Feature-based Algorithm Research Articles

A Novel One-Camera-Five-Mirror Three-Dimensional Imaging Method for Reconstructing the Cavitation Bubble Cluster in a Water Hydraulic Valve

In order to study the bubble morphology, a novel experimental and numerical approach was implemented in this research focusing on the analysis of a transparent throttle valve made by Polymethylmethacrylate (PMMA) material. A feature-based algorithm was written using the MATLAB software, allowing the 2D detection and three-dimensional (3D) reconstruction of bubbles: collapsing and clustered ones. The valve core, being an important part of the throttle valve, was exposed to cavitation; hence, to distinguish it from the captured frames, the faster region-based convolutional neural network (R-CNN) algorithm was used to detect its morphology. Additionally, the main approach grouping the above listed techniques was implemented using an optimized virtual stereo vision arrangement of one camera and five plane mirrors. The results obtained during this study validated the robust algorithms and optimization applied.

An Improvised Feature-Based Method for Sentiment Analysis of Product Reviews

In today’s society, sentiment analysis has gained due importance as it provides useful information about products that are used by variety of users. It gives a sneak peek of users’ reactions towards the products that are available in the market at an early stage. It thus intimates users’ perception and charts out a path that is beneficial for the market to grow as a whole. Although a lot of research is done to exploit the product based sentiment analysis but due to increase demand of the detailed components based products and their associated features, a novel method is desired to meet these criteria. So far, no such method is explored that analyses the product’s components and their features simultaneously, on the basis of sentiments of the users. This paper proposes an improvised Feature Based Algorithm (FBA) for the sentiment analysis of product reviews while formulating a tree structure of product, components, and associated features. In addition, evaluation of double negative sentences, detecting questions and emotions from the review sentences are measured which increases efficiency of the FBA method. The comparison of product’s components reviews is done with other existing algorithmsTF, TF-IDF and Naive Bayes to demonstrate that the proposed FBA is coherent and auspicious.

Abstract 2286: Feature-based clinical interpretation of whole exome and transcriptome data for precision cancer medicine

Abstract Background: Clinical interpretation algorithms and knowledge bases (e.g. PHIAL, OncoKB) are being used for clinical decision making. These approaches are typically based generally limited to “first order” genomic relationships (e.g. BRAFV600E & RAF/MEK inhibition). The increasing complexity of molecular data generated at the point of care, including whole exome and transcriptome results, along with the expanded therapeutic landscape in cancer, necessitate novel algorithms to enable robust and modern clinical interpretation of a cancer patient's molecular data to accelerate precision cancer medicine. We introduce a paired feature-based clinical interpretation algorithm and knowledge system for cancer genomic data to inform treatment decisions at the point of care and provide researchers with rapid assessment of tumor actionability. Methods: We expand upon PHIAL to predict actionability based on first-order genomics using SNVs (from both whole-exome sequencing and bulk RNA-seq), InDels, SCNAs, and fusions to further to infer global features of an individual tumor such as mutational burden, mutational signature profile, MSI-status, somatic-germline interaction, and connections between events. Predictive implication values were assigned to reflect the validities of the database's drug sensitivity, resistance, and prognostic claims. Results: We benchmarked the feature-based approach against PHIAL & TARGET across two cohorts that include both whole exome and transcriptomic data - 150 castrate resistant prostate cancers and 110 metastatic melanomas. PHIAL identified 1281 putatively actionable or biologically relevant alterations, with a median of 3 events per patient and 94% of patients having at least 1 event. The feature-based approach identified 1767 putatively actionable or biologically relevant variants or features, with a median of 5 events per patient and 97% of patients having at least 1 event. Of the these patients, 27% had at least 1 variant associated with an FDA-approved therapy and 18% had events associated with a clinical trial. It also identified 29% of samples as having a putatively actionable global feature. Conclusion: A DNA and RNA based interpretation method was able to identify and rank more putatively actionable first-order genomic alterations than PHIAL & TARGET, while also providing insight to global features of individual tumors. Increased accessibility of clinical interpretation through our cloud-based web portals and genomic reports may aid in sample contextualization, especially at the point of care. Source code and an interactive web portal for this project are available at chips.broadinstitute.org. Citation Format: Brendan Reardon, Nathanael Moore, Eric Kofman, Eliezer Mendell Van Allen. Feature-based clinical interpretation of whole exome and transcriptome data for precision cancer medicine [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2286.

