中国科学院机构知识库网格系统: 检索

浏览/检索结果: 共6条，第1-6条

帮助

条数/页：排序方式：
	Template Matching Based on Geometric Invariance in Deep Neural Network 期刊论文 OAI收割 IEEE ACCESS, 2019, 卷号: 7, 页码: 82174-82182 作者: Cao, Yaming; Yang, Zhen; Wang, Haijiao; Peng, Xiaodong; Gao, Chen \| 收藏 \| 浏览/下载：74/0 \| 提交时间：2019/08/26 Deep neural network geometric invariant interpretability template match adversarial example
	Precise motion compensation based on weighted sub-pixel image matching (EI CONFERENCE) 会议论文 OAI收割 International Symposium on Photoelectronic Detection and Imaging 2011: Advances in Imaging Detectors and Applications, May 24, 2011 - May 26, 2011, Beijing, China Liang H. G. 收藏 \| 浏览/下载：23/0 \| 提交时间：2013/03/25 This paper proposed a sub-pixel image correlation algorithm that can get more Precise result its principle is apply the distribute of relativity peak to get weighted multi-pixel comprehensive of location. Image correlation be as to calculates the greyscale relativity of image template and matching image the relativity of correspond location where match best with template will be most high and in its neighbour range the relativity will be still keep high too. We used these pixel in this local area of calculated match point to get sub-pixel accuracy the relativity of every pixel be used as its weight for participate the sub-pixel calculation. The sub-pixel location is more accuracy than the integer one we applied this method to perform background compensation in processing the target detecting for video image sequence. At the end of this paper some experiment data be proposed it proved this sub-pixel image correlation can obtain better result. 2011 SPIE.
	A matching algorithm on statistical properties of Harris corner (EI CONFERENCE) 会议论文 OAI收割 2011 International Conference on Information and Automation, ICIA 2011, June 6, 2011 - June 8, 2011, Shenzhen, China 作者: He B. 收藏 \| 浏览/下载：21/0 \| 提交时间：2013/03/25 The fundamental goal of target recognition and video tracking is to match target template with source image. Most matching methods are based on image intensity or multi-feature points. And the latter method is more popular for its high accuracy and small calculation. Image Registration Based on Feature Points focus on effective feature extraction of image points and paradigm. Harris corner in the image rotation gray noise and viewpoint change conditions has an ideal match results is more recent application of one feature point. This paper extract the Harris corner deviation and covariance firstly experiments show that the two features exclusive then applied them to image registration for the first time. A set of actual images have shown this proposed method not only overcomes the complicated background gray uneven distribution problems but also pan and zoom the image has a good resistance. 2011 IEEE.
	An image matching algorithm based on sub-block coding (EI CONFERENCE) 会议论文 OAI收割 2nd International Workshop on Computer Science and Engineering, WCSE 2009, October 28, 2009 - October 30, 2009, Qingdao, China 作者: Li S. 收藏 \| 浏览/下载：29/0 \| 提交时间：2013/03/25 In order to improve the speed of matching algorithm and simplify the processing of existing sub-block coding matching a new template matching method combined local gray value encoding matching and phase correlation is proposed. Matching process is divided into rough matching and fine matching. Rough matching divides the image into certain size blocks called R-block sums the gray value of each R-block pixel encodes the R-block according to the gray value distribution of R-block with the adjacent R-block and matches by step between the template and each search sub-image. Then fine matching results are obtained using phase correlation according to initial match parameters. The time complexity of the proposed method is (M2) .The new algorithm is faster than traditional algorithm by two orders of magnitude and the speed has improved twice compared with existing sub-block coding method. Experiments demonstrate that the new algorithm is robust to the linear transformation of pixel grey value and image noise and it also has the stability of small-angle rotation. 2009 IEEE.
	Real-time matching algorithm of navigation image based on corner detection (EI CONFERENCE) 会议论文 OAI收割 International Symposium on Photoelectronic Detection and Imaging 2009: Advances in Imaging Detectors and Applications, June 17, 2009 - June 19, 2009, Beijing, China 作者: Zhang T. 收藏 \| 浏览/下载：43/0 \| 提交时间：2013/03/25 In order to meet requirement of real-time and high accuracy in image matching aided navigation SSDA algorithm is used to match remote sensing image and template image coarsely a fast and effective algorithm of remote sensing image matching based on corner detection is put forward. With the combination of rough and fine match when the matching result is bigger than one to count absolute value sum of energy difference of characteristic point energy to realize fine match of remote sensing image and template image to locate the position of template image in remote sensing image accurately. Simulation experiment proves that the matching of a remote sensing image resolution of 10181530 and a template image resolution of 15090 can be fulfilled within 2.392 second wavelet transform is used to acquire low frequency component to realize image compression to decrease calculation work and increase matching speed. Harris corner detection algorithm is used to detect corner of remote sensing image and template image and energy of every corner is calculated the algorithm is robust and effective real time image navigation can be achieved. 2009 SPIE.
	Affine object recognition and affine parameters estimation based on covariant matrix (EI CONFERENCE) 会议论文 OAI收割 2008 International Symposium on Information Science and Engineering, ISISE 2008, December 20, 2008 - December 22, 2008, Shanghai, China 作者: Wang Y.; Wang Y.; Wang Y.; Wang Y.; Wang Y. 收藏 \| 浏览/下载：121/0 \| 提交时间：2013/03/25 A new method of affine object recognition and affine parameters estimation is presented. For a real-time image and a group of templates in addition firstly on the basis of correct recognition we segment the object regions in them and compute their covariant matrices. Secondly it can estimate affine parameters exactly normalize the ellipse regions defined by covariant matrices to circle regions to get rotational invariants and the estimated error is within 3%. 2008 IEEE. and compute the similarity function value between rotational invariants of real-time image and every template respectively. Then compare the values with threshold set in advance if more than one value is larger than threshold take the corresponding templates as candidates and compute affine matrix between real-time image and every candidate. Finally transform the realtime image with every affine matrix and match the result with corresponding candidate by classical matching methods. Experimental results show that the presented method is robust to illumination with low computational complexity and it can realize recognition of different affine objects