Selected Feature Points Research Articles

Moving object detection in aerial infrared images with registration accuracy prediction and feature points selection

Moving object detection in aerial infrared sequences is widely used in many applications, such as automatic traffic monitoring, border protection, and area surveillance. Feature-based image registration is the key step to compensate the background motion before detecting the moving objects. However, in practice it is often difficult to ensure the accurate registration, resulting in incomplete background compensation and thus decreasing detection performance. We find that the feature points in the object regions (FPOs) always impede the accurate registration of background. Besides, when the inlier points chosen by RANSAC excessively concentrate over the image, it will lead to inaccurate registration, especially in the case of insufficient feature points. These two issues are seldom studied before. In this paper, we propose a new framework of moving object detection, in which feature points selection and registration accuracy prediction are devised to improve detection accuracy. To do this, the detection information of previous frame is fed back to feature extraction of current frame for eliminating FPOs. Moreover, a quantitative metric is presented to measure the concentration of inlier points’ distribution over the image, with which one can predict the registration accuracy and determine whether to introduce additional cues for improving detection accuracy. Furthermore, we comprehensively investigate several classical feature extractors for aerial infrared image registration in terms of accuracy and running time, and suggest Speeded Up Robust Features (SURF) as feature detector and Local Difference Binary (LDB) as feature descriptor for image registration in aerial infrared sequences. Experimental results on benchmark infrared sequences show that our proposed method can greatly reduce false positives and also suppress false negatives, and performs favorably in terms of accuracy and efficiency.

Total variance based feature point selection and applications

Feature points are used to capture geometric characteristics of an object and are usually associated with certain anatomical significance or geometric meaning. The selection of feature points is a fundamental problem with various applications, for example, in shape registration, cross-parameterization, sparse shape reconstruction, parametric shape design, and dimension construction.In the literature, feature points are usually selected on a single shape by their differential property or saliency, and the information of similar shapes in the population are not considered. Though carefully chosen feature points can represent the corresponding shape well, the variations among different shapes within the population are overlooked.In this paper, through statistical shape modeling, we evaluate the feature points by the amount of variance they capture of the shape population, which leads to an algorithm that sequentially selects and ranks the feature points. In this way, the selected feature points explicitly incorporate the population information of the shapes. Then, we demonstrate how the proposed feature point selection approach can be integrated in the applications of sparse shape reconstruction, construction of new dimensions and shape classification through sparse measurements.The numerical examples have validated the effectiveness and efficiency of the proposed approach.

A Non-Cooperative Satellite Feature Point Selection Method for Vision-Based Navigation System.

The number of feature points on the surface of a non-cooperative target satellite used for monocular vision-based relative navigation affects the onboard computational load. A feature point selection method called the quasi-optimal method is proposed to select a subset of feature points with a good geometric distribution. This method, with the assumption that all of the feature points are in a plane and have the same variance, is based on the fact that the scattered feature points can provide higher accuracy than that of them grouped together. The cost is defined as a function of the angle between two unit vectors from the projection center to feature points. The redundancy of a feature point is calculated by summing all costs associated with it. Firstly, the feature point with the most redundant information is removed. Then, redundancies are calculated again with the second feature point removed. The procedures above are repeated until the desired number of feature points is reached. Dilution of precision (DOP) represents the mapping relation between the observation variance and the estimated variance. In this paper, the DOP concept is used in a vision-based navigation system to verify the performance of the quasi-optimal method. Simulation results demonstrate the feasibility of calculating the relative position and attitude by using a subset of feature points with a good geometric distribution. It also shows that the feature points selected by the quasi-optimal method can provide a high accuracy with low computation time.

