Stereo matching su GPU

La Stereo Vision è un popolare argomento di ricerca nel campo della Visione Artificiale; esso consiste nell’usare due immagini di una stessa scena,prodotte da due fotocamere diverse, per estrarre informazioni in 3D. L’idea di base della Stereo Vision è la simulazione della visione binoculare umana:le due fotocamere sono disposte in orizzontale per fungere da “occhi” che guardano la scena in 3D. Confrontando le due immagini ottenute, si possono ottenere informazioni riguardo alle posizioni degli oggetti della scena.In questa relazione presenteremo un algoritmo di Stereo Vision: si tratta di un algoritmo parallelo che ha come obiettivo di tracciare le linee di livello di un area geografica. L’algoritmo in origine era stato implementato per la Connection Machine CM-2, un supercomputer sviluppato negli anni 80, ed era espresso in *Lisp, un linguaggio derivato dal Lisp e ideato per la macchina stessa. Questa relazione tratta anche la traduzione e l’implementazione dell’algoritmo in CUDA, ovvero un’architettura hardware per l’elaborazione pa- rallela sviluppata da NVIDIA, che consente di eseguire codice parallelo su GPU. Si darà inoltre uno sguardo alle difficoltà che sono state riscontrate nella traduzione da *Lisp a CUDA.

Mitigating non-Lambertian surfaces issues in Stereo Matching with Neural Radiance Fields

Depth estimation from images has long been regarded as a preferable alternative compared to expensive and intrusive active sensors, such as LiDAR and ToF. The topic has attracted the attention of an increasingly wide audience thanks to the great amount of application domains, such as autonomous driving, robotic navigation and 3D reconstruction. Among the various techniques employed for depth estimation, stereo matching is one of the most widespread, owing to its robustness, speed and simplicity in setup. Recent developments has been aided by the abundance of annotated stereo images, which granted to deep learning the opportunity to thrive in a research area where deep networks can reach state-of-the-art sub-pixel precision in most cases. Despite the recent findings, stereo matching still begets many open challenges, two among them being finding pixel correspondences in presence of objects that exhibits a non-Lambertian behaviour and processing high-resolution images. Recently, a novel dataset named Booster, which contains high-resolution stereo pairs featuring a large collection of labeled non-Lambertian objects, has been released. The work shown that training state-of-the-art deep neural network on such data improves the generalization capabilities of these networks also in presence of non-Lambertian surfaces. Regardless being a further step to tackle the aforementioned challenge, Booster includes a rather small number of annotated images, and thus cannot satisfy the intensive training requirements of deep learning. This thesis work aims to investigate novel view synthesis techniques to augment the Booster dataset, with ultimate goal of improving stereo matching reliability in presence of high-resolution images that displays non-Lambertian surfaces.

Fast stereo matching by iterated dynamic programming and quadtree subregioning

The application of energy minimisation methods for stereo matching has been demonstrated to produce high quality disparity maps. However the majority of these methods are known to be computationally expensive, requiring minutes or even hours of computation. We propose a fast minimisation scheme that produces strongly competitive results for significantly reduced computation, requiring only a few seconds of computation. In this paper, we present our iterated dynamic programming algorithm along with a quadtree subregioning process for fast stereo matching.

Aerial video geo-registration using terrain models from dense and coherent stereo matching

In the context of aerial imagery, one of the first steps toward a coherent processing of the information contained in multiple images is geo-registration, which consists in assigning geographic 3D coordinates to the pixels of the image. This enables accurate alignment and geo-positioning of multiple images, detection of moving objects and fusion of data acquired from multiple sensors. To solve this problem there are different approaches that require, in addition to a precise characterization of the camera sensor, high resolution referenced images or terrain elevation models, which are usually not publicly available or out of date. Building upon the idea of developing technology that does not need a reference terrain elevation model, we propose a geo-registration technique that applies variational methods to obtain a dense and coherent surface elevation model that is used to replace the reference model. The surface elevation model is built by interpolation of scattered 3D points, which are obtained in a two-step process following a classical stereo pipeline: first, coherent disparity maps between image pairs of a video sequence are estimated and then image point correspondences are back-projected. The proposed variational method enforces continuity of the disparity map not only along epipolar lines (as done by previous geo-registration techniques) but also across them, in the full 2D image domain. In the experiments, aerial images from synthetic video sequences have been used to validate the proposed technique.

