Estimating the height of buildings only based on monocular satellite images is a difficult task due to limited annotated data quantity, subpar data quality, changes in building kinds, and varying angles of satellite perspective, light and shadow. Given the task of predicting a single height value for each building in a satellite image, this work proposes a method for instance segmentation based building height estimation. In a single end to end multitask model, the system predicts contours and corresponding heights of buildings from a monocular remote sensing image. Using a dataset curated for this study, we present initial results on building height classification with three classes. Additionally we use transfer learning from a public building contour dataset to improve results. Overall, an average classification accuracy of 70% is achieved for buildings in each standalone height class.
In this study, SPICE simulation of the CISPR 25 test setup has been carried out to observe the effect of different structures on the electronic board designed for the LED driver of side marker lamp of a car on the conducted emission (CE) level. Therefore, four different SPICE models have been created. In addition to these, the CE test for this electronic board has been carried out in a semi-anechoic chamber. Comparing simulation results with the actual test results, the closest results have been found in the analysis of the fourth model. In the frequency range of 150 kHz - 30 MHz, the highest CE values have been found as 58.13 dBnV at 150 kHz in the fourth model (added vias), while it has been 90.17 dBnV at 213.13kHz in the actual test results. Results for 30 MHz - 108 MHz have been found as 71.47 dBnV at 105 MHz and 88.82 dBnV at 59.39 MHz, respectively.
We show how to address nonlinearities in power amplifiers (PAs), which limit the power efficiency of mobile devices, increase the error vector magnitude, using an deep neural-network (DNN) method. DPD is frequently performed using polynomial-based algorithms that employ an indirect-learning architecture (ILA), which can be computationally complex, particularly on mobile devices, and highly sensitive to noise. By first training a DNN to model the PA and then training a predistorter using PA data through the PA DNN model. The DNN DPD successfully learns the unique PA distortions that a polynomial-based model may struggle to fit, and therefore may provide a nice balance between computation cost and DPD efficiency. We use two different DNN models to show the performance of our DNN approach and examine the complexity tradeoffs.
Adaptive Beamforming is an array signal processing problem in which a beam pattern is generated in the direction of desired signal and nulls are placed in the directions of undesired signals. Neural networks are widely employed for adaptive beamforming problems. In this paper, a autoencoder(AE) guided radial basis function(RBF) neural network(NN) is proposed. The proposed AE guided RBF neural network is also named as hybrid neural network since it is a combination of AE and RBF. Feature extraction and data compression are achieved by using the encoder part of AE in the input and first hidden layer of the hybrid neural network. The proposed hybrid neural network model is compared with a classical 3-layered RBF neural network according to Signal to Interferance Ratio(SIR) performance, Half Power Beamwidth(HPBW) and processing load. Simulation results show that the proposed hybrid neural network reduces the processing load without decreasing the SIR performance.
In recent years, quantum computer development projects have progressed considerably in terms of quality and quantity. These studies also pose a serious risk for widely used security systems based on classical public-key cryptography algorithms. Therefore, development of hardware and software applications that use post-quantum methods are considered a necessity both for today and for the future. Kyber is one of the most promising key encapsulation mechanism in the final round of NIST post-quantum cryptography standardization process. In this study, we present a 64-bit application specific transport triggered architecture processor which has custom operations for accelerating the Kyber algorithm. Moreover, we compare the processor and its various versions with a 64-bit RISC-V core in terms of performance, total energy consumption and resource utilization. According to the test results, our custom processor provides at least 2x better performance and 1.5x lower energy consumption while increasing the required chip area approximately 1.4x.
Population growth and developments in the world increase the demand for electrical energy every year. With the investments made to meet this demand, high voltage facilities are enlarged, voltage levels are increased and power equipment capacities are further challenged. Therefore, insulation coordination is of great importance for the continuity of the power systems. Since breakdown strength is the factor that determines the usage area and service life of insulation materials, it should be taken into consideration for the proper insulation design. One of the factors affecting the breakdown strength is the type of applied voltage. For this reason, the breakdown performance of insulating materials should be investigated under different types of voltages. In this study, the breakdown performances of presspaper, polyethylene terephthalate (PET) and styrene-butadiene rubber/natural rubber (SBR/NR) samples were investigated under AC, positive DC(+) and negative DC(-) voltages. Experiments were carried out in Yildiz Technical University High Voltage Laboratory.
