Random Sample And Consensus Research Articles

Compact rover surveying and laser scanning for BIM development.

This paper presents a custom made small rover based surveying, mapping and building information modeling solution. Majority of the commercially available mobile surveying systems are larger in size which restricts their maneuverability in the targeted indoor vicinities. Furthermore their functional cost is unaffordable for low budget projects belonging to developing markets. Keeping in view these challenges, an economical indigenous rover based scanning and mapping system has developed using orthogonal integration of two low cost RPLidar A1 laser scanners. All the instrumentation of the rover has been interfaced with Robot Operating System (ROS) for online processing and recording of all sensorial data. The ROS based pose and map estimations of the rover have performed using Simultaneous Localization and Mapping (SLAM) technique. The perceived class 1 laser scans data belonging to distinct vicinities with variable reflective properties have been successfully tested and validated for required structural modeling. Systematically the recorded scans have been used in offline mode to generate the 3D point cloud map of the surveyed environment. Later the structural planes extraction from the point cloud data has been done using Random Sampling and Consensus (RANSAC) technique. Finally the 2D floor plan and 3D building model have been developed using point cloud processing in appropriate software. Multiple interiors of existing buildings and under construction indoor sites have been scanned, mapped and modelled as presented in this paper. In addition, the validation of the as-built models have been performed by comparing with the actual architecture design of the surveyed buildings. In comparison to available surveying solutions present in the local market, the developed system has been found faster, accurate and user friendly to produce more enhanced structural results with minute details.

Combining Feature Correspondence With Parametric Chamfer Alignment: Hybrid Two-Stage Registration for Ultra-Widefield Retinal Images.

We propose a novel hybrid framework for registering retinal images in the presence of extreme geometric distortions that are commonly encountered in ultra-widefield (UWF) fluorescein angiography. Our approach consists of two stages: a feature-based global registration and a vessel-based local refinement. For the global registration, we introduce a modified RANSAC (random sample and consensus) that jointly identifies robust matches between feature keypoints in reference and target images and estimates a polynomial geometric transformation consistent with the identified correspondences. Our RANSAC modification particularly improves feature point matching and the registration in peripheral regions that are most severely impacted by the geometric distortions. The second local refinement stage is formulated in our framework as a parametric chamfer alignment for vessel maps obtained using a deep neural network. Because the complete vessel maps contribute to the chamfer alignment, this approach not only improves registration accuracy but also aligns with clinical practice, where vessels are typically a key focus of examinations. We validate the effectiveness of the proposed framework on a new UWF fluorescein angiography (FA) dataset and on the existing narrow-field FIRE (fundus image registration) dataset and demonstrate that it significantly outperforms prior retinal image registration methods in accuracy. The proposed approach enhances the utility of large sets of longitudinal UWF images by enabling: (a) automatic computation of vessel change metrics such as vessel density and caliber, and (b) standardized and co-registered examination that can better highlight changes of clinical interest to physicians.

Accuracy of a Smartphone-based freeware solution and two shape reconstruction algorithms in log volume measurements

Enhancing accuracy, mobility, safety, ergonomics, data transfer and traceability functions in log measurement applications are prerequisites of modern business in the wood supply chain. Various methods and methodologies have been tested so far to overcome the current limitations in the attempt to move the wood supply chain to Forestry 4.0 concepts, many of which still share the same problem, namely the high costs incurred by data acquisition. A freeware application was used in this study to collect point cloud data on 612 logs of various sizes and two external shape reconstruction algorithms, namely Random Sampling and Consensus (RANSAC) and Poisson interpolation, were then used to produce volume estimates under a mobile scanning approach (MS). These were compared against the estimates produced by a highly-detailed manual measurement (MM). At the sample level, which included logs reaching volumes of more than 3.5 m3, the differential bias of the methods ranged between 0.037 and 0.075 m3, while the errors metrics such as the mean absolute error (MAE), mean squared error (MSE) and root mean square error (RMSE) were between 0.047 and 0.081, 0.006 to 0.023 and 0.076 to 0.151 m3. The results indicate that, although there was a proportional bias in MS estimates, the approach would be useful for logs in size of 0.25 to 0.40 m3, while these log sizes largely resemble the characteristics of the wood supply in many areas of the world. Further studies should focus on the capability and accuracy of the tested application to instantly produce accurate estimates, as well as on the integration of deep learning to produce better estimates.