Sharpness Matching in Stereo Images

When stereo images are captured under less than ideal conditions, there may be inconsistencies between the two images in brightness, contrast, blurring, etc. When stereo matching is performed between the images, these variations can greatly reduce the quality of the resulting depth map. In this paper we propose a method for correcting sharpness variations in stereo image pairs which is performed as a pre-processing step to stereo matching. Our method is based on scaling the 2D discrete cosine transform (DCT) coefficients of both images so that the two images have the same amount of energy in each of a set of frequency bands. Experiments show that applying the proposed correction method can greatly improve the disparity map quality when one image in a stereo pair is more blurred than the other.

Sistema di visione stereo su architettura ZYNQ

Lo scopo della tesi è creare un’architettura in FPGA in grado di ricavare informazioni 3D da una coppia di sensori stereo. La pipeline è stata realizzata utilizzando il System-on-Chip Zynq, che permette una stretta interazione tra la parte hardware realizzata in FPGA e la CPU. Dopo uno studio preliminare degli strumenti hardware e software, è stata realizzata l’architettura base per la scrittura e la lettura di immagini nella memoria DDR dello Zynq. In seguito l’attenzione si è spostata sull’implementazione di algoritmi stereo (rettificazione e stereo matching) su FPGA e nella realizzazione di una pipeline in grado di ricavare accurate mappe di disparità in tempo reale acquisendo le immagini da una camera stereo.

Interocular suppression is gated by interocular feature matching

We present a new form of contrast masking in which the target is a patch of low spatial frequency grating (0.46 c/deg) and the mask is a dark thin ring that surrounds the centre of the target patch. In matching and detection experiments we found little or no effect for binocular presentation of mask and test stimuli. But when mask and test were presented briefly (33 or 200 ms) to different eyes (dichoptic presentation), masking was substantial. In a 'half-binocular' condition the test stimulus was presented to one eye, but the mask stimulus was presented to both eyes with zero-disparity. This produced masking effects intermediate to those found in dichoptic and full-binocular conditions. We suggest that interocular feature matching can attenuate the potency of interocular suppression, but unlike in previous work (McKee, S. P., Bravo, M. J., Taylor, D. G., & Legge, G. E. (1994) Stereo matching precedes dichoptic masking. Vision Research, 34, 1047) we do not invoke a special role for depth perception. © 2004 Elsevier Ltd. All rights reserved.

Performance Evaluation and Calibration of Machine Vision Measurement

This thesis presents the calibration and comparison of two systems, a machine vision system that uses 3 channel RGB images and a line scanning spectral system. Calibration. is the process of checking and adjusting the accuracy of a measuring instrument by comparing it with standards. For the RGB system self-calibrating methods for finding various parameters of the imaging device were developed. Color calibration was done and the colors produced by the system were compared to the known colors values of the target. Software drivers for the Sony Robot were also developed and a mechanical part to connect a camera to the robot was also designed. For the line scanning spectral system, methods for the calibrating the alignment of the system and the measurement of the dimensions of the line scanned by the system were developed. Color calibration of the spectral system is also presented.

La représentation de la machine dans la trilogie d’anticipation scientifique d’Albert Robida : du texte à l’image et de l’image au texte

