2019
2019
-
Record 61 of
Title:An ethnic costumes classification model with optimized learning rate
Author(s):Zhang, Shuang(1,2); Song, Zongxi(1)Source: Proceedings of SPIE - The International Society for Optical Engineering Volume: 11179 Issue: DOI: 10.1117/12.2539608 Published: 2019Abstract:In the convolutional neural network model, the learning rate represents the magnitude of the neural network parameters that change at each iteration. When the learning rate is too high, the loss function will oscillate without converging. When the learning rate is too low, the loss function converges slowly. How to set the appropriate value of the learning rate becomes an important issue. Based on the learning rate annealing algorithm, this paper sets the segmentation attenuation and adds periodic pulse perturbation. The learning rate gradually declines and rises at the end of the cycle, and then continues to fall. This prevents the network from oscillating at a local minimum or saddle point in the late training period due to the low learning rate [1]. Finally, this paper verifies the method in the application of ethnic clothing classification using the transfer learning VGG-16 model. ? COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.Accession Number: 20193907475093 -
Record 62 of
Title:Study on high precision and smooth tracking technology of pod based on mirror control
Author(s):Xie, Meilin(1,2); Cao, Yu(1); Huang, Wei(1); Lian, Xuezheng(1); Liu, Peng(1,2); Jing, Feng(1,2)Source: Proceedings of 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference, ITAIC 2019 Volume: Issue: DOI: 10.1109/ITAIC.2019.8785553 Published: May 2019Abstract:High precision stabilization aiming and tracking technology is the core performance of photoelectric pod. Its accuracy determines the quality of image acquisition and affect the final technical and tactical indicators. In order to achieve full spectrum detection, the large pod adopts the open form, and realizes the stabilization and tracking function by controlling the two-dimensional motion of the mirror. In this paper, we focus on the research of airborne pod environment and high-precision stable tracking technology based on mirror control. Firstly, the composition of the pod system is introduced and the coupling relationship between the three-axis perturbation of the aircraft and the mirror is analyzed. Secondly, a high precision sensor information fusion and matching strategy is designed, and the synthetic angle delay compensation algorithm is briefly introduced. Then, the multi-sensor feedback control algorithm is proposed, and the acceleration loop is introduced to improve the ability of the control system to suppress disturbance torque. Finally, the proposed method is introduced into Simulink simulation. The results show that the tracking accuracy can be improved to within 0.2 mrad, which proves the effectiveness of the control strategy and can be applied to other optoelectronic platforms. ? 2019 IEEE.Accession Number: 20193507360360 -
Record 63 of
Title:Air quality prediction based on LSTM-kalman model
Author(s):Song, Xijuan(1,2); Huang, Jijiang(1); Song, Dawei(1,2)Source: Proceedings of 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference, ITAIC 2019 Volume: Issue: DOI: 10.1109/ITAIC.2019.8785751 Published: May 2019Abstract:In this paper, a time prediction model based on LSTM and Kalman filtering, LSTM-Kalman model, is proposed for the prediction of time series data with long-term and short-term characteristics. The LSTM-Kalman model uses the unique memory feature of LSTM to "store" the information contained in the pre-order data. Then it is used to obtain the underlying time series of the predicted problem in subsequent processing. Next Kalman filtering model dynamically adjusts the basic time data sequence obtained by LSTM processing. In the end, we will get the adjusted predicted value. Here we select the observation indicators in the air quality data set and builds training and test set samples to train the LSTM-Kalman model and test the performance of the sample model. As a comparison we also trained the performance of the LSTM model. In this work, we compare the RMSE(Root Mean Square Error) and R-Squared (determination coefficient) of the LSTM model with the LSTM-Kalman model predictions. The results show that the LSTM-Kalman model is better than the LSTM model, and the LSTM-Kalman model has a better fit to the predicted values of the model data. ? 2019 IEEE.Accession Number: 20193507360151 -
Record 64 of
