Training the convolutional neural network by adopting a back propagation algorithm, and determining a network error function and a learning rate; Through the calculation of the convolutional neural network, the network outputs a matching cost space diagram of the left and right image blocks; And performing matching cost aggregation,disparity selection and disparity refinement on the cost space graph, and selecting a pixel point with the minimum cost as a matching point to obtain a final disparity map. The image can be directly used as network input, the overall disparity map obtained through the convolutional neural network matching algorithm is relatively smooth, a relatively good matching effect can be achieved in a non-texture region and a depth value abrupt change region, and relatively good robustness is still achieved for an image pair with illumination change and incomplete correction.

The invention discloses a stereo matching algorithm for a weak texture area. Firstly, Gaussian sampling is performed on an input image to construct to-be-matched image pairs at different scales; and then, the canny algorithm is used for performing edge detection on an original image to detect the edge of the object and mark off a low texture area; a smaller cost aggregation window is set for the edge area; for the low texture area, the constrained consistency of the same pixel between different scales is emphasized; the improved sliding window method is used for calculating the image pixel pair matching cost at each scale; the feature point matching result is used for delimiting the parallax search range; the aggregate matching cost corresponding to each parallax value in the parallax range is calculated through a multi-scale cost aggregation function; and the parallax value corresponding to the minimum cost is selected as a pixel parallax value; and a dense parallax map is obtained bytraversing the original image pixel. The stereo matching algorithm for a weak texture area in the invention has high matching precision for the low texture area and can effectively reduce mismatching.

The invention discloses a stereoscopic vision matching method based on an adaptive feature window, and belongs to the technical field of image information processing. The stereoscopic vision matching method comprises the steps of associating feature selection with an intrinsic feature of an image, and training a feature selection classifier on the basis of a logistic regression method; applying the obtained trained feature selection classifier to stereoscopic matching in an actual scene, selecting a proper feature window, carrying out matching cost calculation by combining a color feature, carrying out matching cost aggregation on the basis of a partition tree method, and finally calculating a parallax result of the stereoscopic matching according to a winner retaining mode. According to the method, the stereoscopic matching is carried out by selecting the proper feature according to the actual application scene, so that the matching performance can be effectively improved, and the parallax calculation precision is improved.

The invention discloses a binocular vision matching cost aggregation optimization method which can be used in the field of computer vision stereo matching. The method comprises the following steps ofmatching left and right images by adopting a census algorithm, calculating the cost of the left image in a corresponding point parallax range in the right image, and obtaining a cost array in each point parallax range; calculating a cost variance in a disparity range of each point, carrying out cost aggregation, eliminating an error cost in combination with a mask scheme, and calculating an aggregation result of each point; and finally, analyzing the correlation of the data in the aggregation, splitting and isolating uncorrelated paths in the cost aggregation process and the data in the same path, and improving the aggregation real-time performance. The whole method can improve the final matching precision and real-time performance.

The invention proposes a stereo matching method based on graph cut. The method comprises the steps: extracting supporting points from left and right images through employing an MSERDoG operator; enabling the pixel gray values of the supporting points to serve as matching cost and fixed windows which are used as cost aggregation for the matching of the supporting points of the left and right images, and obtaining the parallax of the supporting points; calculating a DAISY descriptor operator of each pixel of the left and right images; enabling the parallax of the supporting points to serve as a symbol in graph cut, and enabling the DAISY descriptor operator of each pixel to serve a data item of an energy function in a graph cut algorithm, wherein the data item is used in the in a graph cut algorithm; finally solving a dense parallax graph through solving the minimum value of the energy function, and achieving stereo matching. The method can improve the matching efficiency under the condition of guaranteeing the imaging matching precision.

The invention discloses a stereo matching method and a stereo matching device, which can carry out stereo matching on a first image and a second image. The method comprises steps: the first image is used as a reference image, and cost images according to different parallax distances are acquired; cost aggregation is carried out on matching cost values of each pixel point in the first image in the case of different parallax distances; according to the matching cost values after cost aggregation, a first preset penalty coefficient and a second preset penalty coefficient, dynamic programming operation is carried out on the cost images corresponding to the different parallax distances in a preset direction respectively, an accumulated matching cost value of each pixel point in the case of different parallax distances in different preset directions is acquired, and according to the accumulated matching cost value, an actual parallax value of a corresponding pixel point is acquired; and refinement processing is carried out on the acquired actual parallax value of each pixel point, and a parallax estimation map for the first image and the second image is acquired. Thus, the stereo matching accuracy can be ensured, and the complexity during the stereo matching process can be reduced.