Parametric Classification of Furniture From Point Cloud Developed Using Low Cost Trolley Based Laser Scanning System

This paper presents a parametric classification methodology to identify common indoor and outdoor furniture objects present in the 3D Cartesian point cloud of the surveyed environment. For this purpose, a low cost custom made trolley based scanning and surveying system has developed using orthogonal integration of two popular Hokuyo-30LX 2D laser scanners. The developed system has been successfully used to generate 3D point cloud of the environment using Simultaneous Localization and Mapping (SLAM) technique. The instrumentation system of the trolley has been interfaced through Robot Operating System (ROS) for online processing and recording of all sensorial data. While classification of the furniture present in point cloud has been done in offline mode using Random Sampling and Consensus (RANSAC) based parametric segmentation technique. The innovative furniture detection has applied on each scan in order to reduce the region of interest in the developed point cloud. In addition, the validation of the classified furniture objects has been performed using Fuzzy Logic. Multiple indoor and outdoor vicinities have been scanned and modelling results have been found accurate nearer to ground truth. In comparison to available surveying solutions present in the local market, the developed system has been found faster and precise to produce more enhanced structural results with minute details.

Automated extraction of geometric primitives with solid lines from unstructured point clouds for creating digital buildings models

Point clouds produced by laser scanners are an invaluable source of data for reconstructing multi-dimensional digital models that reflect the as-is conditions of built facilities. However, previous studies aimed to reconstruct models by overlaying the dataset on top of ground-truth reference models to manually adjust the accuracy of the output. Therefore, this paper describes the extraction of geometric primitives with solid lines—the simplest form of objectified data that computer-aided design systems can handle—from unorganized data points and creation of digital models of built facilities in a form of floor plan. The geometric primitives are extracted from 3D points by hybridizing machine learning algorithms, which are mean-shift clustering, non-convex hull, and random sample and consensus (RANSAC). This paper provides a solution for creating a new form of as-built model with high accuracy and robustness from scratch without the involvement of ground-truth solutions or manual adjustments.

Classification of airborne 3D point clouds regarding separation of vegetation in complex environments.

Classification of outdoor point clouds is an intensely studied topic, particularly with respect to the separation of vegetation from the terrain and manmade structures. In the presence of many overhanging and vertical structures, the (relative) height is no longer a reliable criterion for such a separation. An alternative would be to apply supervised classification; however, thousands of examples are typically required for appropriate training. In this paper, an unsupervised and rotation-invariant method is presented and evaluated for three datasets with very different characteristics. The method allows us to detect planar patches by filtering and clustering so-called superpoints, whereby the well-known but suitably modified random sampling and consensus (RANSAC) approach plays a key role for plane estimation in outlier-rich data. The performance of our method is compared to that produced by supervised classifiers common for remote sensing settings: random forest as learner and feature sets for point cloud processing, like covariance-based features or point descriptors. It is shown that for point clouds resulting from airborne laser scans, the detection accuracy of the proposed method is over 96% and, as such, higher than that of standard supervised classification approaches. Because of artifacts caused by interpolation during 3D stereo matching, the overall accuracy was lower for photogrammetric point clouds (74-77%). However, using additional salient features, such as the normalized green-red difference index, the results became more accurate and less dependent on the data source.

Fundamental Matrix Computing Based on 3D Metrical Distance

To reconstruct point geometry from multiple images, computation of the fundamental matrix is always necessary. With a new optimization criterion, i.e., the re-projective 3D metric geometric distance rather than projective space under RANSAC (Random Sample And Consensus) framework, our method can reveal the quality of the fundamental matrix visually through 3D reconstruction. The geometric distance is the projection error of 3D points to the corresponding image pixel coordinates in metric space. The reasonable visual figures of the reconstructed scenes are shown but only some numerical result were compared, as is standard practice. This criterion can lead to a better 3D reconstruction result especially in 3D metric space. Our experiments validate our new error criterion and the quality of fundamental matrix under the new criterion.