Title:An inquiry on image enhancement methods of remote sensing images
Author(s):Liu, Jiwei(1,2); Pu, Xiaoqiu(1,2); Wen, Desheng(1); Song, Zongxi(1); Chen, Junyu(1,2)Source: Proceedings of SPIE - The International Society for Optical Engineering Volume: 11179 Issue: DOI: 10.1117/12.2540274 Published: 2019Abstract:Remote sensing image enhancement algorithm, which can enhance the contrast of image without changing its information, is crucial for the acquisition and analysis of remote sensing information. In this paper, we compare the performance of several state-of-the-art enhancement algorithms in processing remote sensing images. We classify and refine the main idea of each algorithm, and evaluate the enhancement effect with subjective evaluation and objective metrics such as entropy, brightness, contrast and mean gradient. In order to avoid color-enhanced interference, this paper took experiments both on RGB pictures and grayscale pictures. The significance and compared results were shown in the following paper. ? COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.Accession Number: 20193907475177 -
Record 65 of
Title:Design of adaptive fuzzy backstepping control algorithm for multi-joint series manipulator
Author(s):Ma, Yuhao(1); Liang, Yanbing(2)Source: ACM International Conference Proceeding Series Volume: Part F147614 Issue: DOI: 10.1145/3314493.3318457 Published: 2019Abstract:As the brain of manipulators, the control system is the core part which directly affects the overall performance indicators of manipulators, such as control precision, robustness and the ability of autonomous decision-making, etc. From the perspective of cybernetics, multi-joint series manipulator is a complicated multiple-input multiple-output (MIMO) system which is characterized by nonlinearity, time variation, strong coupling and uncertainty. As a result, traditional methods have poor control efficiency on multi-joint manipulator. To solve the above problems, an adaptive fuzzy backstepping controller is designed in this paper. First, the complex nonlinear system is decomposed into several subsystems with inversion method, and the virtual control for each of them is adopted until the whole design of the control law is completed. Then, the model information of manipulator is approximated with adaptive fuzzy system, which realizes the model-free control and reduces the uncertain influence of external disturbance. Finally, the controller can make the manipulator to track a predetermined trajectory. What's more, the adaptive fuzzy backstepping control algorithm and the simple backstepping control algorithm are simulated by MATLAB SIMULINK, respectively. The results show that the adaptive fuzzy backstepping control algorithm greatly reduces the trajectory tracking error and achieves high precision control of the manipulator. ? 2019 Association for Computing Machinery.Accession Number: 20191906897056 -
Record 66 of
Title:A high-speed multi-scale kernel correlation filter tracking algorithm
Author(s):Fu, Bin(1,2); Song, Zongxi(1); Wang, Feng(1); Gao, Wei(1); Zhang, Shuang(1); Liu, Jiwei(1)Source: Proceedings of SPIE - The International Society for Optical Engineering Volume: 11179 Issue: DOI: 10.1117/12.2539759 Published: 2019Abstract:Recently, Kernel Correlation Filter (KCF) is recognized as one of the best visual tracking algorithms which provide excellent tracking performance and high possessing speed. However, how to adapt to the scale change of target in real time and accurately is still an open problem. In this paper, a multi-scale improved method based on one-dimensional scale correlation filtering is proposed. We use the kernel function to solve the regularized least squares classifier, and the translation filter template is obtained to estimate the target position. Then, we collect a series of samples of different scales at the target position to construct a one-dimensional scale filter template for estimating the most suitable scale. The filter is learned and updated online in the tracking process. In scale estimation, we employed the dimension reduction method to reduce the computational complexity, and the interpolation method is used to increase the number of multi-scale to obtain more accurate scale location. A number of experimental results and data show that the target region overlap rate and tracking rate of the improved algorithm are improved by 9% and 60% on average compared with the best of other scale adaptive tracking algorithms. Compared with the KCF algorithm, the average position tracking error is reduced by 23.9% and the overlap rate is increased by 46%. ? COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.Accession Number: 20193907475230 -