Validating an Algorithm for Automatic Scoring of Inspiratory Flow Limitation Within a Range of Recording Settings.

Inspiratory Flow Limitation (IFL) is a phenomenon associated with narrowing of the upper airway, preventing an increase in inspiratory airflow despite an elevation in intrathoracic pressure. It has been shown that quantification of IFL might complement information provided by standard indices such as the apnea-hypopnea index (AHI) in characterizing sleep disordered breathing and identifying subclinical disease. Defining guidelines for visual scoring of IFL has been of increasing interest, and automated methods are desirable to avoid inter-scorer variability and allow analysis of large datasets. In addition, as recording instrumentation and practices may vary across hospitals and laboratories, it is useful to assess the influence of the recording parameters on the accuracy of the automated classification. We employed nasal pressure signals recorded as part of polysomnography (PSG) studies in 7 patients. Two experts independently classified approximately 2000 breaths per subject as IFL or non-IFL, and we used the consensus scoring as the gold standard. For each breath, we derived features indicative of the shape and frequency content of the signals and used them to train and validate a Support Vector Machine (SVM) to distinguish IFL from non-IFL breaths. We also assessed the effect of signal filtering (down-sampling and baseline-removal) on classification performance. The performance of the classifier was excellent (accuracy ~93%) for the raw signals (collected at 125 Hz with no filtering), and decreased for increasing high-pass cut-off frequencies (fc = [0.05, 0.1, 0.15, 0.2] Hz) down to 84% for fc= 0.2 Hz and for decreasing sampling rate (fs = [20, 50, 75, 100] Hz) down to ~85% for fs=20 Hz. Loss of performance was minimized when the classifier was re-trained using data with matched filtering characteristics (accuracy > 89%). We can conclude that the SVM feature-based algorithm provides a reliable and efficient tool for breath-by-breath classification.

Wideband Cyclostationary Signal Processing Using Sparse Subsets of Narrowband Subchannels

The problem of signal detection and modulation recognition is addressed in both the blind and nonblind contexts. Many relevant modulation recognition algorithms have been created over the past three decades. The essential engineering tradeoff that fuels the algorithm-invention process is between generality and optimality. Optimal (e.g., maximum-likelihood) algorithms can be devised for narrow subsets of signal types, whereas highly general but suboptimal feature-based algorithms can be devised for wide subsets of types. It has previously been shown that cyclostationary-based classifiers possess highly desirable properties, such as tolerance to noise and cochannel interference, but also that they are computational costly and so are resistant to real-time hardware implementation. In this paper, we present a reduced-complexity method of signal detection and modulation recognition that exploits the cyclostationarity of over-the-air observed communication signals. The key idea is to represent wideband input data as a set of narrowband, contiguous, nonoverlapping subchannels, and then to extract the desired cyclostationary features for the wideband data using only sparse subsets of the subchannels.

