Fast Feature Extraction Research Articles

Robust gait-based gender classification using depth cameras

This article presents a new approach for gait-based gender recognition using depth cameras, that can run in real time. The main contribution of this study is a new fast feature extraction strategy that uses the 3D point cloud obtained from the frames in a gait cycle. For each frame, these points are aligned according to their centroid and grouped. After that, they are projected into their PCA plane, obtaining a representation of the cycle particularly robust against view changes. Then, final discriminative features are computed by first making a histogram of the projected points and then using linear discriminant analysis. To test the method we have used the DGait database, which is currently the only publicly available database for gait analysis that includes depth information. We have performed experiments on manually labeled cycles and over whole video sequences, and the results show that our method improves the accuracy significantly, compared with state-of-the-art systems which do not use depth information. Furthermore, our approach is insensitive to illumination changes, given that it discards the RGB information. That makes the method especially suitable for real applications, as illustrated in the last part of the experiments section.

A novel approach for Face Recognition using Local Binary Pattern

This paper presents Local Binary pattern (LBP) as an approach for face recognition with the use of some global features also. Face recognition has received quite a lot of attention from researchers in biometrics, pattern recognition, and computer vision communities. The idea behind using the LBP features is that the face images can be seen as composition of micro-patterns which are invariant with respect to monotonic grey scale transformations and robust to factors like ageing. Combining these micro-patterns, a global description of the face image is obtained. Efficiency and the simplicity of the proposed method allows for very fast feature extraction giving better accuracy than the other algorithms. The proposed method is tested and evaluated on ORL datasets combined with other university dataset to give a good recognition rate and 89% classification accuracy using LBP only and 98% when global features are combined with LBP. The method is also tested for real images to give good accuracy and recognition rate. The experimental results show that the method is valid and feasible.

Automatic Registration Based on Covariance Matrix Eigenvector Direction of Feature Point

A number of range images taken from different views to be merged to construct an entire 3D model are a common problem which is called registration in RE. The enormous computational cost is required, because this process is usually repeated on searching optimal rigid body coordinate transformation parameters and evaluated in a statistical error distance between two data. This paper proposes a new registration method based on the covariance matrix eigenvector direction of feature point. First, a very fast feature extraction method is presented. Then, covariance matrix eigenvector direction of feature points and those neighborhood points are calculated. Finally, an initial estimate for relative rigid-body transform can be realized, matching these eigenvector directions using an approach of Hough transform. Experimental results of 3D images taken by laser scanner are carried out to compare the convergence and registration error. The proposed registration approach can realize automatic registration without any assumptions about their initial positions and overcome the problems of traditional ICP in low overlapping and bad initial estimate.

Improvement of the kernel minimum squared error model for fast feature extraction

The kernel minimum squared error (KMSE) expresses the feature extractor as a linear combination of all the training samples in the high-dimensional kernel space. To extract a feature from a sample, KMSE should calculate as many kernel functions as the training samples. Thus, the computational efficiency of the KMSE-based feature extraction procedure is inversely proportional to the size of the training sample set. In this paper, we propose an efficient kernel minimum squared error (EKMSE) model for two-class classification. The proposed EKMSE expresses each feature extractor as a linear combination of nodes, which are a small portion of the training samples. To extract a feature from a sample, EKMSE only needs to calculate as many kernel functions as the nodes. As the nodes are commonly much fewer than the training samples, EKMSE is much faster than KMSE in feature extraction. The EKMSE can achieve the same training accuracy as the standard KMSE. Also, EKMSE avoids the overfitting problem. We implement the EKMSE model using two algorithms. Experimental results show the feasibility of the EKMSE model.

Fully affine invariant SURF for image matching

Fast and robust feature extraction is crucial for many computer vision applications such as image matching. The representative and the state-of-the-art image features include Scale Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), and Affine SIFT (ASIFT). However, neither of them is fully affine invariant and computation efficient at the same time. To overcome this problem, we propose in this paper a fully affine invariant SURF algorithm. The proposed algorithm makes full use of the affine invariant advantage of ASIFT and the efficient merit of SURF while avoids their drawbacks. Experimental results on applications of image matching demonstrate the robustness and efficiency of the proposed algorithm.