Nonuniform multiview color texture mapping of image sequence and three-dimensional model for faded cultural relics with sift feature points

For faded relics, such as Terracotta Army, the 2D–3D registration between an optical camera and point cloud model is an important part for color texture reconstruction and further applications. This paper proposes a nonuniform multiview color texture mapping for the image sequence and the three-dimensional (3D) model of point cloud collected by Handyscan3D. We first introduce nonuniform multiview calibration, including the explanation of its algorithm principle and the analysis of its advantages. We then establish transformation equations based on sift feature points for the multiview image sequence. At the same time, the selection of nonuniform multiview sift feature points is introduced in detail. Finally, the solving process of the collinear equations based on multiview perspective projection is given with three steps and the flowchart. In the experiment, this method is applied to the color reconstruction of the kneeling figurine, Tangsancai lady, and general figurine. These results demonstrate that the proposed method provides an effective support for the color reconstruction of the faded cultural relics and be able to improve the accuracy of 2D–3D registration between the image sequence and the point cloud model.

A novel image registration approach via combining local features and geometric invariants.

Image registration is widely used in many fields, but the adaptability of the existing methods is limited. This work proposes a novel image registration method with high precision for various complex applications. In this framework, the registration problem is divided into two stages. First, we detect and describe scale-invariant feature points using modified computer vision-oriented fast and rotated brief (ORB) algorithm, and a simple method to increase the performance of feature points matching is proposed. Second, we develop a new local constraint of rough selection according to the feature distances. Evidence shows that the existing matching techniques based on image features are insufficient for the images with sparse image details. Then, we propose a novel matching algorithm via geometric constraints, and establish local feature descriptions based on geometric invariances for the selected feature points. Subsequently, a new price function is constructed to evaluate the similarities between points and obtain exact matching pairs. Finally, we employ the progressive sample consensus method to remove wrong matches and calculate the space transform parameters. Experimental results on various complex image datasets verify that the proposed method is more robust and significantly reduces the rate of false matches while retaining more high-quality feature points.

3D object recognition using scale-invariant features

As 3D scanning technology develops, it becomes easier to acquire various 3D surface data; thus, there is a growing need for 3D data registration and recognition technology. Many existing studies use local descriptors using local surface patches, and most of them use a fixed support radius, so they cannot cope perfectly when the model and scene have different scales. In this study, we propose a perfectly scale-invariant feature selection algorithm by extending the 2D SIFT algorithm (Lowe in Int J Comput Vis 60(2):91–110, 2004) to a 3D mesh. The feature selection method proposed in this study can obtain highly repeatable feature points and support radii regardless of mesh scale. The selected features can effectively describe the local information by the new shape descriptor proposed in this study. Unlike existing shape descriptors, it is possible to perform scale-invariant 3D object recognition and achieve a high recognition rate when combined with the feature point selection algorithm proposed in this study by using the gradients of the scalar functions defined on the 3D surface. We also reduced the searching space and lowered the false positive rate by suggesting a new RANSAC-based transformation hypotheses generation algorithm. Our 3D object recognition algorithm achieves recognition rates of 100 and 98.5%, respectively, when tested on the U3OR and CFVD datasets, exceeding the results of previous studies.

EXPERIMENTAL DEM EXTRACTION FROM ASTER STEREO PAIRS AND 3D REGISTRATION BASED ON ICESAT LASER ALTIMETRY DATA IN UPSTREAM AREA OF LAMBERT GLACIER, ANTARCTICA

Abstract. DEM Extraction from ASTER stereo pairs and three-dimensional registration by reference to ICESat laser altimetry data are carried out in upstream area of Lambert Glacier, East Antarctica. Since the study area is located in inland of East Antarctica where few textures exist, registration between DEM and ICESat data is performed. Firstly, the ASTER DEM generation is based on rational function model (RFM) and the procedure includes: a) rational polynomial coefficient (RPC) computation from ASTER metadata, b) L1A image product de-noise and destriping, c) local histogram equalization and matching, d) artificial collection of tie points and bundle adjustment, and e) coarse-to-fine hierarchical matching of five levels and grid matching. The matching results are filtered semi-automatically. Hereafter, DEM is interpolated using spline method with ground points converted from matching points. Secondly, the generated ASTER DEM is registered to ICESat data in three-dimensional space after Least-squares rigid transformation using singular value decomposition (SVD). The process is stated as: a) correspondence selection of terrain feature points from ICESat and DEM profiles, b) rigid transformation of generated ASTER DEM using selected feature correspondences based on least squares technique. The registration shows a good result that the elevation difference between DEM and ICESat data is low with a mean value less than 2 meters and the standard deviation around 7 meters. This DEM is generated and specially registered in Antarctic typical region without obvious ground rock control points and serves as true terrain input for further radar altimetry simulation.

