|
|
Acquisition of inputs by image processing for controlling an autonomous vehicle
Midrla, Daniel ; Píštěk, Václav (referee) ; Kučera, Pavel (advisor)
This master’s thesis deals with data acquisition by image processing in order to control an autonomous vehicle. Firstly, the thesis offers a summary of theoretical knowledge relevant to the given topic. Then follows a description of creating an algorithm, which acquires basic inputs for autonomous vehicle control with the use of a stereo camera and an object detection neural network. The inputs gained from this algorithm are the class of the detected object and its distance. Finally, an experimental evaluation of the correct functionality is performed with an emphasis on optimizing the accuracy and range of the distance computation. An assessment of the ability to deploy the created algorithm in real time on a compact computer with limited computing power is also performed.
|
|
|
3D Map Building Based on Stereo Vision for Mobile Robot Navigation
Babinec, Adam ; Orság, Filip (referee) ; Herman, David (advisor)
This thesis is dedicated to the subject of passive stereo vision in modern robotics. The work includes the design and implementation of autonomous passive binocular stereo vision system for mobile robot navigation. A three-dimensional local map is built up by aggregation of point clouds created by reprojection of the image pairs taken with stereocamera. The image pairs are reprojected using disparity map obtained with application of block matching algorithms on the image pairs. The local map is represented by voxel grid stored in an octree and it supports detection of moving obstacles on ray-casting principle. Visual odometry is calculated by tracking reprojected paired image features detected in series of image pairs. The system allows user to choose different approaches to the problem solving, it is platform and hardware-independent and provides graphic user interface.
|
|
|
Differentiable Depth Estimation for Bin Picking
Černý, Marek ; Klusáček, David (advisor) ; Šikudová, Elena (referee)
The goal of this thesis was to investigate the neural 3D surface reconstruction from multiple views with the intent to use the resulting depth maps for bin picking. Survey of papers from 2014 to 2018 showed that none of the state of the art methods would be used to control a robot arm in our setup. Therefore we decided to create our low-level neural approach which we called the EmfNet. The network is based on a pyramidal resolution refining approach. At each pyramid's layer, there are three separate networks that take part in the computation. Each of them has a definite goal, which gives us almost complete understanding of what is going on inside the network. The EmfNet model was partially usable, but we nevertheless extended it to EmfNet-v2. First, another measuring layer was added, which freed EmfNet from depending on an unnecessary hyperparameter. Second, we used constraints on geometry for the network not to be confused by occlusions (cases where a certain part of the surface is visible only from a single camera). Both networks were implemented and tested on a corpus that was created as a part of this thesis. A corpus containing rendered as well as real data. The process of correspondence pairing inside the network can be observed using the visualization tool. We designed a way how to use a robotic arm...
|
|
|
Acquisition of inputs by image processing for controlling an autonomous vehicle
Midrla, Daniel ; Píštěk, Václav (referee) ; Kučera, Pavel (advisor)
This master’s thesis deals with data acquisition by image processing in order to control an autonomous vehicle. Firstly, the thesis offers a summary of theoretical knowledge relevant to the given topic. Then follows a description of creating an algorithm, which acquires basic inputs for autonomous vehicle control with the use of a stereo camera and an object detection neural network. The inputs gained from this algorithm are the class of the detected object and its distance. Finally, an experimental evaluation of the correct functionality is performed with an emphasis on optimizing the accuracy and range of the distance computation. An assessment of the ability to deploy the created algorithm in real time on a compact computer with limited computing power is also performed.
|
|
|
Differentiable Depth Estimation for Bin Picking
Černý, Marek ; Klusáček, David (advisor) ; Šikudová, Elena (referee)
The goal of this thesis was to investigate the neural 3D surface reconstruction from multiple views with the intent to use the resulting depth maps for bin picking. Survey of papers from 2014 to 2018 showed that none of the state of the art methods would be used to control a robot arm in our setup. Therefore we decided to create our low-level neural approach which we called the EmfNet. The network is based on a pyramidal resolution refining approach. At each pyramid's layer, there are three separate networks that take part in the computation. Each of them has a definite goal, which gives us almost complete understanding of what is going on inside the network. The EmfNet model was partially usable, but we nevertheless extended it to EmfNet-v2. First, another measuring layer was added, which freed EmfNet from depending on an unnecessary hyperparameter. Second, we used constraints on geometry for the network not to be confused by occlusions (cases where a certain part of the surface is visible only from a single camera). Both networks were implemented and tested on a corpus that was created as a part of this thesis. A corpus containing rendered as well as real data. The process of correspondence pairing inside the network can be observed using the visualization tool. We designed a way how to use a robotic arm...
|
|
|
3D Map Building Based on Stereo Vision for Mobile Robot Navigation
Babinec, Adam ; Orság, Filip (referee) ; Herman, David (advisor)
This thesis is dedicated to the subject of passive stereo vision in modern robotics. The work includes the design and implementation of autonomous passive binocular stereo vision system for mobile robot navigation. A three-dimensional local map is built up by aggregation of point clouds created by reprojection of the image pairs taken with stereocamera. The image pairs are reprojected using disparity map obtained with application of block matching algorithms on the image pairs. The local map is represented by voxel grid stored in an octree and it supports detection of moving obstacles on ray-casting principle. Visual odometry is calculated by tracking reprojected paired image features detected in series of image pairs. The system allows user to choose different approaches to the problem solving, it is platform and hardware-independent and provides graphic user interface.
|
|
|
Space in photography and image
Šimeček, Michal
Space perception in pictures or photos is different from space perception of real scenes. The differences are significant but obviously unsuspected. The space perception in pictures and photos is allowed by our experience with many realistic pictures and photos. The effect of relations between picture maker (camera) and observer are important. For example complexity is conducive to uncertainty to estimate egocentric distances in the picture. The space perception is modified by properties of picture (e.g. picture size) or observer’s position.
|
|
|
Visual space perception: Introduction
Šikl, Radovan ; Šimeček, Michal
The topic visual space perception is widely investigated for a long time. During the history, it was of the principle interest for philosophy, visual arts, geometry, optics, physiology and other disciplines. Also, in the times beginning of psychology as a science, the research was often concerned with the questions related to depth perception. In this paper, we introduce the principle questions studied in the field together with the experimental methods used and characteristic results. Finally, we present our own research focused on trade-off between particular spatial descriptors and on selected aspects of the relationship between visual and physical space.
|
|
|
Issues in the research of visual space perception: The geometry of visual space, relationship between the spatial descriptors, and the metric vs. ordinal data
Šimeček, Michal ; Šikl, Radovan
In the article, the authors summarize the main results of their research activity in the field of visual space perception. In the particular experimental projects, they focused on the relationship between the partial spatial descriptors in the process of forming the complex three-dimensional percept. Also, the geometry of the optical space was dealt in with respect to the Euclidean geometry of the physical space. The principle finding concerns the notion of anisotropic curvature of visual space. Finally, the question of plausibility of experiments requiring the perceptual judgements as expressed in metrical units (Euclidean tasks) was raised. The reason for this approach being used is to gather data directly comparable with the physical properties of space. On the other hand, in Euclidean tasks, the ecological validity of knowledge gained is inevitably limited and the danger of intrusion of nonperceptual factors is pervasive.
|
guest ::
