Changes in illumination conditions have a significant effect on the performance of robot vision tasks such as object recognition. One way to handle varying illumination is to apply illumination normalization as a pre-...
详细信息
Changes in illumination conditions have a significant effect on the performance of robot vision tasks such as object recognition. One way to handle varying illumination is to apply illumination normalization as a pre-processing step. We compare several illumination normalization methods for the task of robot soccer. Our results show improved recognition performance under changes in intensity and brightness.
A new approach for object extraction from high-resolution satellite images is presented in this paper. The new approach integrates image fusion, multi-spectral classification, feature extraction and feature segmentati...
详细信息
image classification plays an important part in the fields of remotesensing, image analysis and patternrecognition. image classification can be done using conventional methods. But conventional methods lead to miscl...
ISBN:
(纸本)9781450304498
image classification plays an important part in the fields of remotesensing, image analysis and patternrecognition. image classification can be done using conventional methods. But conventional methods lead to misclassification due to strictly convex boundaries. Textural features are included for better classification but are inconvenient for conventional methods. The proposed system uses textural feature based image classification using neural network. Textural features are extracted using Gray level co-occurrence matrix and artificial neural network is developed for the classification of images into different classes. Neural network is trained by supervised learning using standard back propagation algorithm for the classification of images.
Combining spectral and spatial information can improve land use classification of high-resolution data. However,the use of spatial information always focus on objects' spatial pattern,whereas not pay enough attent...
详细信息
Combining spectral and spatial information can improve land use classification of high-resolution data. However,the use of spatial information always focus on objects' spatial pattern,whereas not pay enough attention to spatial relationship,which is more convenient and effective in remotesensing *** letter proposes a spectral-spatial information method,which aims to exploit objects' spatial relationships in high resolution imagery,and then integrate it with spectral information in remotesensing *** experiment on urban mapping based on spectral-spatial information using Quickbird imagery,and compare its result with supervised classification methods like maximum likelihood classification,and support vector machine (SVM) *** results show that the proposed method yield better performance than the others in both precision and rationality.
Fingerprint matching is an important issue in automatic fingerprint identification systems. There are difficulties about fingerprint matching based on neighborhood. One is the size of the neighborhood can not be deter...
详细信息
Fingerprint matching is an important issue in automatic fingerprint identification systems. There are difficulties about fingerprint matching based on neighborhood. One is the size of the neighborhood can not be determined readily, the other is the feature in the neighborhood can be affected by the noise. To deal with these problem, we developed a novel algorithm for fingerprint matching based on local structures to efficiently extract neighboring minutiae features. Neighboring features present the information of peripheral minutiae which directly connect with the central minutiae on topology. We use one feature vector to present neighboring features from different samples. The samples considered as the same class can make the proposed algorithm robust to rotation and translation of fingerprint images. The experiments are conducted on FVC2002, and the results illustrate the effectiveness of the proposed algorithm.
N-FINDR is a widely used endmember extraction algorithm in hyperspectral imagery. Nevertheless, its computational complexity is high. Plaza's parallel implementation of N FINDR, namely, P-FINDR, demonstrates an ex...
详细信息
N-FINDR is a widely used endmember extraction algorithm in hyperspectral imagery. Nevertheless, its computational complexity is high. Plaza's parallel implementation of N FINDR, namely, P-FINDR, demonstrates an excellent way to improve the computing performance of N-FINDR by incorporating with parallel computing technique. In this paper, three parallel implementation patterns, i.e., synchronous, asynchronous and grouping asynchronous pattern, are presented. Two versions of N-FINDR, i.e. iterative N-FINDR and successive N-FINDR, are considered to be implemented in parallel by these patterns respectively. Thus obtains six parallel N-FINDR algorithms. In experiment, both solution quality and parallel performance of these algorithms are compared and suitable patterns for parallel implementation of N-FINDR are obtained.
This study presents a supervised subspace learning classification method which can be applied directly to the original set of spectral bands of hyperspectral data for land cover classification purpose. The CLAss-Featu...
详细信息
This study presents a supervised subspace learning classification method which can be applied directly to the original set of spectral bands of hyperspectral data for land cover classification purpose. The CLAss-Featuring Information Compression (CLAFIC) method is used to generate the appropriate feature subspace for each class on the training data set by Karhunen-Loeve transform (also known as the principal component analysis). Then, using the iterative learning technology of averaged learning subspace methods (ALSM) to rotate the subspaces slowly for optimizes the subspaces to get better classification accuracy. We carried out experiments with 68 spectral bands Compact Airborne Spectrographic imager-3 (CASI-3) data set. Experimental results show that Subspace method is a valid and effective alternative to other patternrecognition approaches for the mapping grass species and monitoring grass health using hyperspectral remotesensing data. Moreover, it is worth noting that the ALSMs are easily applied (i.e. they only request to set two parameters and can be directly applied to hyperspectral data) and they can entirely identify the training samples in a finite number of steps.
暂无评论