The invention discloses a stereoscopic matching method on the basis of variable-weight cost computation and S-census transformation. The stereoscopic matching method includes carrying out census transformation by the aid of average values of the minimum uniformity of star neighborhoods instead of central pixel gray values; utilizing variable-weight Hamming distances as matching costs and acquiring initial disparity by the aid of non-local cost aggregation processes; further processing the initial disparity by the aid of disparity refining processes on the basis of average value shifting and processing incredible disparity regions to obtain high-precision disparity maps. The stereoscopic matching method has the advantages that the shortcomings of the traditional census transformation are analyzed, the variable-weight cost computation and the S-census transformation are proposed, and accordingly the anti-interference capacity of algorithms can be effectively enhanced; the matching accuracy can be effectively improved by selecting and analyzing processes with the matching costs which are the variable-weight hamming distances, and the mismatching rate can be reduced; the initial disparity can be acquired by the aid of the non-local cost aggregation processes, and accordingly the discrimination of disparity can be improved; the shortcoming of limitation such as low local stereoscopic matching precision and weak anti-interference capacity can be overcome, and high-precision stereoscopic matching can be implemented.

The invention discloses a binocular stereo vision measuring method, including target detection: positioning a first target detection frame and acquiring the initial image position information of a target; binocular calibration: obtaining the relevant parameters of a binocular camera; binocular stereo matching: forming a secondary parallax space and the three-dimensional information of each point in the image; and target segmentation: finely segmenting the target to obtain a second target detection frame, and obtaining a distance between the target and the binocular camera or a protection target by using the three-dimensional information in the second target detection frame. The binocular stereo vision measuring method can improve the robustness of the binocular stereo matching, increases the cost aggregation efficiency in the process, and can quickly obtain high-quality parallax, thereby contributing to the rapid warning of invaders in a power grid and achieving automatic transmissionline intrusion early warning.

The invention relates to an improved matching cost aggregation stereo matching algorithm, which comprises the following steps: carrying out epipolar rectification on left and right two color images; selecting left and right images as reference images respectively, calculating matching cost through combination of image brightness information and gradient information, and for the left and right images, carrying out the following steps respectively: converting the images to logRGB color spaces respectively; defining a weight factor S and defining color weight and space distance weight; defining self-adaptive weight w (x, y) between pixel points x and y, calculating a disparity value for a disparity space through a WTA rule and obtaining an initial disparity map; and carrying out left and right consistency verification on the left and right initial disparity maps to remove unmatched points, and carrying out parallax filling through a background filling method to obtain a more accurate disparity map. The stereo matching algorithm has the advantages of high efficiency and high accuracy.

The present invention provides an edge-based local stereo matching method. After an initial disparity map is obtained; grid point sampling is performed on the disparity map; unsteady grid points are deleted, and stable grid points are preserved; Canny edge points are extracted from the image; stable edge points and the stable grid points are merged into a stable point set; and interpolation is performed on the other points according to disparity corresponding to the stable points, so that an accurate dense disparity map is obtained. According to the method provided by the invention, gradient is adopted as matching cost; a fixed window cost aggregation mode is adopted to calculate the matching cost; and the initial disparity map is obtained through left and right consistency detection. The method of the invention has the advantages of low complexity, high speed and high precision, and is suitable for occasions having high requirements for real-time performance.

The present invention proposes a full-automatic foreground and background separation method based on the binocular vision. The method comprises: in the initialization phase, using the local matching algorithm to generate a disparity map preliminarily, then using the initial disparity map to generate a trimap, and according to a color linear based hypothesis theoretical method proposed by Levin and other people, solving the initial opacity alfa; in the iterative optimization phase, integrating the opacity information of the left and right images into the cost aggregation function to enhance the disparity map, especially the boundary region; using the enhanced disparity map to provide a more reliable trimap for matting, and using the disparity gradient as a diffusion map to insert into the matting formula; and carrying out constant iteration in the entire optimization process until a satisfactory result is obtained. The experimental results show that the proposed method has good robustness, significantly reduces the error at the boundary of the disparity map and improves the accuracy of the image matting.