Fast and Robust Color Feature Extraction for Content-based Image Retrieval

Fast and Robust Color Feature Extraction for Content-based Image Retrieval

Laplacian smoothing transform for face recognition

In this paper, we investigate how to extract the lowest frequency features from an image. A novel Laplacian smoothing transform (LST) is proposed to transform an image into a sequence, by which low frequency features of an image can be easily extracted for a discriminant learning method for face recognition. Generally, the LST is able to be an efficient dimensionality reduction method for face recognition problems. Extensive experimental results show that the LST method performs better than other pre-processing methods, such as discrete cosine transform (DCT), principal component analysis (PCA) and discrete wavelet transform (DWT), on ORL, Yale and PIE face databases. Under the leave one out strategy, the best performance on the ORL and Yale face databases is 99.75% and 99.4%; however, in this paper, we improve both to 100% with a fast linear feature extraction method for the first time.

A Fast Feature Extraction Algorithm for Detection of Foreign Fiber in Lint Cotton within a Complex Background

A Fast Feature Extraction Algorithm for Detection of Foreign Fiber in Lint Cotton within a Complex Background

SU-GG-I-26: The Effect of Tomosynthesis Acquisition Geometry on Automated Feature Extraction Applied to Image Guided Radiation Therapy of the Prostate

Purpose: Gantry mounted kV imaging enables volumetric imaging of patients in the treatment position using cone beam computerized tomography(CBCT)reconstruction algorithms. Tomosynthesis techniques offer the ability to produce volumetric imaging with reduced set of projections, at the cost of significantly reduced image quality, but may be advantageous in terms of patient dose and scan time. Adaptive radiation therapy requires imaging and fast objective feature extraction algorithms to localize anatomic landmarks and determine treatment adjustments. This work analyzes the relation between tomosynthesis scan angle and localization accuracy, including automated feature extraction, in the setting of image guided radiation therapy(IGRT) of the prostate using a rectal balloon for daily immobilization. Method and Materials: A FDK‐based CBTS reconstruction was applied to selected sets of projection data to simulate volumetric imaging achievable from tomosynthesis for a range of scan angles. Projection data was calculated from planning CTs of 10 patients treated with rectal balloon immobilization. An intensity based edge detection algorithm was used to objectively identify over 50 points at the tissue‐air interface. Point locations were compared to the location of the corresponding points extracted for full‐rotation‐CBCT reconstruction. A similar process was applied to rectal‐balloon phantom measurements. Results: Reducing the tomosynthesis scan angle reduces the ability to identify the prostate‐rectum interface. However a level based thresholding algorithm can be used to automatically identify the air‐tissue interface. The accuracy is quantified in terms of the mean discrepancy as a function of reconstruction angle as well as the fraction of points identified as a function of distance to agreement to full‐rotation‐CBCT. Conclusion: This work quantifies and presents the accuracy of automatic extraction of rectal balloon‐prostate interface from tomosynthesis as a function of scan angle. Such results may guide the choice of scan parameters used for patient localization based on the desired localization accuracy.

A Fast Feature Extraction Algorithm for Detection of Foreign Fiber in Lint Cotton within a Complex Background

A novel algorithm is presented in this paper to extract the features of foreign fibers in lint cotton within a complex background. The 2D wavelet transform is used to implement the edge detection based on the gray contrast between foreign fibers and cotton background, while the color features are extracted in CrCgCb color cube to solve the problem of luminance fluctuation. Morphological analysis is a critical procedure of the algorithm and the discontinuity of object features and operation time must be considered. Therefore, the proposed approach integrates a two-level connected component labeling algorithm and a morphological identification algorithm based on equivalent length-width ratio. Tests on five typical kinds of foreign fibers were implemented, and the results show that the identification rate of the above-mentioned algorithm is about 95 %. The experimental results demonstrate that the feature extraction algorithm can identify foreign fibers effectively and can be used in real-time application.