IMU-based gait analysis in lower limb prosthesis users: Comparison of step demarcation algorithms

BackgroundInertial Measurement Unit (IMU)-based gait analysis algorithms have previously been validated in healthy controls. However, little is known about the efficacy, performance, and applicability of these algorithms in clinical populations with gait deviations such as lower limb prosthesis users (LLPUs). Research questionTo compare the performance of 3 different IMU-based algorithms to demarcate steps from LLPUs. MethodsWe used a single IMU sensor affixed to the midline lumbopelvic region of 17 transtibial (TTA), 16 transfemoral (TFA) LLPUs, and 14 healthy controls (HC). We collected acceleration and angular velocity data during overground walking trials. Step demarcation was evaluated based on fore-aft acceleration, detecting either: (i) maximum acceleration peak, (ii) zero-crossing, or (iii) the peak immediately preceding a zero-crossing. We quantified and compared the variability (standard deviation) in acceleration waveforms from superposed step intervals, and variability in step duration, by each algorithm. ResultsWe found that the zero-crossing algorithm outperformed both peak detection algorithms in 65% of TTAs, 81% of TFAs, and 71% of HCs, as evidenced by lower standard deviations in acceleration, more consistent qualitative demarcation of steps, and more normally distributed step durations. SignificanceThe choice of feature-based algorithm with which to partition IMU waveforms into individual steps can affect the quality and interpretation of estimated gait spatiotemporal metrics in LLPUs. We conclude that the fore-aft acceleration zero-crossing serves as a more reliable feature for demarcating steps in the gait patterns of LLPUs.

Pilot study of feature-based algorithm for breech face comparison

A novel feature-based method, which is scale invariant feature transform (SIFT) and RANdom SAmple Consensus (RANSAC) integration algorithm, is introduced to promote the automated identification of the breech face impression, the most common mark left on the cartridge used for firearm evidence. SIFT algorithm is employed to extract the local extrema from examined impression as keypoints representing its invariant features, and to build the feature descriptor for each keypoint based on its local gradients in neighborhood. RANSAC is used to improve the matching performance among these keypoints and feature descriptors. With hypothesize-and-verify methods, RANSAC is able to construct the best model fitting initial matching pairs of keypoints and to guarantee the robust comparison result. Validation tests using 40 cartridge cases fired from pistols with 10 consecutively manufactured slides yielded a clear separation result, which strongly supports the effectiveness of the ensemble algorithm of SIFT and RANSAC. This application indicates the practical feasibility of feature-based algorithm and image processing technique in forensic science.

An Automatic Evaluation Platform for Feature Matching Algorithms Based on an Orbital Optical Pushbroom Stereo Imaging System

Thanks to the high robustness of feature-based matching algorithms, they can be applied in remote sensing (RS) applications with complex image changes. However, evaluating feature matching algorithms on RS images is still challenging, because creating ground truth matching data sets of RS images costs a lot of both computing time and human operator time. Therefore, in this letter, we present an evaluation platform for simulating RS ground truth data sets of feature-points correspondences based on a customizable orbital optical pushbroom stereo imaging system. With the help of the proposed platform, evaluating feature matching algorithms could be fully automatic and customized. The performance of three state-of-the-art feature matching algorithms based on local transformation constraint is evaluated and discussed on the proposed platform with comprehensive experiments. The evaluation results of the platform are also compared with the manual evaluation results of 10 pairs of real RS stereo images. The evaluation results show that the proposed platform indeed offers an efficient way for evaluating RS feature matching algorithms.

PCA/SVM-Based Method for Pattern Detection in a Multisensor System

This paper presents a multivariate analysis framework for pattern detection in a multisensor system; the proposed principal component analysis (PCA)/support vector machine- (SVM-) based supervision scheme can identify patterns in the multisensory system. Although the PCA and SVM are commonly used in pattern recognition, an effective methodology using the PCA/SVM for multisensory system remains unexplored. Pattern detection in a multisensor system has long been a challenge. For example, object inspections in multisensor systems are difficult to perform because inspectors might fail to use multiple sensing devices when concurrently detecting different patterns. Therefore, to resolve this issue, this study proposes a novel framework for establishing indicators and corresponding thresholds to identify patterns in the system; it employs a feature-based scheme that integrates principal component analysis (PCA) with an SVM for effectively detecting patterns in the system. Experiments were conducted using a tactile and optical measurement system. The experimental results demonstrated that the proposed method can effectively identify patterns in multisensor systems by using a feature-based algorithm that combines PCA and SVM classification for detecting various patterns. Moreover, the proposed framework established alarm indicators and corresponding thresholds that can be used for pattern detection.