Robust Image Stitching and Reconstruction of Rolling Stocks Using a Novel Kalman Filter With a Multiple-Hypothesis Measurement Model

This work introduces a novel algorithm for the reconstruction of rolling stocks from a sequence of images. The research aims at producing an accurate and wide image model that can be used as a Digital Twin (DT) for diagnosis, fault prediction, maintenance, and other monitoring operations. When observing large surfaces with nearly constant textures, metallic reflections, and repetitive patterns, motion estimation algorithms based on whole image error minimization and feature pairing with Random Sampling and Consensus(RANSAC) or Least Median of Squares (LMedS) fail to provide appropriate associations. To overcome such an issue, we propose a custom Kalman Filter (KF) modified by adding multiple input-noise sources represented as a Gaussian mixture distribution (GM), and specific algorithms to select appropriate data and variance to use for state prediction and correction. The proposed algorithm has been tested on images of train vessels, having a high number of windows, and large metallic paintings with constant or repetitive patterns. The approach here presented showed to be robust in the presence of high environmental disturbances and a reduced number of features. A large set of rolling stocks has been collected during a six months campaign. The set was employed to demonstrate the validity of the proposed algorithm by comparing the reconstructed twin versus known data. The system showed an overall accuracy in length estimation above 99%.

Random sampling and probabilistic consensus for identifying outliers in road surface datasets

Road surface measurement plays a crucial role in the modelling and simulation of vehicles as the road surface is one of the primary means of excitation. A prevalent technique for measuring road surfaces utilises scanning lasers whose measurements produce a non-uniform, 3-dimensional point cloud representation, in which statistical outliers typically manifest. In this work, a novel, axiomatic, probabilistic method for simultaneously identifying outliers and estimating the road surface height at uniformly spaced grid nodes is developed. The method expands on the concepts used in the seminal model fitting algorithm, random sampling and consensus (RANSAC), to address a situation in which multiple underlying models may exist in a neighbourhood of the data. The proposed method, called random sampling and probabilistic consensus (RSPC), is evaluated on a 2-dimensional simulated road surface dataset containing 60% outliers in order to demonstrate its effectiveness at identifying outliers and simultaneously estimating grid node heights.

AWARE: An Algorithm for the Automated Characterization of EUV Waves in the Solar Atmosphere

Extreme ultraviolet (EUV) waves are large-scale propagating disturbances observed in the solar corona, frequently associated with coronal mass ejections and flares. They appear as faint, extended structures propagating from a source region across the structured solar corona. Since their discovery, over two hundred papers discussing their properties, causes and physical nature have been published. However, despite this their fundamental properties and the physics of their interactions with other solar phenomena are still not understood. To further the understanding of EUV waves, we have constructed the Automated Wave Analysis and REduction (AWARE) algorithm for the measurement of EUV waves. AWARE is implemented in two stages. In the first stage, we use a new type of running difference image, the running difference persistence image, which enables the efficient isolation of propagating, brightening wavefronts as they propagate across the corona. In the second stage, AWARE detects the presence of a wavefront, and measures the distance, velocity and acceleration of that wavefront across the Sun. The fit of propagation models to the wave progress isolated in the first stage is achieved using the Random Sample and Consensus (RANSAC) algorithm. AWARE is tested against simulations of EUV wave propagation, and is applied to measure EUV waves in observational data from the Atmospheric Imaging Assembly (AIA). We also comment on unavoidable systematic errors that bias the estimation of wavefront velocity and acceleration. In addition, the full AWARE software suite comes with a package that creates simulations of waves propagating across the disk from arbitrary starting points.