Bilinear Lanczos components for fast dimensionality reduction and feature extraction

Matrix-based methods such as generalized low rank approximations of matrices (GLRAM) have gained wide attention from researchers in pattern recognition and machine learning communities. In this paper, a novel concept of bilinear Lanczos components (BLC) is introduced to approximate the projection vectors obtained from eigen-based methods without explicit computing eigenvectors of the matrix. This new method sequentially reduces the reconstruction error for a Frobenius-norm based optimization criterion, and the resulting approximation performance is thus improved during successive iterations. In addition, a theoretical clue for selecting suitable dimensionality parameters without losing classification information is presented in this paper. The BLC approach realizes dimensionality reduction and feature extraction by using a small number of Lanczos components. Extensive experiments on face recognition and image classification are conducted to evaluate the efficiency and effectiveness of the proposed algorithm. Results show that the new approach is competitive with the state-of-the-art methods, while it has a much lower training cost.

Fast extraction of wavelet-based features from JPEG images for joint retrieval with JPEG2000 images

In this paper, some fast feature extraction algorithms are addressed for joint retrieval of images compressed in JPEG and JPEG2000 formats. In order to avoid full decoding, three fast algorithms that convert block-based discrete cosine transform (BDCT) into wavelet transform are developed, so that wavelet-based features can be extracted from JPEG images as in JPEG2000 images. The first algorithm exploits the similarity between the BDCT and the wavelet packet transform. For the second and third algorithms, the first algorithm or an existing algorithm known as multiresolution reordering is first applied to obtain bandpass subbands at fine scales and the lowpass subband. Then for the subbands at the coarse scale, a new filter bank structure is developed to reduce the mismatch in low frequency features. Compared with the extraction based on full decoding, there is more than 72% reduction in computational complexity. Retrieval experiments also show that the three proposed algorithms can achieve higher precision and recall than the multiresolution reordering, especially around the typical range of compression ratio.

Visual seam tracking system for butt weld of thin plate

An image-based visual seam tracking system is designed for butt weld of thin plate in this paper. The torch deviation from the joint in the horizontal direction is measured from the seam image. Since the weld seam is so narrow, its image is based on natural lighting. Due to the hardness of measuring depth information using monocular vision, structured laser light is used for detecting the torch deviation in the vertical direction. For steady seam tracking, a robust and fast feature extraction algorithm is developed. Moreover, a reliable controller is designed with error filtering and output pulse verification. Finally, experiments are conducted to verify the proposed seam tracking system.

Fast Object-Level Change Detection for VHR Images

A novel approach is presented for change detection of very high resolution images, which is accomplished by fast object-level change feature extraction and progressive change feature classification. Object-level change feature is helpful for improving the discriminability between the changed class and the unchanged class. Progressive change feature classification helps improve the accuracy and the degree of automation, which is implemented by dynamically adjusting the training samples and gradually tuning the separating hyperplane. Experiments demonstrate the effectiveness of the proposed approach.

リバースエンジニアリングにおける自動位置合せに関する検討(機械力学,計測,自動制御)

This paper proposes a kind of feature extraction based efficient registration algorithm including coarse to fine registrations. During coarse registration, a very fast feature extraction method is present. Then an initial estimate for relative rigid-body transform can be brought to realization, through matching these regions with eigenvectors of covariance matrix of feature-closed point clouds, During coarse registration, a new fine registration method will be present in this paper based Iterative Closest Point (ICP) algorithm which had been used in fine registration to align view pairs after the coarse registration. Experimental results of 3D images taken by laser scanner are carried out to compare the convergence and registration result. The proposed registration approach can realize automatic registration without any assumptions about their initial positions, and can overcome the problems of traditional ICP in low overlapping and bad initial estimate. The proposed feature extraction based ICP algorithm and measure procedure management provides an efficient 3D modeling for computer-aided engineering and computer-aided design.