Sea Ice Drift Tracking From Sequential SAR Images Using Accelerated-KAZE Features

In this paper, we propose a feature-tracking algorithm for sea ice drift retrieval from a pair of sequential satellite synthetic aperture radar (SAR) images. The method is based on feature tracking comprising feature detection, description, and matching steps. The approach exploits the benefits of nonlinear multiscale image representations using accelerated-KAZE (A-KAZE) features, a method that detects and describes image features in an anisotropic scale space. We evaluated several state-of-the-art feature-based algorithms, including A-KAZE, Scale Invariant Feature Transform (SIFT), and a very fast feature extractor that computes binary descriptors known as Oriented FAST and Rotated BRIEF (ORB) on dual polarized Sentinel-1A C-SAR extra wide swath mode data over the Arctic. The A-KAZE approach outperforms both ORB and SIFT up to an order of magnitude in ice drift. The experimental results showed high relevance of the proposed algorithm for retrieval of ice drift at subkilometre resolution from a pair of SAR images with 100-m pixel size. From this paper, we found that feature tracking using nonlinear scale-spaces is preferable due to its high efficiency against noise with respect to image features compared with other existing feature tracking alternatives that make use of Gaussian or linear scale spaces.

Number of Transmit Antennas Detection Using Time-Diversity of the Fading Channel

Detection of the number of transmit antennas plays an important role in the design of adaptive multiantenna wireless communication systems. This paper develops a detection algorithm based on second- and fourth-order statistics of the received signal and time-varying nature of the wireless channel. In contrast to existing feature-based algorithms, the proposed algorithm does not require a priori information about the transmitted signals, such as coding scheme, modulation type, and pilot patterns. Unlike information-theoretic algorithms, it does not require the number of receive antennas to be larger than that of transmit antennas. In fact, the proposed algorithm is capable of operating with a single receive antenna even for large numbers of transmit antennas. The performance analysis of the proposed algorithm is provided, and the results are verified through simulations. The algorithm exhibits a good performance over a wide range of signal-to-noise ratios, and is robust to the carrier frequency and timing offsets.

A morphing-based scheme for large deformation analysis with stereo-DIC

A key step in the DIC-based image registration process is the definition of the initial guess for the non-linear optimization routine aimed at finding the parameters describing the pixel subset transformation. This initialization may result very challenging and possibly fail when dealing with pairs of largely deformed images such those obtained from two angled-views of not-flat objects or from the temporal undersampling of rapidly evolving phenomena. To address this problem, we developed a procedure that generates a sequence of intermediate synthetic images for gradually tracking the pixel subset transformation between the two extreme configurations. To this scope, a proper image warping function is defined over the entire image domain through the adoption of a robust feature-based algorithm followed by a NURBS-based interpolation scheme. This allows a fast and reliable estimation of the initial guess of the deformation parameters for the subsequent refinement stage of the DIC analysis.The proposed method is described step-by-step by illustrating the measurement of the large and heterogeneous deformation of a circular silicone membrane undergoing axisymmetric indentation. A comparative analysis of the results is carried out by taking as a benchmark a standard reference-updating approach. Finally, the morphing scheme is extended to the most general case of the correspondence search between two largely deformed textured 3D geometries. The feasibility of this latter approach is demonstrated on a very challenging case: the full-surface measurement of the severe deformation (> 150% strain) suffered by an aluminum sheet blank subjected to a pneumatic bulge test.