Cette thèse étudie la représentation de la machine chez Robida. La partie centrale de notre recherche s’intéresse à révéler ses significations et interroge sa mise en scène littéraire et visuelle dans chacun des romans de la trilogie d’anticipation scientifique la plus connue de l’auteur-illustrateur. La quête se transforme en un voyage continu entre le lisible et le visible, le dit et le non-dit, la description littéraire et l’imagination, la réalité et la fiction. Nous nous intéressons à l’évolution de la vision de Robida : dans Le Vingtième siècle, l’image de la machine bienfaisante, facilitant la vie de l’homme, économisant du temps et de l’argent, et contribuant largement à son bonheur et à son divertissement, à part quelques accidents très limités, se traduit par une complémentarité avantageuse entre le texte d’une part et les vignettes, les tableaux et les hors-textes se trouvant dans le récit, d’autre part. Celle-ci se transforme, dans La Guerre au vingtième siècle, en une inquiétude vis-à-vis de l’instrumentalisation de la machine pour la guerre, qui s’exprime par une projection de la narration vers l’illustration in-texte, et sensibilise le lecteur en montrant le caractère violent et offensif d’appareils uniquement nommés. Celle-ci devient finalement, dans La Vie électrique, synonyme d’un pessimisme total quant à l’implication de la machine dans la société et à la puissance du savoir scientifique dans l’avenir, qui s’affiche dans des hors-textes sombres et maussades. Dans ce cadre, la machine illustrée exige une lecture iconotextuelle, une importance accordée au détail, aux éléments présents ou absents, aux modalités de passage d’un mode de présentation à l’autre, à la place anticipée ou tardive de l’illustration, au rapport entre le texte, le dessin et sa légende, aux mots qui migrent vers le dessin et surtout au reste du décor incomplet. Chez Robida, les louanges qui passent à la critique et l’humour qui se fait cynisme, sont assez représentatifs des espoirs et des craintes suscités par la découverte et la mise en application de l’électricité, par ses vertus, mais aussi par son aspect incontrôlable.

Correspondência estéreo usando imagens em multiresolução com profundidade variável

We propose a multi-resolution, coarse-to-fine approach for stereo matching, where the first matching happens at a different depth for each pixel. The proposed technique has the potential of attenuating several problems faced by the constant depth algorithm, making it possible to reduce the number of errors or the number of comparations needed to get equivalent results. Several experiments were performed to demonstrate the method efficiency, including comparison with the traditional plain correlation technique, where the multi-resolution matching with variable depth, proposed here, generated better results with a smaller processing time

Uma proposição para o cálculo de mapas de disparidade de imagens estéreo usando um interpolador neural baseado em funções de base radial

This study aims to seek a more viable alternative for the calculation of differences in images of stereo vision, using a factor that reduces heel the amount of points that are considered on the captured image, and a network neural-based radial basis functions to interpolate the results. The objective to be achieved is to produce an approximate picture of disparities using algorithms with low computational cost, unlike the classical algorithms

Methodologies for visual correspondence

Visual correspondence is a key computer vision task that aims at identifying projections of the same 3D point into images taken either from different viewpoints or at different time instances. This task has been the subject of intense research activities in the last years in scenarios such as object recognition, motion detection, stereo vision, pattern matching, image registration. The approaches proposed in literature typically aim at improving the state of the art by increasing the reliability, the accuracy or the computational efficiency of visual correspondence algorithms. The research work carried out during the Ph.D. course and presented in this dissertation deals with three specific visual correspondence problems: fast pattern matching, stereo correspondence and robust image matching. The dissertation presents original contributions to the theory of visual correspondence, as well as applications dealing with 3D reconstruction and multi-view video surveillance.

Metodologie di apprendimento automatico applicate alla generazione di dati 3D

Il framework in oggetto, è un ambiente ideato con lo scopo di applicare tecniche di Machine Learning (in particolare le Random Forest) alle funzionalità dell'algoritmo di stereo matching SGM (Semi Global Matching), al fine di incrementarne l'accuratezza in versione standard. Scopo della presente tesi è quello di modificare alcune impostazioni di tale framework rendendolo un ambiente che meglio si adatti alla direzionalità delle scanline (introducendo finestre di supporto rettangolari e ortogonali e il training di foreste separate in base alla singola scanline) e ampliarne le funzionalità tramite l'aggiunta di alcune nuove feature, quali la distanza dal più vicino edge direzionale e la distintività calcolate sulle immagini Left della stereo pair e gli edge direzionali sulle mappe di disparità. Il fine ultimo sarà quello di eseguire svariati test sui dataset Middlebury 2014 e KITTI e raccogliere dati che descrivano l'andamento in positivo o negativo delle modifiche effettuate.