Efficient and robust lane marking extraction from mobile lidar point clouds

Surveys of roadways with Mobile Laser Scanning (MLS) are now being conducted on a regular basis by many transportation agencies to provide detailed geometric information to support a wide range of applications, including asset management. Most MLS systems provide intensity (return signal strength) data as a point attribute in georeferenced point clouds, which may be used to estimate retro-reflectivity of pavement markings for effective maintenance. Nevertheless, the extraction of pavement markings from mobile lidar data remains an open challenge, due to variable noise, degree of wear on the markings, and road conditions. This paper addresses these challenges, presenting a novel approach for efficient, reliable extraction of lane markings, including those that have been significantly worn. First, using the MLS trajectory information, the lidar data is discretized into smaller sections, and then transformed to the local coordinate system, such that the road surface is near-horizontal for reliable extraction on roads with significant grade. Subsequently, the road surface is extracted using the constrained Random Sampling and Consensus (RANSAC) algorithm and then rasterized into a 2D intensity image to apply image processing techniques, namely: image segmentation to separate the lane markings from the road pavement, and a morphological opening operation to remove small objects. However, the extracted lane markings are prone to over-segmentation, due to occlusions or worn portions caused by moving vehicles. To rectify this, topologically-similar lane markings are associated with each other by computing line parameters (i.e., orientation and distance from the origin), which enables the gaps to be filled among the associated lanes. Finally, the remaining incorrect lane markings are detected and removed through a noise filtering phase using Dip test statistics. Examples of the effectiveness and application of the methodology are shown for a variety of sites with stripes of variable condition to highlight the robustness of the approach. Using optimized parameter values, the algorithm achieved F1 scores of 89–97% when tested on a variety of datasets encompassing a wide range of road scene types.

DIRSAC: A directed sample and consensus algorithm for localization with quasi-degenerate data

We propose a new method that uses an iterative closest point (ICP) algorithm to fit three-dimensional points to a prior geometric model for the purpose of determining the position and orientation (pose) of a sensor with respect to a model. We use a method similar to the Random Sample and Consensus (RANSAC) algorithm. However, where RANSAC uses random samples of points in the fitting trials, DIRSAC DIRects the sampling by ordering the points according to their contribution to the solution constraints. This is particularly important when the data is quasi-degenerate; meaning that some of the degrees of freedom of the pose are under constrained. In this case, the standard RANSAC algorithm often fails to find the correct solution. Our approach uses mutual information to avoid redundant points that result in degenerate sample sets. We demonstrate our approach on real data and show that in the case of quasi-degenerate data, the proposed algorithm outperforms RANSAC.

Sketch and Validate for Big Data Clustering

In response to the need for learning tools tuned to big data analytics, the present paper introduces a framework for efficient clustering of huge sets of (possibly high-dimensional) data. Building on random sampling and consensus (RANSAC) ideas pursued earlier in a different (computer vision) context for robust regression, a suite of novel dimensionality- and set-reduction algorithms is developed. The advocated sketch-and-validate (SkeVa) family includes two algorithms that rely on K-means clustering per iteration on reduced number of dimensions and/or feature vectors: The first operates in a batch fashion, while the second sequential one offers computational efficiency and suitability with streaming modes of operation. For clustering even nonlinearly separable vectors, the SkeVa family offers also a member based on user-selected kernel functions. Further trading off performance for reduced complexity, a fourth member of the SkeVa family is based on a divergence criterion for selecting proper minimal subsets of feature variables and vectors, thus bypassing the need for K-means clustering per iteration. Extensive numerical tests on synthetic and real data sets highlight the potential of the proposed algorithms, and demonstrate their competitive performance relative to state-of-the-art random projection alternatives.

Research on 3D Virtual Environment Modeling Technology for Space Tele-robot

To overcome the time delay in tele-robot system, a new 3D virtual environment modeling technology is proposed in this paper. The interactions between slave robot and environment can be attained in advance in a 3D virtual environment in the master side, therefore an accurate operation of space tele-robot can be realized and the effect of time delay would be minimized. Firstly, 3D virtual interaction scenario is modeled as a point cloud data image, and the target objects are recognized from the image and reconstructed by using Random Sample And Consensus (RANSAC) algorithm. Secondly, an effective method is proposed to modify the position of virtual robot and calculate the virtual interactive force between the virtual robot and the virtual objects. Lastly, the experiment is completed with the time delay. The errors are lower (10% F.S.), while virtual interaction force is compared with the real force measured by the force sensor. Experimental results show that this 3D virtual environment modeling method is effective and reliable for space tele-robot control.