Star identification based on euclidean distance transform, voronoi tessellation, and k-nearest neighbor classification

In this paper, an image-based lost in space star identification algorithm is developed for a typical daytime star tracker that uses uniform selection of the 2 Micron All Sky Survey (2MASS) infrared point source catalog. The proposed algorithm uses Euclidean distance transform of the image followed by Voronoi tessellation of stars and k-nearest-neighbor cell classification. This algorithm recognizes more than 95% of 1000 frames containing more than 50 spikes using Monte Carlo simulation within a desirable update frequency. Performance of image-based algorithm is compared to traditional feature-based algorithms in the presence of spikes. The feature-based method uses star triangle planar and spatial angles followed by the k-vector search technique and the geometric voting concept in an iterative process. The algorithm is optimized using multiobjective genetic algorithm Pareto front analysis. Monte Carlo simulation of the feature-based method using optimal parameters shows unacceptable reduction of robustness in the presence of more than 12 spikes. In order to analyze the effects of hardware, both algorithms are evaluated in a hardware-in-the-loop simulation using prototype star tracker Nasir-I, available at the SSDI laboratory. The results of 100 sample images from the 2MASS catalog accord with the software simulation. The proposed image-based algorithm shows acceptable performance in small fields of view requiring modest memory and computation cost, making it suitable for small satellites as well.

Determining woody-to-total area ratio using terrestrial laser scanning (TLS)

Accurately determining woody-to-total area ratio (WTA) is a key step to indirectly retrieve leaf area index (LAI) from terrestrial laser scanning (TLS) data. In this work, we first collected both individual tree and forest plot point cloud data (PCD) from broadleaf and coniferous tree species and leaf characteristics using both side-lateral and full field-of-view TLS field setups with scan distances between 2.5 to 28m. Using a local geometrical feature-based algorithm, the generated PCD were automatically classified into three different categories including photosynthetic canopy components, non-photosynthetic canopy components, and bare earth. To convert each classified point into a surface area, we then developed and validated a novel approach that considers sampling space, laser incidence angle, and leaf orientation information. The estimated surface areas from this approach showed strong agreements with validation datasets for single leaf (91.44%), photosynthetic (95.64%), and non-photosynthetic canopy components (89.60%) of an artificial tree and stems of an old-growth coniferous tree (93.53%), two individual broadleaf trees (98.31% and 97.46%) and a broadleaf forest plot (90.26%). By doing this, we computed the parameter WTA for an individual artificial tree (10.90%), an old-growth coniferous tree (29.97%), two individual broadleaf tree (14.83% and 4.27%) and four natural forest stands ranging from 7.74%–15.57%, respectively. The proposed method can effectively improve the accuracy of retrieving true LAI by removing the effects of woody components and converting each point into a surface area.

A comparative study of the performance of local feature-based pattern recognition algorithms

Several feature-based pattern recognition algorithms have been developed during the past decade. These algorithms rely on identifying keypoints in an image and assigning a descriptor to each point based on the composition of their surrounding region. Comparison of the descriptors of keypoints found in two images enables these algorithms to match similar objects within those images. The dependence of these algorithms’ performance on the similarity of the internal structure of objects makes them susceptible to modifications that change this internal structure. In this paper, we first compare the relative performance of some major feature-based algorithms in finding similar objects surrounded by geometrical noise. Next, we add several noise and transformation types to target objects and re-evaluate the performance of these algorithms under the resulting structural changes. Our results provide insights on the relative strengths of these algorithms in the presence and absence of several noise and transformation types. In addition, these findings allow us to identify modification types that can better inhibit the performance of these algorithms. The resulting insight can be used in applications that need to build resistance against such algorithms, e.g., in developing CAPTCHAs that need to be resistant to recognition attacks.

Real Time Monocular Visual Odometry using ORB Features for Indoor Environment

Navigation is a key process in many intelligent systems. The most common way to do navigation is using Global Positioning System (GPS). However, in indoor environment, GPS is inaccurate due to multi-path problem. To solve this problem, a dead reckoning system may be implemented. Visual odometry is one of the dead reckoning method which can help solving this problem. Among many types of visual odometry algorithm, feature-based monocular visual odometry algorithm is the one which will be built in this work. The feature detector used in this work is ORB. The features is used for triangulation and the obtained 3D structure will be used as the basic information for poses determination. Pose determination is done by solving PnP problem. After the visual odometry algorithm has been built, the performance of this algorithm is tested. The result of the performance test shows that the visual odometry algorithm can determine the position and orientation with good accuracy.