The embodiment of that invention relate to the technical field of image processing, and in particular, the invention relates to a power equipment stereo matching method based on global optimization. The method comprises: obtaining a first image and a second image; constructing initial matching cost function, calculating a matching generation value of the first image and the second image accordingto the initial matching cost function, building a minimum spanning tree, using the minimum spanning tree to aggregate the cost of matching generation value to obtain the aggregate value of the cost. The improved tree structure dynamic programming method is used to optimize the cost aggregation value globally and obtain the optimized generation value. The disparity maps of the first image and the second image are obtained by calculating and screening a plurality of new cost aggregation values according to the optimized generation value, and the dense disparity maps are obtained by calculating aplurality of new cost aggregation values according to the disparity maps and screening. The method can effectively improve the matching accuracy of the images under the scene of the electric power equipment.

As a key component of 4D products, a digital surface model (DSM) has a wide use range and use value in the field of national basic mapping. In order to accurately, stably and efficiently generate a digital surface model, the present invention discloses a dense DSM generation method based on object-side semi-global multi-view image matching. The method uses feature matching and triangulation network (TIN) interpolation to establish an initial DSM, introduces multi-view image matching of object-side geometric constraints through object-side space grid division, forms the DSM by matching cost aggregation of object-side space three-dimensional coordinates, and finally uses a series of data post-processing to remove the mismatched area and fill the default elevation area. After a lot of experiments, the method greatly improves the accuracy, reliability and efficiency of digital surface model matching.

The invention provides a high-efficiency focusing stereo matching method based on image segmentation. A reference image is segmented into different image blocks by utilizing an image segmentation method; Census transformation is simultaneously performed on the reference image and a to-be-matched image to obtain sequence values; dynamic rectification is performed on a to-be-matched pixel parallax scope by utilizing neighborhood pixel parallax; a matching cost function is employed, segmentation blocks are used as support windows, and a cost calculation is performed on the sequence values; and acost aggregation value in each segmentation block is calculated by the movement of windows in a horizontal direction, and a horizontal offset amount corresponding to a minimum cost aggregation value represents an optimal parallax value of a segmentation block. According to the invention, the matching precision of deep discontinuous regions is improved, the calculation efficiency is high, and robustness is good in different illumination environments.

A method and system for stereo correspondence. The method for stereo correspondence includes a matching cost computation step, a cost aggregation step, a disparity computation step, and a disparity optimization step. The matching cost computation step acquires a left disparity space image and a right disparity space image by using horizontal gradients and vertical gradients of intensities of all component channels of every pixel in a left image and a right image. Utilizing the invention, accurate disparity maps may be acquired quickly.

An embodiment of the application discloses a binocular stereoscopic vision based distance measuring method and a distance measuring system. The method comprises the following steps of obtaining an original reference image and an original target image; correcting the original reference image and the original target image according to a camera parameter to obtain corresponding corrected reference image and corrected target image; carrying out matching cost calculation on the corrected reference image and the corrected target image to obtain a matching cost between each pixel point in the corrected reference image and a corresponding point in the corrected target image, wherein the matching cost comprises a characteristic matching cost and a divergence matching cost; carrying out cost aggregation on the matching cost to obtain a corresponding cumulative cost; and calculating a parallax value corresponding to each pixel point in the corrected reference image according to the cumulative cost, so as to complete distance measurement. The distance measuring method has quite good distance measuring accuracy no matter being used for carrying out short-distance distance measurement or long-distance distance measurement; and an effective measurement range and the distance measuring accuracy of the distance measuring system applying the method are improved.

An apparatus for generating depth information of an example may include an image downsizer configured to downsize an input stereo image by a predetermined magnification to generate a downsized stereo image; a rawcost calculator configured to calculate rawcosts for the downsized stereo image; a mapper configured to map the rawcosts to the input stereo image; an interpolator configured to perform interpolation and cost aggregation for the rawcosts; and a depth information generator configured to calculate an optimal disparity for each pixel of the input stereo image by referring to the rawcosts.