Fakulta elektrotechniky a informatiky / Faculty of Electrical Engineering and Informatics
Stálý URI pro tuto komunituhttps://hdl.handle.net/10195/3847
Práce obhájené před rokem 2008 jsou uloženy pouze v kolekci Vysokoškolské kvalifikační práce
Procházet
18 výsledky
Search Results
Článekpeer-reviewedpostprint Omezený přístup Genetic Algorithm-Based Task Assignment for Fleet of Unmanned Surface Vehicles in Dynamically Changing Environment(Taylor & Francis Inc, 2023) Dvořák, Miroslav; Doležel, Petr; Štursa, Dominik; Chouai, MohamedUnmanned vehicles are gaining the attention of professional operators and the general public. The implementation of unmanned vehicles is evident in, among other fields, emergency management, agriculture, traffic monitoring, post-disaster operations, and delivery of goods. Naturally, a group of unmanned vehicles can cooperatively complete operations more proficiently than a single vehicle. However, several issues must be resolved before a stable and reliable group of unmanned vehicles can be generally deployed to solve tasks in civil infrastructures and in industrial facilities. Here, a framework for the guidance of a fleet of unmanned surface vehicles is proposed. The framework utilizes several levels of control, namely Global Planning Level, Local Planning Level, and Low-Level Control. While the individual vehicles are completely autonomous in their operational locomotion and obstacle avoidance (low-level control and local planning), the task assignment for each vehicle (or group of them) is provided by a global planning process, based on the genetic algorithm. The framework provides a concept to solve complex tasks for the fleet of unmanned surface vehicles (USVs). This includes, but is not necessarily limited to, a dynamically changing environment, different types of USVs with special abilities, multiple types of areal restrictions and obstacles, different restrictions for individual USVs, cooperation of multiple USVs to solve their subtasks, energy consumption optimization, etc. The framework can be advantageously applied to tasks such as warehouse logistics, surface maintenance, area exploration, etc. At the end of the study, the application of the framework is presented using a simulated example of cooperative problem solving using six vehicles.Článekpeer-reviewedpostprint Omezený přístup How the Presence of Crystalline Phase Affects Structural Relaxation in Molecular Liquids: The Case of Amorphous Indomethacin(MDPI AG (Multidisciplinary Digital Publishing Institute), 2023) Svoboda, Roman; Pakosta, Marek; Doležel, PetrThe influence of partial crystallinity on the structural relaxation behavior of low-molecular organic glasses is, contrary to, e.g., polymeric materials, a largely unexplored territory. In the present study, differential scanning calorimetry was used to prepare a series of amorphous indomethacin powders crystallized to various extents. The preparations stemmed from the two distinct particle size fractions: 50-125 mu m and 300-500 mu m. The structural relaxation data from the cyclic calorimetric measurements were described in terms of the phenomenological Tool-Narayanaswamy-Moynihan model. For the 300-500 mu m powder, the crystalline phase forming dominantly on the surface led to a monotonous decrease in the glass transition by similar to 6 degrees C in the 0-70% crystallinity range. The activation energy of the relaxation motions and the degree of heterogeneity within the relaxing matrix were not influenced by the increasing crystallinity, while the interconnectivity slightly increased. This behavior was attributed to the release of the quenched-in stresses and to the consequent slight increase in the structural interconnectivity. For the 50-125 mu m powder, distinctly different relaxation dynamics were observed. This leads to a conclusion that the crystalline phase grows throughout the bulk glassy matrix along the internal micro-cracks. At higher crystallinity, a sharp increase in T-g, an increase in interconnectivity, and an increase in the variability of structural units engaged in the relaxation motions were observed.Článekpeer-reviewedpostprint Omezený přístup CSU-Net: Contour Semantic Segmentation Self-Enhancement for Human Head Detection(IEEE (Institute of Electrical and Electronics Engineers), 2023) Chouai, Mohamed; Doležel, PetrThe computer vision community has made tremendous progress in solving a variety of semantic image understanding tasks, such as classification and segmentation. With the advancement of imaging technology and hardware, image semantic segmentation, through the use of deep learning, is among the most common topics which have been worked on in the last decade. However, image semantic segmentation suffers from several drawbacks such as insufficient detection of object boundaries. In this study, we present a new convolutional neural network architecture called CSU-Net that aims to self-enhance the results of semantic segmentation. The proposed model consists of two strongly concatenated encoder-decoder blocks. With this design, we reduced requirements on computing power and memory size to decrease costs and increase the training/prediction speed. This study also demonstrates the advantage of the proposed system for small training data sets. The proposed approach has been implemented on our private dataset, as well as on a publicly available dataset. A comparative analysis was carried out with four popular segmentation models and three other recently introduced architectures to show the efficiency of the proposed system. CSU-Net outperformed the other competing neural networks that we considered for the comparative study. As an example, it succeeded in improving the traditional U-Net result by approximately 50% in mean Intersection over Union (mIoU) for both tested datasets. Based on our experience, the CSU-Net can improve results of semantic segmentation in many applications.