Haptic Object Recognition in Underwater and Deep‐sea Environments

Tactile sensors, because of their intrinsic insensitivity to lighting conditions and water turbidity, provide promising opportunities for perception and object recognition in underwater and deep‐sea environments. However, the limited availability of tactile sensors for underwater use has led to limited research in this domain. Recently, we have developed a deep‐sea‐capable tactile sensing system, with high spatial and force resolutions, which has made underwater haptic exploration possible for the first time. This paper presents a tactile sensor‐based object recognition and localization methodology for structured underwater and deep‐sea applications. Our approach is based on database matching using a local feature‐based Random Sampling and Consensus (RANSAC) algorithm, and sequentially evolving the resulting hypotheses over the course of object exploration. It can handle a large database of three‐dimensional objects of complex shapes, and it performs a complete six degree of freedom localization of a static object. An approach to utilize both contact and free‐space measurements is presented. Extensive experimentation is performed in underwater environments for validating both the sensor system and the algorithms. To our knowledge, this is the first instance of haptic object recognition and localization in underwater and deep‐sea environments.

Object Recognition and Localization: The Role of Tactile Sensors

Tactile sensors, because of their intrinsic insensitivity to lighting conditions and water turbidity, provide promising opportunities for augmenting the capabilities of vision sensors in applications involving object recognition and localization. This paper presents two approaches for haptic object recognition and localization for ground and underwater environments. The first approach called Batch Ransac and Iterative Closest Point augmented Particle Filter (BRICPPF) is based on an innovative combination of particle filters, Iterative-Closest-Point algorithm, and a feature-based Random Sampling and Consensus (RANSAC) algorithm for database matching. It can handle a large database of 3D-objects of complex shapes and performs a complete six-degree-of-freedom localization of static objects. The algorithms are validated by experimentation in ground and underwater environments using real hardware. To our knowledge this is the first instance of haptic object recognition and localization in underwater environments. The second approach is biologically inspired, and provides a close integration between exploration and recognition. An edge following exploration strategy is developed that receives feedback from the current state of recognition. A recognition by parts approach is developed which uses the BRICPPF for object sub-part recognition. Object exploration is either directed to explore a part until it is successfully recognized, or is directed towards new parts to endorse the current recognition belief. This approach is validated by simulation experiments.

A hierarchical approach for speech-instrumental-song classification.

Audio classification acts as the fundamental step for lots of applications like content based audio retrieval and audio indexing. In this work, we have presented a novel scheme for classifying audio signal into three categories namely, speech, music without voice (instrumental) and music with voice (song). A hierarchical approach has been adopted to classify the signals. At the first stage, signals are categorized as speech and music using audio texture derived from simple features like ZCR and STE. Proposed audio texture captures contextual information and summarizes the frame level features. At the second stage, music is further classified as instrumental/song based on Mel frequency cepstral co-efficient (MFCC). A classifier based on Random Sample and Consensus (RANSAC), capable of handling wide variety of data has been utilized. Experimental result indicates the effectiveness of the proposed scheme.Electronic supplementary materialThe online version of this article (doi:10.1186/2193-1801-2-526) contains supplementary material, which is available to authorized users.

Laser Scan Matching by FAST CVSAC in Dynamic Environment

Localization and mapping are very important for safe movement of robots. One possible way to assist with this functionality is to use laser scan matching. This paper describes a method to implement this functionality. It is based on well-known random sampling and consensus (RANSAC) and iterative closest point (ICP). The proposed algorithm belongs to the class of point to point scan matching approach with its matching criteria rule. The performance of the proposed algorithm is examined in real environment and found applicable in real-time application.

Notes on a Robust Plane Detection Approach in 3D.

This paper addresses the issue of plane detection in 3 dimensional (3D) range images. The identification of planar structures is a crucial task in many visual-aided autonomous robotic applications. The proposed method consists in implementing, in cascade, two algorithms: Random Sample and Consensus (RANSAC) and the more recent Least Entropy-like Estimator (LEL), a nonlinear prediction error estimator that minimizes a cost function inspired by the definition of Gibbs entropy. LEL estimators allow to improve RANSAC performances while maintaining its robustness; kernel density estimation is used to classify data into inliers and outliers. The method has been experimentally applied to 3D images acquired by a Time-Of-Flight camera and compared with a stand alone RANSAC solution. The proposed solution does not require an accurate estimation of the noise variance or outlier scale. This is of fundamental practical importance as the outlier scale, while severely influencing standard RANSAC, is usually unknown a priori and hard to estimate.