|
|
Structural Methods of Objects Identification for Industrial Robot Operation
Minařík, Martin ; Šlapal, Josef (referee) ; Konečný, Vladimír (referee) ; Šťastný, Jiří (advisor)
This PhD thesis deals with the use of structural methods of objects identification for industrial robots operation. First, the present state of knowledge in the field is described, i.e. the whole process of objects recognition with the aid of common methods of the syntactic analysis. The main disadvantage of these methods is that is impossible to recognize objects whose digitalized image is corrupted in some ways (due to excessive noise or image disturbances), objects are therefore deformed. Further, other methods for the recognition of deformed objects are described. These methods use structural description of objects for object recognition, i.e. methods which determine the distance between attribute descriptions of images. The core part of this PhD thesis begins in Chapter 5, where deformation grammars, capable of description of all possible object deformations, are described. The only complication in the analysis is the ambiguity of the deformation grammar, which lowers the effectiveness of the analysis. Further, PhD thesis deals with the selection and modification of a proper parser, which is able to analyze a deformation grammar effectively. Three parsers are described: the modified Earley parser, the modified Tomita parser and the modified hybrid LRE(k) parser. As for the modified Earley’s parser, ways of its effective implementation are described. One of the necessary parts of the object recognition is providing the invariances, which this PhD thesis covers in detail, too. Finally, the results of described algorithms are mentioned (successfulness and speed of deformed objects recognition) and suggested testing environment and implemented algorithms are described. In conclusion, all determined possibilities of deformation grammars and their results are summarized.
|
|
|
Methods of Segmentation and Identification of Deformed Vertebrae in 3D CT Data of Oncological Patients
Jakubíček, Roman ; Flusser, Jan (referee) ; Kozubek, Michal (referee) ; Jan, Jiří (advisor)
In this doctoral thesis, the design of algorithms enabling the implementation of a fully automatic system for vertebrae segmentation in 3D computed tomography (CT) image data of possibly incomplete spines, in patients with bone metastases and vertebral compressions is presented. The proposed algorithm consists of several fundamental problems: spine detection and its axis determination, individual vertebra localization and identification (labeling), and finally, precise segmentation of vertebrae. The detection of the spine, specifically identifying its ends, and determining the course of the spinal canal, combines several advanced methods, including deep learning-based approaches. A novel growing circle method has been designed for tracing the spinal cord canal. Further, the innovative spatially variant filtering of brightness profiles along the spine axis leading to intervertebral disc localization has been proposed and implemented. The discs thus obtained are subsequently identified via comparing the tested vertebrae and model of vertebrae provided by a machine-learning process and optimized by dynamic programming. The final vertebrae segmentation is provided by the deformation of the complete-spine intensity model, utilizing a proposed multilevel registration technique. The complete proposed algorithm has been validated on testing databases, including also publicly available datasets. This way, it has been proven that the newly proposed algorithms provide results at least comparable to other author’s algorithms, and in some cases, even better. The main strengths of the algorithms lie in high reliability of the results and in the robustness to even strongly distorted vertebrae of oncological patients and to the occurrence of artifacts in data; moreover, they are capable of identifying the vertebra labels even in incomplete spinal CT scans. The strength is also in the complete automation of the processing and in its relatively low computational complexity enabling implementation on standard PC hardware. The system for fully automatic localization and labeling of distorted vertebrae in possibly incomplete spinal CT data is presented in this doctoral thesis. The design of algorithms enabling the implementation utilizes several novel approaches, which were presented at international conferences and published in the journal Jakubicek et al. (2020). Based on the results of the experimental validation, the proposed algorithms seem to be routinely usable and capable of providing fully acceptable input data (identified and precisely segmented vertebrae) as needed in the subsequent automatic spine bone lesion analysis.
|
|
|
Robot vision with industrial robots Kuka
Krutílek, Jan ; Pochylý, Aleš (referee) ; Kubela, Tomáš (advisor)
Diploma thesis deals with a robot vision and its application to the problem of manipulation of coincidentally placed objects. There is mentioned an overview of current principles of the most frequently used vision systems on the market. With regard to the required task to be solved, there are mentioned various possibilities of using basic softsensors during the recognition of different objects. The objective of this Diploma thesis is also programming and realization of a demonstration application applying knowledge of PLC programming, knowledge of expert programming KRL language (for KUKA robots), knowledge of designing scripts for smart camera in Spectation software and knowledge of network communication among all devices used in this case.
|
|
|
Design of a vision system with Kuka robot
Rusnák, Jakub ; Pochylý, Aleš (referee) ; Kubela, Tomáš (advisor)
Diploma thesis deals with applications of vision system with KUKA robot in field of identification and sorting bigger amount of different objects. Introductory and theoretical part of the thesis describes present situation on industrial vision systems market and their usage. Diploma thesis include practical application of object (coin) recognition with SICK IVC 2D vision system and their sorting by industrial robot KUKA KR 3. Application is also concerned with network communication between camera and robot via PLC, programming in KRL language and programm for object recognition in IVC Studio.
|
|
|
Synthetic data generator aimed at development of drone detectors
Zlatníčková, Marie ; Dobrovský, Ladislav (referee) ; Škrabánek, Pavel (advisor)
This diploma thesis deals with the issue of creating images of realistic-looking images from 3D models of drones. The search section of this thesis explains the basic concepts in digital image processing and the use of neural networks in the detection and recognition of objects in the image. The practical part of this work deals with the implementation of a software solution that creates tagged colored images from digital 3D drone models. These images can contain one or more drones in different flight phases, with different light, rotation or blur.