Článekpeer-reviewedpostprint (accepted) Otevřený přístup RGB images-driven recognition of grapevine varieties using a densely connected convolutional network(Oxford University Press, 2022) Škrabánek, Pavel; Doležel, Petr; Matousek, RadomilWe present a pocket-size densely connected convolutional network (DenseNet) directed to classification of size-normalized colour images according to varieties of grapes captured in those images. We compare the DenseNet with three established small-size networks in terms of performance, inference time and model size. We propose a data augmentation that we use in training the networks. We train and evaluate the networks on in-field images. The trained networks distinguish between seven grapevine varieties and background, where four and three varieties, respectively, are of red and green grapes. Compared to the established networks, the DenseNet is characterized by near state-of-the-art performance, short inference time and minimal model size. All these aspects qualify the network for real-time, mobile and edge computing applications. The DenseNet opens possibilities for constructing affordable selective harvesters in accordance with agriculture 4.0.Článekpeer-reviewedpublished Otevřený přístup Centroid based person detection using pixelwise prediction of the position(Elsevier Science BV, 2022) Doležel, Petr; Škrabánek, Pavel; Štursa, Dominik; Zanon, Bruno Baruque; Adrian, Hector Cogollos; Kryda, PavelImplementations of person detection in tracking and counting systems tend towards processing of orthogonally captured images on edge computing devices. The ellipse-like shape of heads in orthogonally captured images inspired us to predict head centroids to determine positions of persons in images. We predict the centroids using a fully convolutional network (FCN). We combine the FCN with simple image processing operations to ensure fast inference of the detector. We experiment with the size of the FCN output to further decrease the inference time. We compare the proposed centroid-based detector with bounding box-based detectors on head detection task in terms of the inference time and the detection performance. We propose a performance measure which allows quantitative comparison of the two detection approaches. For the training and evaluation of the detectors, we form original datasets of 8000 annotated images, which are characterized by high variability in terms of lighting conditions, background, image quality, and elevation profile of scenes. We propose an approach which allows simultaneous annotation of the images for both bounding box-based and centroid-based detection. The centroid-based detector shows the best detection performance while keeping edge computing standards.Článekpeer-reviewedpublished Otevřený přístup Sequence of U-Shaped Convolutional Networks for Assessment of Degree of Delamination Around Scribe(Atlantis Press, 2022) Rozsívalová, Veronika; Doležel, Petr; Štursa, Dominik; Rozsíval, PavelThe application of protective layers is the primary method of keeping metallic structures resistant to degradation. The measurement of the layer resistance to delamination is one of the important indicators of the protection quality. Therefore, ISO 4628 standard has been issued to handle and quantify the main coating defects. Here, an innovative assessment of degree of delamination around a scribe according to ISO 4628 standard has been practically realized. It utilizes an computer-driven deep learning-based method. The assessment method is composed of two shallow U-shaped convolutional networks in a row; the first for preliminary and the second for refined detection of delamination area around a scribe. The experiments performed on 586 samples showed that the proposed sequence of U-shaped convolutional networks meets the edge computing standards, provides good generalization capability, and provides precise delamination area detection for a large variability of surfaces.Článekpeer-reviewedpublished version Otevřený přístup Optimization of a Depiction Procedure for an Artificial Intelligence-Based Network Protection System Using a Genetic Algorithm(MDPI, 2021) Doležel, Petr; Holík, Filip; Merta, Jan; Štursa, DominikThe current demand for remote work, remote teaching and video conferencing has brought a surge not only in network traffic, but unfortunately, in the number of attacks as well. Having reliable, safe and secure functionality of various network services has never been more important. Another serious phenomenon that is apparent these days and that must not be discounted is the growing use of artificial intelligence techniques for carrying out network attacks. To combat these attacks, effective protection methods must also utilize artificial intelligence. Hence, we are introducing a specific neural network-based decision procedure that can be considered for application in any flow characteristic-based network-traffic-handling controller. This decision procedure is based on a convolutional neural network that processes the incoming flow characteristics and provides a decision; the procedure can be understood as a firewall rule. The main advantage of this decision procedure is its depiction process, which has the ability to transform the incoming flow characteristics into a graphical structure. Graphical structures are regarded as very efficient data structures for processing by convolutional neural networks. This article's main contribution consists of the development and improvement of the depiction process using a genetic algorithm. The results presented at the end of the article show that the decision procedure using an optimized depiction process brings significant improvements in comparison to previous experiments.