中国科学院机构知识库网格系统: 检索

Federated Data Quality Assessment Approach: Robust Learning With Mixed Label Noise 期刊论文 OAI收割
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2023, 页码: 15
作者: Zeng, Bixiao; Yang, Xiaodong; Chen, Yiqiang

|

收藏 | 浏览/下载：30/0 | 提交时间：2023/12/04

Noise measurement Servers Task analysis Adaptation models Data models Data integrity Computers Data quality assessment federated learning (FL) noise-robust algorithm

A shape context based Hausdorff similarity measure in image matching 会议论文 OAI收割
5th International Symposium on Photoelectronic Detection and Imaging (ISPDI) - Infrared Imaging and Applications, Beijing, June 25-27, 2013
作者: Ma TL(马天磊); Liu YP(刘云鹏); Shi ZL(史泽林); Yin J(尹健)

收藏 | 浏览/下载：35/0 | 提交时间：2013/12/26

The traditional Hausdorff measure, which uses Euclidean distance metric (L2 norm) to define the distance between coordinates of any two points, has poor performance in the presence of the rotation and scale change although it is robust to the noise and occlusion. To address the problem, we define a novel similarity function including two parts in this paper. The first part is Hausdorff distance between shapes which is calculated by exploiting shape context that is rotation and scale invariant as the distance metric. The second part is the cost of matching between centroids. Unlike the traditional method, we use the centroid as reference point to obtain its shape context that embodies global information of the shape. Experiment results demonstrate that the function value between shapes is rotation and scale invariant and the matching accuracy of our algorithm is higher than that of previously proposed algorithm on the MEPG-7 database.

A fast target recognition algorithm based on MSA and MSR (EI CONFERENCE) 会议论文 OAI收割
2012 International Conference on Industrial Control and Electronics Engineering, ICICEE 2012, August 23, 2012 - August 25, 2012, Xi'an, China
作者: Wang Y.; Liu G.; Wang Y.; Wang Y.; Wang Y.

收藏 | 浏览/下载：27/0 | 提交时间：2013/03/25

This paper presents a new fast target recognition algorithm the proposed method is based on Multi-scale Auto convolution(MSA) and Multi-scale Retinex(MSR). As shown by the comparison with original MSA it appears that this new technique solves the problem that MSA algorithm is sensitive to illumination and the computational load is significantly reduced to 1/8th of that of the original MSA algorithm it is also robust to affine transform light projective transform noise thin fog occlusion and illumination change. the performed experiments show that it has fast searching speed and can accurately recognize and locate target in real scenes. 2012 IEEE.

Rotation and scaling invariant feature lines for image matching (EI CONFERENCE) 会议论文 OAI收割
2011 International Conference on Mechatronic Science, Electric Engineering and Computer, MEC 2011, August 19, 2011 - August 22, 2011, Jilin, China
作者: Wang Y.; Wang Y.; Wang Y.; Wang Y.; Wang Y.

收藏 | 浏览/下载：24/0 | 提交时间：2013/03/25

Image matching has been one of the most fundamental issues computer vision over the decades. In this paper we propose a novel method based on making use of feature lines in order to achieve more robust image matching. The feature lines have the properties of rotation and scaling invariance coined RIFLT(Rotation invariant feature line transform). Experimental results demonstrate the effectiveness and efficiency of the proposed method. Compare with the famous powerful algorithm Scale Invariant Feature Transform(SIFT) the proposed method is more insensitive to noise. And for certain sequence of images which contain clear lines the proposed method is more efficiency. Using the feature lines obtained by our method it is possible to matching two scene images with different rotation angles scale and light distort. 2011 IEEE.

Extracting sea-sky-line based on improved local complexity (EI CONFERENCE) 会议论文 OAI收割
2010 International Conference on Computer, Mechatronics, Control and Electronic Engineering, CMCE 2010, August 24, 2010 - August 26, 2010, Changchun, China
作者: Zhang Y.-F.

收藏 | 浏览/下载：20/0 | 提交时间：2013/03/25

Sea-sky-line extraction under complicated sea-sky background is an important aspect of long-range target tracking research. An algorithm based on improved local complexity is proposed according to the feature that the sky and the sea usually show up at different gray levels in sea-sky background images. Median filter is applied first to remove peak noise and then the point set of sea-sky-line region which has the greatest change in image is obtained by calculating the local complexity of image and selecting a segmentation threshold. Finally Hough transform is used to extract the sea-sky-line. The experiment results indicate that this method can extract sea-sky-line under simple and complicated sea-sky backgrounds which is robust and can improve noise immunity of the algorithm. 2010 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.

