检索结果-内蒙古大学图书馆

Conference on computer vision and pattern recognition (CVPR)

作者： Dirk Padfield One Research Circle GE Global Research Center Niskayuna NY USA

With the explosion in the usage of mobile devices and other smart electronics, embedded devices are becoming ubiquitous. Most such embedded architectures utilize fixed-point rather than floating-point computation to meet power, heat, and speed requirements leading to the need for integer-based processing algorithms. Operations involving Gaussian kernels are common to such algorithms, but the standard methods of constructing such kernels result in approximations and lack a property that enables efficient bitwise shift operations. To overcome these limitations, we present how to precisely combine the power of integer arithmetic and bitwise shifts with intrinsically real valued Gaussian kernels. We prove mathematically that there exist a set of what we call “magic sigmas” for which the integer kernels exactly represent the Gaussian function whose values are all powers-of-two, and we discovered that the maximum sigma that leads to such properties is about 0.85. We also designed a simple and precise algorithm for designing kernels composed exclusively of integers given any arbitrary sigma and show how this can be exploited for Gaussian filter design. Considering the ubiquity of Gaussian filtering and the need for integer computation for increasing numbers of embedded devices, this is an important result for both theoretical and practical purposes.

关键词： Kernel Equations Approximation methods Hardware Algorithm design and analysis Mathematical model computer vision

来源：评论

学校读者我要写书评

暂无评论

computer vision based date fruit grading system: Design and implementation

引用

JOURNAL OF KING SAUD UNIVERSITY-computer AND INFORMATION SCIENCES 2011年第1期23卷 29-36页

作者： Al Ohali, Yousef King Saud Univ Coll Comp & Informat Sci Dept Comp Sci Riyadh Saudi Arabia

The Kingdom of Saudi Arabia is the world's largest producer of date fruit. It produces almost 400 date varieties in bulk. During the harvesting season the date grading and sorting pose problems for date growers. Since it is a labor intensive and time consuming process, it delays the post harvesting operations which costs them dearly. The date grading and sorting is a repetitive process. In practice, it is carried out by humans manually through visual inspection. The manual inspection poses further problems in maintaining consistency in grading and uniformity in sorting. To speed up the process as well as maintain the consistency and uniformity we have designed and implemented a prototypical computer vision based date grading and sorting system. We have defined a set of external quality features. The system uses RGB images of the date fruits. From these images, it automatically extracts the aforementioned external date quality features. Based on the extracted features it classifies dates into three quality categories (grades 1, 2 and 3) defined by experts. We have studied the performance of a back propagation neural network classifier and tested the accuracy of the system on preselected date samples. The test results show that the system can sort 80% dates accurately. (C) 2010 King Saud University. Production and hosting by Elsevier B.V. All rights reserved.

关键词： Fruit classification pattern recognition Image processing Neural networks Automation in agriculture

来源：评论

学校读者我要写书评

暂无评论

An Empirical Study of Visual Features for Part Based Model

An Empirical Study of Visual Features for Part Based Model

引用

第一届亚洲模式识别会议

作者： Junge Zhang Yinan Yu Shuai Zheng Kaiqi Hua National Laboratory of Pattern Recognition Institute of Automation Chinese Academy of Sciences National Laboratory of Pattern RecognitionInstitute of Automation Chinese Academy of Sciences

ISBN: (纸本)9781457701221

Object detection is a fundamental task in computer vision. Deformable part based model has achieved great success in the past several years, demonstrating very promising performance. Many papers emerge on part based model such as structure learning, learning more discriminative features. To help researchers better understand the existing visual features’ potential for part based object detection and promote the deep research into part based object representation, we propose an evaluation framework to compare various visual features’ performance for part based model. The evaluation is conducted on challenging PASCAL VOC2007 dataset which is widely recognized as a benchmark database. We adopt Average Precision (AP) score to measure each detector’s performance. Finally, the full evaluation results are present and discussed.