Článekpeer-reviewedpublished version Otevřený přístup New End-to-End Strategy Based on DeepLabv3+Semantic Segmentation for Human Head Detection(MDPI, 2021) Chouai, Mohamed; Doležel, Petr; Štursa, Dominik; Němec, ZdeněkIn the field of computer vision, object detection consists of automatically finding objects in images by giving their positions. The most common fields of application are safety systems (pedestrian detection, identification of behavior) and control systems. Another important application is head/person detection, which is the primary material for road safety, rescue, surveillance, etc. In this study, we developed a new approach based on two parallel Deeplapv3+ to improve the performance of the person detection system. For the implementation of our semantic segmentation model, a working methodology with two types of ground truths extracted from the bounding boxes given by the original ground truths was established. The approach has been implemented in our two private datasets as well as in a public dataset. To show the performance of the proposed system, a comparative analysis was carried out on two deep learning semantic segmentation state-of-art models: SegNet and U-Net. By achieving 99.14% of global accuracy, the result demonstrated that the developed strategy could be an efficient way to build a deep neural network model for semantic segmentation. This strategy can be used, not only for the detection of the human head but also be applied in several semantic segmentation applications.Článekpeer-reviewedpublished version Otevřený přístup Memory Efficient Grasping Point Detection of Nontrivial Objects(IEEE (Institute of Electrical and Electronics Engineers), 2021) Doležel, Petr; Štursa, Dominik; Kopecký, Dušan; Jecha, JiříRobotic manipulation with a nontrivial object providing various types of grasping points is of an industrial interest. Here, an efficient method of simultaneous detection of the grasping points is proposed. Specifically, two different 3 degree-of-freedom end effectors are considered for simultaneous grasping. The method utilizes an RGB data-driven perception system based on a specifically designed fully convolutional neural network called attention squeeze parallel U-Net (ASP U-Net). ASP U-Net detects grasping points based on a single RGB image. This image is transformed into a schematic grayscale frame, where the positions and poses of the grasping points are coded into gradient geometric shapes. In order to approve the ASP U-Net architecture, its performance was compared with nine competitive architectures using metrics based on generalized intersection over union and mean absolute error. The results indicate its outstanding accuracy and response time. ASP U-Net is also computationally efficient enough. With a more than acceptable memory size (77 MB), the architecture can be implemented using custom single-board computers. Here, its capabilities were tested and evaluated on the NVIDIA Jetson NANO platform.Článekpeer-reviewedpublished version Otevřený přístup Person Detection for an Orthogonally Placed Monocular Camera(Hindawi limited, 2020) Škrabánek, Pavel; Doležel, Petr; Němec, Zdeněk; Štursa, DominikCounting of passengers entering and exiting means of transport is one of the basic functionalities of passenger flow monitoring systems. Exact numbers of passengers are important in areas such as public transport surveillance, passenger flow prediction, transport planning, and transport vehicle load monitoring. To allow mass utilization of passenger flow monitoring systems, their cost must be low. As the overall price is mainly given by prices of the used sensor and processing unit, we propose the utilization of a visible spectrum camera and data processing algorithms of low time complexity to ensure a low price of the final product. To guarantee the anonymity of passengers, we suggest orthogonal scanning of a scene. As the precision of the counting is relevantly influenced by the precision of passenger recognition, we focus on the development of an appropriate recognition method. We present two opposite approaches which can be used for the passenger recognition in means of transport with and without entrance steps, or with split level flooring. The first approach is the utilization of an appropriate convolutional neural network (ConvNet), which is currently the prevailing approach in computer vision. The second approach is the utilization of histograms of oriented gradients (HOG) features in combination with a support vector machine classifier. This approach is a representative of classical methods. We study both approaches in terms of practical applications, where real-time processing of data is one of the basic assumptions. Specifically, we examine classification performance and time complexity of the approaches for various topologies and settings, respectively. For this purpose, we form and make publicly available a large-scale, class-balanced dataset of labelled RGB images. We demonstrate that, compared to ConvNets, the HOG-based passenger recognition is more suitable for practical applications. For an appropriate setting, it defeats the ConvNets in terms of time complexity while keeping excellent classification performance. To allow verification of theoretical findings, we construct an engineering prototype of the system.