Record 67 of
Title:Research on intelligent target detection and coder-decoder technology based on embedded platform
Author(s):Zhao, Xiaodong(1); Zhang, Xunying(1); Cheng, Xuemei(2); Chen, Fayang(3); Zhou, Zuofeng(4); Xu, Tao(1)Source: 2019 IEEE International Conference on Unmanned Systems and Artificial Intelligence, ICUSAI 2019 Volume: Issue: DOI: 10.1109/ICUSAI47366.2019.9124858 Published: November 2019Abstract:In order to meet the embedded application requirements of machine learning algorithm, the intelligent target detection and recognition algorithm based on convolutional neural network and corresponding optimal process are studied. Detailed network structure analysis and network performance analysis are carried out. Based on GPU embedded platform, TensorRT technology is used to accelerate the embedded application of intelligent target detection and recognition algorithm, including fp16 and int8 inference modes. Satisfactory verification results are achieved on embedded platform. In addition, an integrated system of real-time machine learning and H.265 encoding and decoding technology is realized. Firstly, the compressed image data sent by the camera is received by embedded platform and decoded in real time in H.265 format. Then the real-time intelligent target detection and recognition algorithm basing on TensorRT technology is done for RGB data obtained by hardware decoding process. Finally, the data is compressed in H.265 format, and subsequently storage and data transmission are carried out. The experimental results show that TensorRT technology can improve the inference speed of neural network in embedded platform. The network structure optimized by TensorRT technology can achieve three times the speed increase, with limited accuracy loss. Hardware coding and decoding of H.265 can also cause corresponding delay to program inevitably. ? 2019 IEEE.Accession Number: 20202908946958 -
Record 68 of
Title:Fast design method of smooth freeform lens with an arbitrary aperture for collimated beam shaping
Author(s):Mao, Xianglong(1); Li, Jinpeng(1); Wang, Fengbiao(1); Gao, Rong(1); Li, Xing(1); Xie, Yongjun(1)Source: Applied Optics Volume: 58 Issue: 10 DOI: 10.1364/AO.58.002512 Published: April 1, 2019Abstract:A method is presented for the fast design of a smooth freeform lens to tailor a collimated light beam with an arbitrary contour. This method begins by calculating an initial surface based on a simplified ray mapping. Then the surface is fitted by a system of Zernike polynomials, whose weights are treated as the optimization variables for further optimization. In the optimization, the objective function is analytically calculated using a partial differential-equation-based approach. To validate the effectiveness of the proposed method, a freeform lens is designed for a collimated Gaussian beam with a spline contour to form a uniform illumination distribution with another spline contour, which takes only 26 s. A freeform lens is also fabricated and experimented, and its practical performance approaches the design. ? 2019 Optical Society of America.Accession Number: 20191406723529 -
Record 69 of
Title:Research of Image Sharpness Assessment Algorithm for Autofocus
Author(s):Her, Lilin(1,2); Yang, Xiaojun(1)Source: 2019 IEEE 4th International Conference on Image, Vision and Computing, ICIVC 2019 Volume: Issue: DOI: 10.1109/ICIVC47709.2019.8980980 Published: July 2019Abstract:With the wide application of imaging system in security monitoring, aerospace, medical image and other fields, how to capture the clear image of the target in real time and automatically is particularly important. Image sharpness evaluation function is the key to evaluate the imaging quality of various imaging systems. The spatial gradient evaluation algorithm is based on the direct processing of image pixels, and the calculation is simple and intuitive. In this paper, we compare the performance of image sharpness evaluation functions in four spatial domains through two sets of atlases with different background richness. Experiments show that Benner algorithm has high scene adaptability and strong anti-jamming ability; Laplace algorithm has high sensitivity and can get results quickly in different size images; Tenengrad algorithm can reduce the occurrence of local extremum after selecting a certain threshold; Robert algorithm has poor unimodality and accuracy in two sets of atlas test. ? 2019 IEEE.Accession Number: 20201908641571 -