关键词： Histograms Object detection Image color analysis Visualization computer vision Deformable models Color

来源：评论

学校读者我要写书评

暂无评论

A robust structure extraction method based on image edge

A robust structure extraction method based on image edge

引用

Congress on Image and Signal Processing, CISP

作者： Guang Yang Lei Hu Qichao Lu Feng Gao Beijing Key Laboratory of Digital Media School of Computer Science and Engineering Beihang University 100191 China

In this paper, we propose a novel approach with desirable universality and expansibility for structural feature extraction. Structural feature is a high-level 2D feature that delivers information including image components and spatial relations of them. In the proposed approach, we first trace edges to line segments, fit line segments to longer lines and extract parallel pairs. Then, we extract contours of image components by grouping lines and parallel pairs. Lastly, we reveal the salient spatial relations of components by theorems and judging rules. The theorems, which are concluded from the step of components extraction, facilitate the following analytical procedure. Experimental results tested on 1960 real images demonstrate that our approach can extract integrated contours of image components and analyze the spatial relations between them in desirable time consuming, and the approach is suitable for many application fields.

关键词： Feature extraction Image edge detection Image segmentation Complexity theory computer vision pattern recognition Target recognition

来源：评论

学校读者我要写书评

暂无评论

Handwritten Street Name recognition for Indian Postal Automation

Handwritten Street Name Recognition for Indian Postal Automa...

引用

International Conference on Document Analysis and recognition

作者： Umapada Pal Ramit Kumar Roy Fumitaka Kimura Computer Vision and Pattern Recognition Unit Indian Statistical Institute Kolkata India Saint Xavier's College Kolkata India Graduate School of Engineering Mie University TSU Japan

Although for postal automation there are many pieces of work towards street name recognition on non-Indian languages, to the best of our knowledge there is no work on street name recognition on Indian languages. In this paper we proposed a scheme for recognition of Indian street name written in Bangla script. Because of the writing style of different individuals some of the characters in a street name may touch with its neighboring characters. Accurate segmentation of such touching into individual characters is a difficult task. To avoid such segmentation, here we consider a street name string as word and the street name recognition problem is treated as lexicon driven word recognition. Some of the street names may contain two or more words and we have concatenated these words to have a single word. In the proposed method, at first, street names are binarized and pre-segmented into possible primitive components (individual characters or its parts) analyzing their cavity portions. Pre-segmented components of a street name are then merged into possible characters to get the best street name. Dynamic programming (DP) is applied for the merging using total likelihood of characters as the objective function. To compute the likelihood of a character, modified quadratic discriminant function (MQDF) is used. Our proposed system shows 99.03% reliability with 18.80% rejection, and 0.79% error rates when tested on 4450 handwritten Bangla street name samples.

关键词： Handwriting recognition Accuracy Cavity resonators Automation Feature extraction Dynamic programming Image segmentation

来源：评论

学校读者我要写书评

暂无评论

A recognition System for 3D Embossed Digits on Non-Smooth Metallic Surface

A Recognition System for 3D Embossed Digits on Non-Smooth Me...

引用

International Conference on computer Communication and Management（ICCCM 2011）

作者： Taweepol Suesut Suphan Gulpanich Kitti Tirasesth Department of Instrumentation and Control Engineering Faculty of Engineering King Mongkut's Institute of Technology Ladkrabang Bangkok 10520 THAILAND Department of Instrumentation and Control Engineering Faculty of Engineering King Mongkut’s Institute of Technology Ladkrabang Bangkok 10520 THAILAND

This paper presents a new solution to the automatic localization and recognition of natural digits embossed on structured and rough surfaces. A new concept of combining periodic (e.g. FFT, DCT) with polynomial moments (e.g. Tchebichef moments) is shown that the new method can eliminate some of the problems associated with the Gibbs phenomena. QR decomposition is used to obtain a unitary basis, minimizing the numerical effort when modeling surfaces. The combined moment method is used to generate a smoothed global model of the surface structure. Following segmentation of the individual digits, pattern recognition based on moment function such as Hu’s moment invariants, Zernike moments, and Orthogonal Fourier Mellin Moments was used to recognize the stamped number after analyzing 3D surface data of the steel block. A prototype system of the laser-scanning instrument was implemented and test measurements were performed in a production line. The results showed the correct functionality of the method.