Fast Hand Feature Extraction Based on Connected Component Labeling, Distance Transform and Hough Transform

In hand gesture recognition or hand tracking systems relied on hand modeling methods, it is usually required to extract from a hand image some hand features. This paper presents a new robust method based on connected component labeling (CCL), distance transform (DT) and Hough transform (HT) to fast and precisely extract the center of the hand, the directions and the fingertip positions of all outstretched fingers on a skin color detection image. First, the method uses a simple but reliable technique that is performed on both the connected component labeling image and the distance transform image to extract the center of the hand and a set of features pixels, which are called distance-based feature pixels. Then, the Hough transform is calculated on these feature pixels to detect all outstretched fingers as lines. From the line detection result, the finger directions and the fingertip positions are determined easily and precisely. This method can be carried out fast and accurately, even when the skin color detection image includes hand, faces and some noise. Moreover, the number of distance-based feature pixels is usually not so high; therefore, the line detection process based on the Hough transform can be performed very fast. That can satisfy the demands of a real-time human-robot interaction system based on hand gestures or hand tracking.

Competitive code–based fast palmprint identification using a set of cover trees

A palmprint identification system recognizes a query palmprint image by searching for its nearest neighbor from among all the templates in a database. When applied on a large-scale identification system, it is often necessary to speed up the nearest-neighbor searching process. We use competitive code, which has very fast feature extraction and matching speed, for palmprint identification. To speed up the identification process, we extend the cover tree method and propose to use a set of cover trees to facilitate the fast and accurate nearest-neighbor searching. We can use the cover tree method because, as we show, the angular distance used in competitive code can be decomposed into a set of metrics. Using the Hong Kong PolyU palmprint database (version 2) and a large-scale palmprint database, our experimental results show that the proposed method searches for nearest neighbors faster than brute force searching.

Combining pores and ridges with minutiae for improved fingerprint verification

This paper presents a fast fingerprint verification algorithm using level-2 minutiae and level-3 pore and ridge features. The proposed algorithm uses a two-stage process to register fingerprint images. In the first stage, Taylor series based image transformation is used to perform coarse registration, while in the second stage, thin plate spline transformation is used for fine registration. A fast feature extraction algorithm is proposed using the Mumford–Shah functional curve evolution to efficiently segment contours and extract the intricate level-3 pore and ridge features. Further, Delaunay triangulation based fusion algorithm is proposed to combine level-2 and level-3 information that provides structural stability and robustness to small changes caused due to extraneous noise or non-linear deformation during image capture. We define eight quantitative measures using level-2 and level-3 topological characteristics to form a feature supervector. A 2 ν -support vector machine performs the final classification of genuine or impostor cases using the feature supervectors. Experimental results and statistical evaluation show that the feature supervector yields discriminatory information and higher accuracy compared to existing recognition and fusion algorithms.

CPU와 GPU의 병렬 처리를 이용한 고속 물체 인식 알고리즘 구현

This paper presents a fast feature extraction method for autonomous mobile robots utilizing parallel processing and based on OpenMP, SSE (Streaming SIMD Extension) and CUDA programming. In the first step on CPU version, the algorithms and codes are optimized and then implemented by parallel processing. The parallel algorithms are debugged to maintain the same level of performance and the process for extracting key points and obtaining dominant orientation with respect to key points is parallelized. After extraction, a parallel descriptor via SSE instructions is constructed. And the GPU version also implemented by parallel processing using CUDA based on the SIFT. The GPU-Parallel descriptor achieves an acceleration up to five times compared with the CPU-Parallel descriptor, but it shows the lower performance than CPU version. CPU version also speed-up the four and half times compared with the original SIFT while maintaining robust performance.

MULTI-VIEW GENDER CLASSIFICATION USING MULTI-RESOLUTION LOCAL BINARY PATTERNS AND SUPPORT VECTOR MACHINES

In this paper, we present a novel method for multi-view gender classification considering both shape and texture information to represent facial images. The face area is divided into small regions from which local binary pattern (LBP) histograms are extracted and concatenated into a single vector efficiently representing a facial image. Following the idea of local binary pattern, we propose a new feature extraction approach called multi-resolution LBP, which can retain both fine and coarse local micro-patterns and spatial information of facial images. The classification tasks in this work are performed by support vector machines (SVMs). The experiments clearly show the superiority of the proposed method over both support gray faces and support Gabor faces on the CAS-PEAL face database. A higher correct classification rate of 96.56% and a higher cross validation average accuracy of 95.78% have been obtained. In addition, the simplicity of the proposed method leads to very fast feature extraction, and the regional histograms and fine-to-coarse description of facial images allow for multi-view gender classification.