In this paper, a flux controlled high frequency fully floating type memristor emulator circuit based on OTA (Operational Transconductance Amplifier) is introduced. The emulator circuit is realized using operational transconductance amplifier, multi output transconductance amplifier, a grounded resistor, and a grounded capacitor. The proposed circuit can be configured in both incremental and decremental topology by changing the connections. Simulations were performed with LTSpice using $0.18-\mu \mathrm{m}$ CMOS technology at a supply voltage of ±1.5V. The memristor characteristic can be electronically tuned by adjusting the transconductance of the OTAs. To obtain the characteristic at higher frequency in the current versus voltage plane, capacitor value can be change. The proposed emulator circuit performs well up to 20MHz.
In this research, a novel phase-BOTDR based distributed acoustic sensing method using the phase change occurring along the sensing fiber due to the ambient temperature has been proposed. In this method, distributed phase change information has been obtained depending on temperature variation on the model built up for making analysis and performing corresponding simulations. Moreover, the differential analysis of the product of fiber core refractive index and quantity change of fiber length has been attained and the linear equations related to the temperature and core refractive index dependencies of phase change have been derived. Furthermore, the temperature dependence of phase change has been obtained as 45.667 rad/°K. For the variations of core refractive index in 1.44183-1.44188 range, values of phase change have been acquired in the range of 0 rad −227.7 rad. In other words, a 1 rad increasing in phase change induces the variation in value of ~ 2.23×10−7 on the core refractive index of the sensing fiber.
Adaptive network based fuzzy inference system (ANFIS) is utilized for efficient detection of counterfeit banknotes. ANFIS is a hybrid system that employs a network architecture to fine-tune the membership functions of a fuzzy inference system. Three input space partitioning methods are investigated for banknote authentication. The average classification performances for 30 runs of 5-fold and 10-fold cross validation experiments are reported. The experimental results and statistical analyses confirm that ANFIS models based on subtractive clustering and grid partitioning offer better classification performance than the ANFIS model based on fuzzy c-means clustering for banknote authentication.
With the cheapening of the technological products, users have started to use intelligent building automation systems for saving purposes especially in the European Market. The impact of rising energy prices and taxes on this situation is quite large. In this study, a novel dimmable lighting control unit has been developed to control and manage lighting systems, which is one of the most important parts of intelligent building automation. By controlling multiple lighting units from a single point, energy savings have been achieved by choosing the optimal operating conditions that will be necessary for the environment. The novel lighting control unit, which also provides control via mobile devices, supports the KNX infrastructure, which is used quite widely in the Global Market. It is also integrated into systems running on a network infrastructure with an Ethernet connection. In addition, it will be used in old buildings with no automation infrastructure with the help of digital inputs.
Generalized Cross Correlation method is commonly used for estimation of a signal Time Difference of Arrival. The method is based on applying cross correlation between signals at reference receiver and signals at principal receiver. This commonly used method may not work well under low signal-to-noise ratio conditions observed in passive positioning or ranging systems, such as navigation systems using opportunity signals or passive radars. This paper presents an enhanced technique for Generalized Cross Correlation method applied on Arrival Time Difference estimation by dividing the received signals into small time windows and then integrating the correlation results for each sample. The Pulse Integration method in the radar literature is the inspiration of this work. More accurate upshots are observed in this work even in low signal to noise ratio conditions.
This paper focuses on channel modeling and characterization of indoor visible light communication (VLC)-based medical body sensor networks (MBSNs) which establish links between light-emitting diodes (LEDs) and MBSNs nodes couple with photodetectors (PDs) placed on the shoulder (D1), wrist (D2), and ankle (D3) of the mobile user who walks over random trajectories in 3D scenarios of ICU ward and family type patient room. We adopt non-sequential ray-tracing to obtain channel impulse responses (CIRs) and channel characteristics over random trajectories. Based on simulation results, it is observed that channel DC gains exhibit sinusoidal behaviour for D1 and D2 except for D3 (i.e., due to the number of diffuse rays received at D3), as the user approaches and moves away from the luminaries. It is also revealed that a flat fading channel can be modeled if a data rate lower than 7.03 Mbit/s, i.e., sufficient for MBSNs applications, is chosen.