A local feature with multiple line descriptors and its speeded-up matching algorithm

This paper introduces a local feature with multiple line descriptors and its unique matching algorithm. Previous approaches describe the local feature based on image patch that uses single feature point as the approximate center. But there is no accurate information about orientation or scale in the image patch. On the contrary, line segment possesses it. For this reason, we extract a line descriptor from a model of line segment that links two randomly selected feature points. There forms a mesh topology due to the fact that a line descriptor links two feature points and meanwhile a feature point links multiple line descriptors. But as a price to pay for it, there comes a large number of line descriptors that is bad for matching descriptors. In order to speed up the matching process, we design a unique matching algorithm by exploiting the mesh topology. The result shows that the local feature with multiple line descriptors outperforms other classical features based on image patch on robustness.

Evolutionary computing enriched ridge regression model for craniofacial reconstruction

Craniofacial reconstruction is one of the dominating research domains having vital significance towards forensic purposes as well as archaeological investigation needs. With a goal to enable an error-resilient and swift craniofacial reconstruction model, in this paper an evolutionary computing assisted enhanced regression model has been proposed. We have developed an enhanced over-fitting resilient regression model called Ridge Regression (RR) as a statistical method to perform craniofacial reconstruction using landmark points and skull-face/(tissue) skin features. Our proposed model incorporates a hybrid evolutionary computing scheme containing Particle Swarm Optimization (PSO) and Differential Evolution (DE) to perform feature point selection and landmark count reduction. Here, the prime objective is to reduce the feature sets and landmark that can eventually make craniofacial reduction process more time efficient and accurate. The performance assessment reveals that the proposed PSO-DEFS based RRM model outperforms existing approaches such as the least square support vector regression (LSSVR) and partial least square regression (PLSR).

Robust image hashing using progressive feature selection for tampering detection

The main problem addressed in this paper is the robust tamper detection of the image received in a transmission under various content-preserving attacks. To this aim the progressive feature point selection method is proposed to extract the feature points of high robustness; with which, the local feature and color feature are then generated for each feature point. Afterwards, the robust image hashing construction method is proposed by using the location-context information of the features. The constructed hash is attached to the image before transmission and it can be used for analyzing at destination to filter out the geometric transformations occurred in the received image. After image restoration, the similarity of the global hashes between the source image and restored image is calculated to determine whether the received image has the same contents as the trusted one or has been maliciously tampered. When the received image being judged as a tampered image, the hashes calculated with the proposed Horizontal Location-Context Hashing (HLCH) and Vertical Location-Context Hashing (VLCH) methods will be used to locate the tampered regions. Experimental results on different images with tampering of arbitrary size and location demonstrate that our image authentication and tampering localization scheme are superior to the state-of-the-art methods under various attacks.

Feature-Based Image Watermarking Algorithm Using SVD and APBT for Copyright Protection

Watermarking techniques can be applied in digital images to maintain the authenticity and integrity for copyright protection. In this paper, scale-invariant feature transform (SIFT) is combined with local digital watermarking and a digital watermarking algorithm based on SIFT, singular value decomposition (SVD), and all phase biorthogonal transform (APBT) is proposed. It describes the generation process of the SIFT algorithm in detail and obtains a series of scale-invariant feature points. A large amount of candidate feature points are selected to obtain the neighborhood which can be used to embed the watermark. For these selected feature points, block-based APBT is carried out on their neighborhoods. Moreover, a coefficients matrix of certain APBT coefficients is generated for SVD to embed the encrypted watermark. Experimental results demonstrate that the proposed watermarking algorithm has stronger robustness than some previous schemes. In addition, APBT-based digital watermarking algorithm has good imperceptibility and is more robust to different combinations of attacks, which can be applied for the purpose of copyright protection.