关键词： Polynomial moments, 3D surface modeling, OCR, machine vision, Gibbs error

来源：评论

学校读者我要写书评

暂无评论

Detect saliency to understand a photo

Detect saliency to understand a photo

引用

International Conference on Multimedia Technology

作者： Lin Cong Ruo-Feng Tong Da-Yong Bao College of Computer Science and Technology University of Zhejiang Hangzhou Zhejiang China

Visual saliency is an extremely hot topic in image and video processing. Our paper addresses a novel method for saliency detection, which conveys the photographers' idea of a scene. It consists of analyzing the photo elements like color, depth-of-field (DOF) and composition. We use the color spatial distribution feature to calculate the initial saliency. A simple but efficient classifier is trained to discriminate the low DOF image. For this kind of images we find the focal area to refine the initial saliency. The saliency map will be further adjusted by an image composition analysis step to generate the final output. We compare our framework with state-of-the-art algorithms. Experimental results show that, our method out-performs the previous methods in many cases, especially in the low DOF situation.

关键词： Image color analysis Visualization Image edge detection Conferences computer vision pattern recognition Computational modeling

来源：评论

学校读者我要写书评

暂无评论

Robust upper body detection in unconstrained posture

Robust upper body detection in unconstrained posture

引用

International Conference on Multimedia Technology

作者： Diping Li Beiji Zou Department of Math & Information Technology Hanshan Normal University Chaozhou China School of Information Science and Engineering Central South University Changsha China

ISBN: (纸本)9781612847719

This paper presents an efficient approach to detecting upper body in unconstrained posture based on Shape Context and Histograms of Orient Gradients. The method contains two steps. Shape context matching is used to get the candidates of upper bodies in the images at first, and then validation of each candidate is performed by using the Histograms of Orient Gradients. The method has two advantages, comparing with existing body detection methods. One is that it can detect upper bodies in all kinds of postures other than upright standing posture, and the other is that it significantly speeds up the computation.

关键词： Shape Context Humans Histograms computer vision Support vector machines pattern recognition

来源：评论

学校读者我要写书评

暂无评论

Neural network based motion segmentation for accelerometer applications

Neural network based motion segmentation for accelerometer a...

引用

IEEE Symposium on 3D User Interfaces (3DUI)

作者： Jong Gwan Lim Sang-Youn Kim Dong-Soo Kwon KAIST South Korea Korea Advanced Institute of Science and Technology Korea University of Technology and Education South Korea

Of several research issues related to motion interaction using inertia measurement units, faster motion segmentation without accuracy loss has recently been raised. Instead of using excessive filtering that produces time delay or tricky use of multiple thresholds that cause difficulty in parameter optimization, this poster demonstrates that time series prediction using neural networks significantly decreases time delay and guarantees rigid motion segmentation by detecting end points in accelerometer signals. According to a general pattern recognition procedure, feature selection is made by a filtering method and the optimal structure is determined by cross validation. Radial basis function networks and Multi-Layer Perceptrons (MLPs) are tested and the results are compared with the conventional methods to evaluate accuracy and time delay in a handwriting case in 3D space. This study confirms that MLP shows the best accuracy and shortens the time delay by 1/4~1/3 compared to the conventional methods.

关键词： Delay effects Acceleration Accelerometers Motion segmentation computer vision pattern recognition Accuracy

来源：评论

学校读者我要写书评

暂无评论

A rapid, one-hour method for enumeration of Escherichia coli based on computer vision technology and artificial neural network

A rapid, one-hour method for enumeration of Escherichia coli...

引用

2011 International Conference on computers, Communications, Control and Automation

作者： Yun Ding Zhihui Qian Editorial department of Journal of Bionic Engineering Jilin University College of Biological and Agricultural Engineering Jilin University

In order to reduce the detection time of Escherichia coli(E. coli) on fresh, minimally processed vegetables, a new close to real-time method based on artificial intelligence technology is presented. The ability of three-layer multilayer perceptron neural network(MLP) models to predict E. coli contaminated condition using feature parameters extracted from E. coli images was compared with multivariate statistical model(discriminant analysis, DA) with respect to the identification accuracy of E. coli cells. The performance of MLP was 10% better than that from the DA. Comparisons of the counts of E. coli by rapid method and by traditional method(multiple-tube fermentation technique) closed correlated(R2=0.9867). Besides the high accuracy, the total detection time was within 1 h and the detection limit was 10 cells /ml. This is a clear improvement compared with the 6 days taken to obtain results from the traditional method.

关键词： Multilayer perceptron neural network computer vision technology Artificial intelligence detection system pattern recognition

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：