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
29 výsledky
Search Results
Konferenční objektpeer-reviewedpostprint Otevřený přístup Comparison of ReLU and linear saturated activation functions in neural network for universal approximation(IEEE (Institute of Electrical and Electronics Engineers), 2019) Štursa, Dominik; Doležel, PetrActivation functions used in hidden layers directly affect the possibilities for describing nonlinear systems using a feedforward neural network. Furthermore, linear based activation functions are less computationally demanding than their nonlinear alternatives. In addition, feedforward neural networks with linear based activation functions can be advantageously used for control of nonlinear systems, as shown in previous authors' publications. This paper aims to compare two types of linear based functions - symmetric linear saturated function and the rectifier linear unit (ReLU) function as activation functions of the feedforward neural network used for a nonlinear system approximation. Topologies with one hidden layer and the combination of defined quantities of hidden layer neurons in the feedforward neural network are used. Strict criteria are applied for the conditions of the experiments; specifically, the Levenberg-Marquardt algorithm is applied as a training algorithm and the Nguyen-Widrow algorithm is used for the weights and biases initialization. Three benchmark systems are then selected as nonlinear plants for approximation, which should serve as a repeatable source of data for testing. The training data are acquired by the computation of the output as a reaction to a specified colored input signal. The comparison is based on the convergence speed of the training for a fixed value of the error function, and also on the performance over a constant number of epochs. At the end of the experiments, only small differences between the performance of both applied activation functions are observed. Although the symmetric linear saturated activation function provides the lesser median of the final error function value across the all tested numbers of neurons in topologies, the ReLU function seems to be also capable of use as the activation function for nonlinear system modeling.Konferenční objektpeer-reviewedpostprint Otevřený přístup Convolutional Neural Network for Sound Processing - Study of Deployed Application(IEEE (Institute of Electrical and Electronics Engineers), 2019) Doležel, Petr; Štursa, Dominik; Honc, DanielPest birds are considered as a special kind of vermin, since, in most of countries, their legal position does not enable their direct extermination. Therefore, in order to protect the agricultural areas indirectly from pest birds, the robust and highly selective pest bird sensor is necessary to design. In this contribution, the pest bird detection unit, based on a convolutional neural network, is presented. The convolutional neural network itself is used for the decision making about the pest bird occurrence, while sound recordings are used as input data. The testings, presented at the end of the contribution, proved a very high accuracy of the detection unit, with the results indispensably improved in comparison to previously presented approaches.Konferenční objektpeer-reviewedpostprint Otevřený přístup On possibilities of human head detection for person flow monitoring system(Springer Nature Switzerland AG, 2019) Doležel, Petr; Štursa, Dominik; Škrabánek, PavelAlong with the development of human society, economy, industry and engineering, as well as with growing population in the world's biggest cities, various approaches to person detection have become the subject of great interest. One approach to developing a person detection system is proposed in this paper. A high-angle video sequence is considered as the input to the system. Then, three classification algorithms are considered: support vector machines, pattern recognition neural networks and convolutional neural networks. The results showed very little difference between the classifiers, with the overall accuracy more than 95% over a testing set.Konferenční objektpeer-reviewedpostprint Otevřený přístup Bin Picking Success Rate Depending on Sensor Sensitivity(IEEE (Institute of Electrical and Electronics Engineers), 2019) Doležel, Petr; Pidanič, Jan; Zálabský, Tomáš; Dvořák, MiroslavThe goal of this contribution is to determine correlation between an applied sensor for object registration and the success rate of the bin-picking problem. In most applications of a bin picking problem in industry, the procedure consists of two consecutive steps. The first step provides an initial guess of both position and rotation angle of the object to be registered, while the second one improves the exact pose accuracy, as required in following tasks. The second step can be, among others, implemented by the Iterative Closest Point Algorithm (ICP). It is well known that the ICP algorithm is very sensitive to the initial guess of the position and rotation angle of the object. Another interesting feature, especially from the technician’s point of view, is the sensitivity of the ICP algorithm in relation to the applied sensor. Therefore, one particular bin picking application, involving complex irregular objects, is examined in this paper. Various kinds of sensors for 3D scene reconstruction are employed and, as a result of this