Detection of low contrast targets based on lifting scheme wavelet transform (EI CONFERENCE) 会议论文 OAI收割
2009 IEEE International Conference on Mechatronics and Automation, ICMA 2009, August 9, 2009 - August 12, 2009, Changchun, China
作者: Chen X.; Wang Y.; Wang Y.; Wang Y.; Wang Y.

收藏 | 浏览/下载：15/0 | 提交时间：2013/03/25

This paper present a fast algorithm for detection of low contrast objects by using wavelet filters based on lifting scheme. The advantage is robust to noise. According Swelden's lifting wavelet filters are biorthogonal wavelet filters containing free parameters. We use reference image of targets to train the lifting terms so that the learnt wavelet filters have the features of targets. Then applying such filters to the images including targets taken from camera system. We can detect the locations where the high frequency components are almost the same as those of the target image. 2009 IEEE.

An improved two-dimensional entropy method for star trail tracing in deep sky (EI CONFERENCE) 会议论文 OAI收割
ICO20: Optical Information Processing, August 21, 2005 - August 26, 2005, Changchun, China
作者: Wang Y.-J.; Yao Z.-J.

收藏 | 浏览/下载：29/0 | 提交时间：2013/03/25

The trace of star trail is an important component of deep sky detection. The stars are low contrast targets and their self-rotation will make their brightness change in cycle. Above all the trail trace is vulnerable to the block and disturbance of other stars. Traditional one-dimensional maximum entropy thresholding algorithm is vulnerable to the noise and the calculation of two-dimensional entropy methods is too large and takes too much time. This paper proposes an improved two-dimensional entropy threshold algorithm. We use recursion iteration method to eliminate the redundancy calculation and reduce the size of two-dimensional histogram based on the deep sky stars characteristic such as low contrast fuzziness and the centralized histogram. We also combine our algorithm with the space trail trace model to forecast the star trace. Experiments results show when the star are blocked or they turn dark the method still can well extrapolate the star trace. Our method improves the capability of trailing the ebb and small star and increases the precision of tracing. It is also robust to the noise so there is a good application foreground for the method.

A novel starting-point-independent wavelet coefficient shape matching (EI CONFERENCE) 会议论文 OAI收割
ICO20: Optical Information Processing, August 21, 2005 - August 26, 2005, Changchun, China
Hu S.; Zhu M.; Wu C.; Song H.-J.

收藏 | 浏览/下载：17/0 | 提交时间：2013/03/25

In many computer vision tasks in order to improve the accuracy and robustness to the noise wavelet analysis is preferred for the natural multi-resolution property. However the wavelet representation suffers from the dependency of the starting point of the sampled contour. For overcoming the problem that the wavelet representation depends on the starting point of the sampled contour the Zernike moments are introduced and a novel Starting-Point-lndependent wavelet coefficient shape matching algorithm is presented. The proposed matching algorithm firstly gains the object contours and give the translation and scale invariant object shape representation. The object shape representation is converted to the dyadic wavelet representation by the wavelet transform. And then calculate the Zernike moments of wavelet representation in different scales. With respect to property of rotation invariant of Zernike moments consider the Zernike moments as the feature vector to calculate the dissimilarity between the object and template image which overcoming the problem of dependency of starting point. The experimental results have proved the proposed algorithm to be efficient precise and robust.

Detection and tracking of low contrast targets based on integertype lifting wavelet transform (EI CONFERENCE) 会议论文 OAI收割
ICO20: Remote Sensing and Infrared Devices and Systems, August 21, 2005 - August 26, 2005, Changchun, China
作者: Wang L.; Wang L.; Wang Y.; Wang Y.; Wang Y.

收藏 | 浏览/下载：17/0 | 提交时间：2013/03/25

This paper presents a method for detecting and tracking of low contrast targets. The new method uses an integer-type lifting wavelet transform and the proposed method doesn't extract patterns similar to a template but finds parts having the same feature in the targets. We utilize one of integer-type lifting wavelet transforms that contains rounding-off arithmetic for mapping integers to integers. The lifting term contains parameters that are learned by using standard training images of targets. We assume that the targets include many high frequency components. In order to obtain the features of the targets the lifting parameters are determined by a condition that high frequency components are vanished in wavelet transform. But the condition cannot be determined by the parameters wholly. So we put an additional condition of minimizing the squared sum of the lifting parameters. The advantage of using integer-type wavelet transform is simple and robust to noise. Simulation illustrated the approach can detect and track the moving targets in dim background. We would test our algorithm in the TV tracking system.