A measurement setup for the 3D validation of fingertip deformation models

Deformable models of the human fingertip are commonly validated by evaluating a finger's silhouette when charged with a line load. To validate models, it would be desirable to measure 3D data of the fingertip at a high frequency. This paper proposes the use of a stereo camera setup to measure the 3D surface of the fingertip's side. The setup offers a theoretical depth resolution of 0.17 mm at a frame rate above 100 Hz. A feature point matching algorithm incorporating scene knowledge is implemented to obtain 3D points to which a paraboloid surface is fitted. Precision is compared with reference to a LIDAR sensor and experiments applying normal and tangential force to the fingertip recorded with a 6-DOF force sensor setup. The results imply that state-of-the-art feature-based algorithms are insufficient or unsuited to match the fingertip surface reliably. Further approaches are proposed to improve robustness.

Efficient image classification via sparse coding spatial pyramid matching representation of SIFT‐WCS‐LTP feature

Shape and texture information are critical to the accuracy of image classification systems. In this study, the authors propose a novel descriptor called weighted centre-symmetric local ternary pattern (WCS-LTP), better characterising the image local texture. Then, based on the proposed WCS-LTP descriptor, they introduce a new local scale invariant feature transform and WCS-LTP (SIFT–WCS-LTP) feature extraction approach. Compared with conventional local CS-LTP and SIFT features, the authors’ proposed SIFT–WCS-LTP feature can not only capture the shape information of images, but also tend to extract more precise texture information. Finally, SIFT–WCS-LTP feature-based sparse coding spatial pyramid matching (ScSPM) representation classification is proposed for image classification. Extensive experimental results demonstrate that the effectiveness of their proposed SIFT–WCS-LTP feature-based ScSPM representation classification algorithm.

Iris Recognition Using the Nonsubsampled Contourlet Transform

Biometric information is widely used in user identification system. Because of the unique and invariant properties of the iris through a lifetime, iris recognition is one of the most stable and reliable means in biometric identification. Extracting distinguishable iris features for iris recognition is very important. In this paper, for capturing effective texture features that represent the complex directional structures of an iris image, a new iris recognition method using the nonsubsampled contourlet transform (NSCT) features is proposed. With the shift-invariance, multiscale, and multidirection properties, significant NSCT coefficient features along the radial and angular directions in an iris image can be represented efficiently. Iris segmentation and normalization are considered at first as pre-processing. The modified normalized iris image is obtained from the normalized iris regions for extracting the robust iris features, and then is filtered with the NSCT to obtain the distinct coefficient features in each directional subband. Next, using the NSCT coefficients in each subband, an iris code vector is constructed for iris matching. Comparison of experimental results of the proposed and existing methods with three databases show the effectiveness of the proposed NSCT feature-based iris recognition algorithm, in terms of the three performance measures.

Neighboring constraint-based pairwise point cloud registration algorithm

Three-dimensional point cloud registration is important in reverse engineering. In this paper, we propose a registration method for large-scale 3D point clouds, which is based on neighborhood constraints of geometrical features. The method consists of initial and exact registration steps.In the process of initial registration, we define a new functon that measures feature similarity by calculating the distance function, and in the process of exact registration, we introduce the angle information that improve the accuracy of iterative closest point algorithm. Compared with the traditional feature-based and iterative closest point algorithms, our method significantly reduced the registration time by 11.9 % and has only 1 % of the registration error of the traditional feature-based algorithm. The proposed algorithm can be used to create efficient 3D models for virtual plant reconstruction and computer-aided design, and the registration results can provide a reference for virtual plant reconstruction and growth.