contribution, a comprehensive set of relations between sensor quality and the ICP algorithm sensitivity is formulated.Konferenční objektpeer-reviewedpostprint Otevřený přístup Comparison Two of Different Technologies for Outdoor Positioning of Robotic Vehicles(Springer Nature, 2019) Dvořák, Miroslav; Doležel, PetrThis paper aims to compare two different technologies, which can determine the exact position of a robotic vehicle. The first method uses wireless technology and is based on the measurement of the signal strength of the bluetooth beacons. Based on these values, you can calculate the distance from beacons. The second method uses laser light and measurement of the reflected pulses. Based on the reference points of reflection, we can determine the distance. Both methods then use 2D triangulation to determine the position of the robotic vehicle. The exact position of the bluetooth beacons or the reference points must be known for the calculation. The paper also describes experiments with a laser, and the conclusion provides an evaluation of both technologies.Konferenční objektpeer-reviewedpostprint Omezený přístup Predictive Controller Based on Feedforward Neural Network with Rectified Linear Units(Springer Nature Switzerland AG, 2019) Doležel, Petr; Honc, Daniel; Štursa, DominikThis paper deals with a nonlinear Model Predictive Control with a special form of the process model. Controller uses for the prediction purposes a locally valid linear sub-models. The sub-models are obtained from a neural model with the rectifier activation function in hidden neurons. Simulation example is given to demonstrate proposed solution - neural model design and predictive controller application.Konferenční objektpeer-reviewedpostprint Omezený přístup Neural Network for Smart Adjustment of Industrial Camera - Study of Deployed Application(Springer, 2018) Doležel, Petr; Honc, DanielSince machine vision is gaining more and more interest lately, it is necessary to deal with correct approaches to visual data acquisition in industry. As a particular part of this complex problematics, a technique for the industrial camera exposure time and image sensor gain tuning is presented in this contribution. In comparison to other approaches, a human expert photographer is used instead of explicitly defined cost function. His knowledge is transformed into an artificial expert system represented by a feedforward neural network. The expert system then provides the suitable exposure time and image sensor gain to gather sharp and balanced images.Konferenční objektpeer-reviewedpostprint Omezený přístup Thermal Process Control Using Neural Model and Genetic Algorithm(Springer Nature Switzerland AG, 2019) Honc, Daniel; Doležel, Petr; Merta, JanPredictive Controller of a laboratory thermal process is presented in the paper. Process model is approximated by a neural network. On-line optimization is done by a genetic algorithm. Control algorithm is tested on the laboratory thermal process and compared to the standard control methods like predictive controller with the transfer and state-space linear model and the quadratic programming optimization method or a PI controller.Článekpeer-reviewedpostprint Omezený přístup On Reporting performance of binary classifiers(2017) Škrabánek, Pavel; Doležel, PetrIn this contribution, the question of reporting performance of binary classifiers is opened in context of the so called class imbalance problem. The class imbalance problem arises when a dataset with a highly imbalanced class distribution is used within the training or evaluation process. In such cases, only measures, which are not biased by distribution of classes in datasets, should be used; however, they cannot be chosen arbitrarily. They should be selected so that their outcomes provide desired information; and simultaneously, they should allow a full comparison of just evaluated classifier performance along, with performances of other solutions. As is shown in this article, the dilemma with reporting performance of binary classifiers can be solved using so called class balanced measures. The class balanced measures are generally applicable means, appropriate for reporting performance of binary classifiers on balanced as well as on imbalanced datasets. On the basis of the presented pieces of information, a suggestion for a generally applicable, fully-valued, reporting of binary classifiers performance is given.Konferenční objektpeer-reviewedpostprint Otevřený přístup An IoT approach to positioning of a robotic vehicle(Springer International Publishing AG, 2018) Dvořák, Miroslav; Doležel, PetrThis paper presents and evaluates one approach to the problems of automatic control of a vehicle movement in a large outdoor area. The positioning of the vehicle in the area is provided by iBeacons, located at the edges of the given surface. The iBeacon is a small and low-power device which periodically transmits its UUID (Universally Unique Identifier) number through the interface of a Bluetooth 4.x. The vehicle should be able to calculate its position according to the power of the signal, considering the location of the iBeacons. To be more specific, the triangulation method is applied to determine the position. According to the set of experiments presented at the end of the paper, the position error of a robotic vehicle is mostly less then 1m.
- «
- 1 (current)
- 2
- 3
- »