Increase in power demand, green initiatives and regulatory requirements force system operators to adopt more complicated distribution system layout which are resulted in dissension from traditional distribution system practises. The main source of this complexity can be listed as; extensive application of embedded generation into distribution systems, different earthing practices, increasing percentage of under-ground cables and lastly, parallel feeders causing loop flows. One of the consequences of the increased complexity is malfunctioning in the protection systems namely; false tripping especially for the earth faults. Thus, conventional non-directional protection relays are, grad-ually, replaced with the directional ones in radial topology. How-ever, directional protection is more expensive since requirement of additional measurement devices is imminent. Consequently, to facilitate cost effective protection system implementation, directional protection and non-directional protection performances should be investigated properly. Distribution system which earthed trough reactance may also suffer directional discrimination because of possibility of unintentional resonance between neutral reactor and cable capacitances. In this study, boundaries of directional and non-directional protections are investigated by taking into account neutral reactance, cable length and parallel feeder number. It is proposed that, a hybrid algorithm selection procedures based on system parameters should be undertaken for a successful and selective protection.
Electronic nose is becoming a popular tool for various application areas. The data of an electronic nose is collected with various chemical sensor arrays and then odors are classified with suitable pattern recognition methods. This paper proposes a convolutional neural network for the the classification task of a wine quality electronic nose dataset. Method was tested on different portions of the dataset and compared with two previous studies. Proposed method managed to obtain high accuracy results within the relatively short time period. Additionally, method was tested by using portions of the sensor responses, hence allowing the user to assess wine quality earlier. Each training was repeated ten times in order to minimize the effects of random data selection.
State of Health (SOH) of lithium-ion batteries is a major criterion to understand the remaining energy left in the battery. For calculating the SOH, among many cumber-some methods, the Incremental Capacity and Delta Voltage (IC/DV) method provides accurate results with less effort. So in this study, IC/DV method was experimentally implemented on a set of 18650 Panasonic cells. It was shown that IC/DV method can successfully predict the SOH of lithium-ion batteries up to 1C charge.
In this study, we trained a convolutional neural network to detect disease in tomato leaves and then examined the effect of hyperparameters and layers used when training a convolutional neural network on the trained model. In our study, it was observed that with hyperparameter tuning, it is possible to increase the validation accuracy of a CNN trained from scratch using the plantvillage dataset from 92% to 98%.
This paper presents a practical method which can be used to estimate the ratio of kinetic energy turned into regenerated electric active energy through regenerative braking by electric locomotives used to tract freight trains. An algorithm implemented in Matlab was conceived and used to process 2 types of data: static (tracted mass) and dynamic (altitudes, speeds and regenerated electric active energies). The dynamic input data were recorded by an onboard equipment every 5 minutes across the same route in a round trip, in order to diminish as much as possible, the influence of the route geographical characteristics. Because the train mass varied with the sense of travel, calculations were made separately for each sense, considering only motion points and 2 different thresholds for the altitude difference between adjacent points. Criteria used to filter the input data and respectively to reject deviant values were conceived and results are discussed.
This work presents our proposed solution to provide interoperability among systems that have different microservice architectures. Publish/Subscribe technologies are targeted, specifically Kafka and RabbitMQ. Interoperability is achieved via a gateway tool which is designed as a microservice-based Publish/Subscribe system allowing more than two systems to work together. Experiments are conducted between a Kafka-based and a RabbitMQ-based system. Results indicate the applicability of the gateway under stress, when subjected to tens of thousands of bi-directional messages transmitted per second.
A new microstrip patch antenna was designed and fabricated for a 2.45 GHz and 5.2-5.9 GHz WLAN receiver. The antenna covers a bandwidth of about 700 MHz for the 5.2-5.9 GHz WLAN band. In order to get a better response for the lower band at 2.45 GHz, a microstrip matching circuit was added. Electromagnetic simulations were used for the design and optimization process. In addition, antenna measurements were performed to validate the simulations. Some measurements will be presented during the conference.
Deep learning has been studied extensively for driver drowsiness detection using video data. However, since the proposed deep learning methods are computationally cumbersome, the commercial driver drowsiness detection methods are still using hand-crafted features such as lane deviation and percentage of eye closure. This study investigates a deep learning model that provides a fair drowsiness detection performance with a lightweight architecture. In the proposed method, Dlib library was used to detect the driver's face in individual frames of video data. The detected faces are fed into a pre-defined convolutional neural network architecture. Then, a long short-term memory network was used to capture the temporal information between the frame sequences to assess the state of drowsiness. The proposed model achieves a detection accuracy of 80% in a popular benchmark dataset. It was also verified that the model could be implemented on a commercial and inexpensive development board with a frame rate of 5 frames per second.