On precise three-dimensional environment modeling via UAV-based photogrammetric systems

Abstract : Images acquired from unmanned aerial vehicles (UAVs) can provide data with unprecedented spatial and temporal resolution for three-dimensional (3D) modeling. Solutions developed for this purpose are mainly operating based on photogrammetry concepts, namely UAV-Photogrammetry Systems (UAV-PS). Such systems are used in applications where both geospatial and visual information of the environment is required. These applications include, but are not limited to, natural resource management such as precision agriculture, military and police-related services such as traffic-law enforcement, precision engineering such as infrastructure inspection, and health services such as epidemic emergency management. UAV-photogrammetry systems can be differentiated based on their spatial characteristics in terms of accuracy and resolution. That is some applications, such as precision engineering, require high-resolution and high-accuracy information of the environment (e.g. 3D modeling with less than one centimeter accuracy and resolution). In other applications, lower levels of accuracy might be sufficient, (e.g. wildlife management needing few decimeters of resolution). However, even in those applications, the specific characteristics of UAV-PSs should be well considered in the steps of both system development and application in order to yield satisfying results. In this regard, this thesis presents a comprehensive review of the applications of unmanned aerial imagery, where the objective was to determine the challenges that remote-sensing applications of UAV systems currently face. This review also allowed recognizing the specific characteristics and requirements of UAV-PSs, which are mostly ignored or not thoroughly assessed in recent studies. Accordingly, the focus of the first part of this thesis is on exploring the methodological and experimental aspects of implementing a UAV-PS. The developed system was extensively evaluated for precise modeling of an open-pit gravel mine and performing volumetric-change measurements. This application was selected for two main reasons. Firstly, this case study provided a challenging environment for 3D modeling, in terms of scale changes, terrain relief variations as well as structure and texture diversities. Secondly, open-pit-mine monitoring demands high levels of accuracy, which justifies our efforts to improve the developed UAV-PS to its maximum capacities. The hardware of the system consisted of an electric-powered helicopter, a high-resolution digital camera, and an inertial navigation system. The software of the system included the in-house programs specifically designed for camera calibration, platform calibration, system integration, onboard data acquisition, flight planning and ground control point (GCP) detection. The detailed features of the system are discussed in the thesis, and solutions are proposed in order to enhance the system and its photogrammetric outputs. The accuracy of the results was evaluated under various mapping conditions, including direct georeferencing and indirect georeferencing with different numbers, distributions and types of ground control points. Additionally, the effects of imaging configuration and network stability on modeling accuracy were assessed. The second part of this thesis concentrates on improving the techniques of sparse and dense reconstruction. The proposed solutions are alternatives to traditional aerial photogrammetry techniques, properly adapted to specific characteristics of unmanned, low-altitude imagery. Firstly, a method was developed for robust sparse matching and epipolar-geometry estimation. The main achievement of this method was its capacity to handle a very high percentage of outliers (errors among corresponding points) with remarkable computational efficiency (compared to the state-of-the-art techniques). Secondly, a block bundle adjustment (BBA) strategy was proposed based on the integration of intrinsic camera calibration parameters as pseudo-observations to Gauss-Helmert model. The principal advantage of this strategy was controlling the adverse effect of unstable imaging networks and noisy image observations on the accuracy of self-calibration. The sparse implementation of this strategy was also performed, which allowed its application to data sets containing a lot of tie points. Finally, the concepts of intrinsic curves were revisited for dense stereo matching. The proposed technique could achieve a high level of accuracy and efficiency by searching only through a small fraction of the whole disparity search space as well as internally handling occlusions and matching ambiguities. These photogrammetric solutions were extensively tested using synthetic data, close-range images and the images acquired from the gravel-pit mine. Achieving absolute 3D mapping accuracy of 11±7 mm illustrated the success of this system for high-precision modeling of the environment.

Convolutional Neural Network Architectures for Template Matching

The Neural Networks customized and tested in this thesis (WaldoNet, FlowNet and PatchNet) are a first exploration and approach to the Template Matching task. The possibilities of extension are therefore many and some are proposed below. During my thesis, I have analyzed the functioning of the classical algorithms and adapted with deep learning algorithms. The features extracted from both the template and the query images resemble the keypoints of the SIFT algorithm. Then, instead of similarity function or keypoints matching, WaldoNet and PatchNet use the convolutional layer to compare the features, while FlowNet uses the correlational layer. In addition, I have identified the major challenges of the Template Matching task (affine/non-affine transformations, intensity changes...) and solved them with a careful design of the dataset.