A high-fidelity multiresolution digital elevation model for Earth systems

Abstract. The impact of topography on Earth systems variability is well recognised. As numerical simulations evolved to incorporate broader scales and finer processes, accurately assimilating or transforming the topography to produce more exact land–atmosphere–ocean interactions, has proven to be quite challenging. Numerical schemes of Earth systems often use empirical parameterisation at sub-grid scale with downscaling to express topographic endogenous processes, or rely on insecure point interpolation to induce topographic forcing, which creates bias and input uncertainties. Digital elevation model (DEM) generalisation provides more sophisticated systematic topographic transformation, but existing methods are often difficult to be incorporated because of unwarranted grid quality. Meanwhile, approaches over discrete sets often employ heuristic approximation, which are generally not best performed. Based on DEM generalisation, this article proposes a high-fidelity multiresolution DEM with guaranteed grid quality for Earth systems. The generalised DEM surface is initially approximated as a triangulated irregular network (TIN) via selected feature points and possible input features. The TIN surface is then optimised through an energy-minimised centroidal Voronoi tessellation (CVT). By devising a robust discrete curvature as density function and exact geometry clipping as energy reference, the developed curvature CVT (cCVT) converges, the generalised surface evolves to a further approximation to the original DEM surface, and the points with the dual triangles become spatially equalised with the curvature distribution, exhibiting a quasi-uniform high-quality and adaptive variable resolution. The cCVT model was then evaluated on real lidar-derived DEM datasets and compared to the classical heuristic model. The experimental results show that the cCVT multiresolution model outperforms classical heuristic DEM generalisations in terms of both surface approximation precision and surface morphology retention.

Feature points selection with flocks of features constraint for visual simultaneous localization and mapping

Simultaneous localization and mapping is a crucial problem for mobile robots, which estimates the surrounding environment (the map) and, at the same time, computes the robot location in it. Most researchers working on simultaneous localization and mapping focus on localization accuracy. In visual simultaneous localization and mapping , localization is to calculate the robot’s position relative to the landmarks, which corresponds to the feature points in images. Therefore, feature points are of importance to localization accuracy and should be selected carefully. This article proposes a feature point selection method to improve the localization accuracy. First, theoretical and numerical analyses are conducted to demonstrate the importance of distribution of feature points. Then, an algorithm using flocks of features is proposed to select feature points. Experimental results show that the proposed flocks of features selector implemented in visual simultaneous localization and mapping enhances the accuracy of both localization and mapping, verifying the necessity of feature point selection.

3D motion estimation via optimized feature point selection

3D motion estimation via feature point matching is an important issue in machine vision. However, most existing methods are not good enough due to various defects, such as being sensitive to matching error, and hard to give an appropriate threshold for feature point selection. To solve these problems, a mathematical model is obtained from massive experiments to describe how the quantity and quality of matched feature points influence the motion estimation accuracy. This model can be used to select the most appropriate feature points for motion parameter calculation, hence threshold setting is avoided while estimation performance is optimized. Experimental results show that the proposed algorithm can improve the accuracy of 3D motion estimation by up to 50% with little effect on computation time. Furthermore, it is especially suitable for real time applications as there is no need to set threshold manually.

Threshold logic based low-level vision sparse object features

Purpose The real-time generation of feature descriptors for object recognition is a challenging problem. In this research, the purpose of this paper is to provide a hardware friendly framework to generate sparse features that can be useful for key feature point selection, feature extraction, and descriptor construction. The inspiration is drawn from feature formation processes of the human brain, taking into account the sparse, modular, and hierarchical processing of visual information.