|
|
|
Methods of Segmentation and Identification of Deformed Vertebrae in 3D CT Data of Oncological Patients
Jakubíček, Roman ; Flusser, Jan (referee) ; Kozubek, Michal (referee) ; Jan, Jiří (advisor)
In this doctoral thesis, the design of algorithms enabling the implementation of a fully automatic system for vertebrae segmentation in 3D computed tomography (CT) image data of possibly incomplete spines, in patients with bone metastases and vertebral compressions is presented. The proposed algorithm consists of several fundamental problems: spine detection and its axis determination, individual vertebra localization and identification (labeling), and finally, precise segmentation of vertebrae. The detection of the spine, specifically identifying its ends, and determining the course of the spinal canal, combines several advanced methods, including deep learning-based approaches. A novel growing circle method has been designed for tracing the spinal cord canal. Further, the innovative spatially variant filtering of brightness profiles along the spine axis leading to intervertebral disc localization has been proposed and implemented. The discs thus obtained are subsequently identified via comparing the tested vertebrae and model of vertebrae provided by a machine-learning process and optimized by dynamic programming. The final vertebrae segmentation is provided by the deformation of the complete-spine intensity model, utilizing a proposed multilevel registration technique. The complete proposed algorithm has been validated on testing databases, including also publicly available datasets. This way, it has been proven that the newly proposed algorithms provide results at least comparable to other author’s algorithms, and in some cases, even better. The main strengths of the algorithms lie in high reliability of the results and in the robustness to even strongly distorted vertebrae of oncological patients and to the occurrence of artifacts in data; moreover, they are capable of identifying the vertebra labels even in incomplete spinal CT scans. The strength is also in the complete automation of the processing and in its relatively low computational complexity enabling implementation on standard PC hardware. The system for fully automatic localization and labeling of distorted vertebrae in possibly incomplete spinal CT data is presented in this doctoral thesis. The design of algorithms enabling the implementation utilizes several novel approaches, which were presented at international conferences and published in the journal Jakubicek et al. (2020). Based on the results of the experimental validation, the proposed algorithms seem to be routinely usable and capable of providing fully acceptable input data (identified and precisely segmented vertebrae) as needed in the subsequent automatic spine bone lesion analysis.
|
|
|
Applications of Machine Learning for Detecting and Counting Objects in Cell Biology
Brázdilová, Květa ; Stopka, Pavel (advisor) ; Hoksza, David (referee)
Modern biological research generates large amounts of data, which require automation for efficient analysis. Lately, machine learning solutions are being developed for many of the problems in this field. This thesis focuses on applications of machine learning for image analysis, such as detecting cells in microscopy images and classifying them based on their phenotype. After a brief introduction to machine learning concepts, eight published methods are presented, which employ machine learning for either detecting and classifying, or counting objects in biological images. Five open-source software tools for biological image analysis, which employ some of the methods mentioned above, are introduced. A new project is also described, which aims to create a convolutional neural network for counting bacterial colonies in images of agar plates. The results of this project are discussed. Keywords: machine learning, neural network, pattern recognition, cell biology, segmentation
|
|
|
Object recognition using 3D convolutional neural networks
Moravec, Jaroslav ; Lokoč, Jakub (advisor) ; Straka, Milan (referee)
Title: Object recognition using 3D convolutional neural networks Author: Jaroslav Moravec Department: Department of Software Engineering Supervisor: RNDr. Jakub Lokoč, Ph.D., Department of Software Engineering Abstract: With the fast development of laser and sensor technologies, it has become easy to scan a real-world object and save it in a digital format into a persistent database. With the rising number of scanned 3D objects, data man- agement and retrieval methods become necessary. For various retrieval tasks, effective retrieval models are required. In our work, we focus on effective classifi- cation and similarity search. The investigated approach is based on convolutional neural networks representing a machine learning method that boomed in recent years. We have designed and trained several architectures of 3D convolutional neural networks and tested them on state-of-the-art benchmark 3D datasets for 3D object recognition and retrieval. We were also able to show that the trained features on one dataset can be then used to predict class labels on another 3D dataset. Keywords: Object recognition, 3D convolution, neural networks
|
|
|
The suitable neural network topology design for object recognition in the grass.
Polívka, Tomáš ; Pavlíček, Josef (advisor) ; Hanzlík, Petr (referee)
This thesis describes testing of artificial neural network depending on learning set for object recognition in grass. Values of experimenting with different topologies in Neuroph Studio were recorded. Testing was conducted on three sets of learning basic shapes. Verification of the ability of learning networks were tested using four basic shapes captured in the grass (square, rectangle, circle and triangle). When evaluating the most appropriate topology was defined by two main criteria. Number of successful recognition of shapes and distribution by recognized shapes.
The theoretical part describes the functionality and features of neural networks, image recognition using edge detection and existing applications for recognition of plant species.
|
|
|
Object Detection Based on Edges
Caha, Jaroslav ; Švub, Miroslav (referee) ; Španěl, Michal (advisor)
This work presents a door detection method in images for mobile robot navigation. The method is able to detect doors in an input picture on the basis of found image edges. It is important to distinguish the door from similar objects like windows, paintings, or floor patterns. Therefore, the picture is divided into more parts (a floor, a wall, a ceiling) so that the potential placement of the door can be better drawn.
|
guest ::