Record 70 of
Title:Optical system design of four-channel dual-band infrared panoramic imaging
Author(s):Ming, Gao(1); Yang, Chen(1); Xibin, Zhang(2)Source: Optical Engineering Volume: 58 Issue: 4 DOI: 10.1117/1.OE.58.4.045104 Published: April 1, 2019Abstract:In order to reduce the volume of a panoramic optical system, a four-channel infrared dual-band panoramic imager was designed using spatial multicamera image mosaicking. Each optical system of the imaging channel was designed in a double imaging configuration with an F-number of 2, working bands of MWIR 3 to 5 μm and LWIR 8 to 12 μm, and a full field of view (FOV) of 122 deg. By adopting refractive-diffractive hybrid optical elements and introducing aspheric designs, the system was made to achieve temperature compensation from -40 ° C to 60°C by means of optical passive athermalization. Results indicate that the system attained almost 100% cold stop efficiency. At the Nyquist frequency of 18 lp/mm, the modulation-transfer-function (MTF) of the MWIR system was higher than 0.70 at the edges of the FOV, whereas the MTF of LWIR system was greater than 0.35 for the same condition, both approaching the diffraction limit. ? The Authors. Published by SPIE under a Creative Commons Attribution 4.0 Unported License.Accession Number: 20192006915795 -
Record 71 of
Title:Integrated polarizers based on graphene oxide in waveguides and ring resonators
Author(s):Wu, Jiayang(1); Yang, Yunyi(1); Qu, Yang(1); Xu, Xingyuan(1); Liang, Yao(1); Chu, Sai T.(2); Little, Brent E.(3); Morandotti, Roberto(4,5,6); Jia, Baohua(1); Moss, David J.(1)Source: arXiv Volume: Issue: DOI: null Published: July 16, 2019Abstract:Integrated waveguide polarizers and polarization-selective micro-ring resonators (MRRs) incorporated with graphene oxide (GO) films are experimentally demonstrated. CMOS-compatible doped silica waveguides and MRRs with both uniformly coated and patterned GO films are fabricated based on a large-area, transfer-free, layer-by-layer GO coating method that yields precise control of the film thickness. Photolithography and lift-off processes are used to achieve photolithographic patterning of GO films with precise control of the placement and coating length. Detailed measurements are performed to characterize the performance of the devices versus GO film thickness and coating length as a function of polarization, wavelength and power. A high polarization dependent loss of ~53.8 dB is achieved for the waveguide coated with 2-mm-long patterned GO films. It is found that intrinsic film material loss anisotropy dominates the performance for less than 20 layers whereas polarization dependent mode overlap dominates for thicker layers. For the MRRs, the GO coating length is reduced to 50 μm, yielding a ~ 8.3-dB polarization extinction ratio between TE and TM resonances. These results offer interesting physical insights and trends of the layered GO films and demonstrate the effectiveness of introducing GO films into photonic integrated devices to realize high-performance polarization selective components. Copyright ? 2019, The Authors. All rights reserved.Accession Number: 20200508123 -
Record 72 of
Title:Analysis of Signal Demodulation Circuit of Eddy Current Sensor Based on Taylor Series
Author(s):Yang, Lulu(1,2); Wang, Rongbin(3); Li, Zhe(1); Qiao, Yongming(1)Source: Proceedings - 2019 11th International Conference on Intelligent Human-Machine Systems and Cybernetics, IHMSC 2019 Volume: 2 Issue: DOI: 10.1109/IHMSC.2019.10138 Published: August 2019Abstract:The alternating current bridge circuit was utilized as the signal demodulation circuit of the eddy current displacement sensor coil, and the sinusoidal alternating current voltage with a frequency of 1 MHz and an amplitude of 24V was used as the excitation source of the bridge circuit. The output voltage of the bridge was analyzed by Taylor series. The phase sensitive detection circuit and the rectifier filter circuit were respectively simulating to demodulate the impedance signal of detection coil by the Multisim software, which verified the accuracy of Taylor series expansion analysis. The simulation results indicated that the coil impedance were more effectively demodulated by rectifier filter circuit than phase sensitive detection circuit when the rate of impedance variation of the detection coil was within 10% and there was no temperature compensation. ? 2019 IEEE.Accession Number: 20200408074390