A New Symbolization and Distance Measure Based Anomaly Mining Approach for Hydrological Time Series

Most of the time series data mining tasks attempt to discover data patterns that appear frequently. Abnormal data is often ignored as noise. There are some data mining techniques based on time series to extract anomaly. However, most of these techniques cannot suit big unstable data existing in various fields. Their key problems are high fitting error after dimension reduction and low accuracy of mining results. This paper studies an approach of mining time series abnormal patterns in the hydrological field. The authors propose a new idea to solve the problem of hydrological anomaly mining based on time series. They propose Feature Points Symbolic Aggregate Approximation (FP_SAX) to improve the selection of feature points, and then measures the distance of strings by Symbol Distance based Dynamic Time Warping (SD_DTW). Finally, the distances generated are sorted. A set of dedicated experiments are performed to validate the authors' approach. The results show that their approach has lower fitting error and higher accuracy compared to other approaches.

Researches on binocular vision pose measurement with selected feature points

Generally, the binocular vision system for pose measurement can identify much more than 3 feature points each time. However, the pose of an object can be measured with 3 feature points. Different combinations of feature points may make different effect on the measurement accuracy of the object pose. This article presents a method to optimize the selection of feature points in real time and consequently ensure optimized accuracy of pose measurement based on binocular vision. At first, the object poses measurement model based on binocular vision is established. Then, pose measurement error model is deduced, and the objective function is presented for optimizing the selection of feature points based on error transferring matrix. At last, simulations are conducted and the results verified the correctness of pose measurement error model and the effectiveness of the selected feature points.

Real Time Facial Emotion Recognition based on Image Processing and Machine Learning

Behaviors, actions, poses, facial expressions and speech; these are considered as channels that convey human emotions. Extensive research has being carried out to explore the relationships between these channels and emotions. This paper proposes a prototype system which automatically recognizes the emotion represented on a face. Thus a neural network based solution combined with image processing is used in classifying the universal emotions: Happiness, Sadness, Anger, Disgust, Surprise and Fear. Colored frontal face images are given as input to the prototype system. After the face is detected, image processing based feature point extraction method is used to extract a set of selected feature points. Finally, a set of values obtained after processing those extracted feature points are given as input to the neural network to recognize the emotion contained.

Reliable Microwave Modeling by Means of Variable-Fidelity Response Features

In this work, methodologies for low-cost and reliable microwave modeling are presented using variable-fidelity response features. The two key components of our approach are: 1) a realization of the modeling process at the level of suitably selected feature points of the responses (e.g., $S$ -parameters versus frequency) of the structure at hand and 2) the exploitation of variable-fidelity EM simulation data, also for the response feature representation. Due to the less nonlinear dependence between the coordinates of the feature points on the geometrical parameters of the structure of interest, the amount of training data can be greatly reduced. Additional cost reduction is obtained by means of generating the majority of the training data at a coarse-discretization EM simulation level and exploiting the correlations between the EM models of various fidelities. We propose two ways of combining the low- and high-fidelity data sets: 1) an external approach, through space mapping (simpler to implement) and 2) an internal approach, using co-kriging (more flexible and potentially offering better accuracy). The operation and performance of our modeling techniques are demonstrated by three microstrip filter examples and a compact rat-race coupler. A comprehensive verification and comparisons with several benchmark techniques, as well as application examples (filter optimization) are also provided.

Video Stabilization Based on Feature Trajectory Augmentation and Selection and Robust Mesh Grid Warping.

We propose a video stabilization algorithm, which extracts a guaranteed number of reliable feature trajectories for robust mesh grid warping. We first estimate feature trajectories through a video sequence and transform the feature positions into rolling-free smoothed positions. When the number of the estimated trajectories is insufficient, we generate virtual trajectories by augmenting incomplete trajectories using a low-rank matrix completion scheme. Next, we detect feature points on a large moving object and exclude them so as to stabilize camera movements, rather than object movements. With the selected feature points, we set a mesh grid on each frame and warp each grid cell by moving the original feature positions to the smoothed ones. For robust warping, we formulate a cost function based on the reliability weights of each feature point and each grid cell. The cost function consists of a data term, a structure-preserving term, and a regularization term. By minimizing the cost function, we determine the robust mesh grid warping and achieve the stabilization. Experimental results demonstrate that the proposed algorithm reconstructs videos more stably than the conventional algorithms.