视觉极限对比测试仪的研制

叙述视觉极限对比测试仪的原理，介绍该仪器系统的设计过程，给出仪器性能参数的实际测试结果。     

Hyperparameter on-line learning of stochastic resonance based threshold networks

Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optimized by various gradient-based optimizers. The introduction of injected noise extends the noise level into the parameter space of the designed threshold network, but leads to a highly non-convex optimization landscape of the loss function. Thus, the hyperparameter on-line learning procedure with respective to network weights and noise levels becomes of challenge. It is shown that the Adam optimizer, as an adaptive variant of stochastic gradient descent, manifests its superior learning ability in training the stochastic resonance based threshold network effectively. Experimental results demonstrate the significant improvement of performance of the designed threshold network trained by the Adam optimizer for function approximation and image classification.     

考虑顾客便利半径和质量阈值的竞争设施选址问题

在竞争设施选址问题中,顾客选择行为是决定设施占领市场份额的重要因素,其描述了需求在设施之间的分配方式。为了贴近顾客真实的光顾行为,本文提出了一种考虑顾客便利半径和质量阈值的顾客选择规则,并研究了在该规则下市场中新进入公司的竞争设施选址问题。提出了一种基于排名的遗传算法(RGA)求解该问题,并将该算法与经典遗传算法(GA)和基于排名的离散优化算法(RDOA)进行了比较,结果说明了算法的有效性以及模型中质量阈值的重要性。     

Fuzzy rule generation methods for high-level computer vision

In many decision making systems involving multiple sources, the decisions made may be considered as the result of a rule-based system in which the decision rules are usually enumerated by experts or generated by a learning process. In this paper, we discuss the various issues involved in the generation of fuzzy rules automatically from training data for high-level computer vision. Features are treated as linguistic variables that appear in the antecedent clauses of the rules. We present methods to generate the corresponding linguistic labels (values) and their membership functions. Rules are generated by constructing a minimal approximate fuzzy aggregation network and then training the network using gradient descent methods. Several examples are given.     

Research on FBG sensor signal wavelength demodulation based on improved wavelet transform

A new Fiber Bragg Grating (FBG) wavelength demodulation scheme is studied in the paper, which consists of an improved de-noising method and Gaussian fitting peak searching algorithm. The improved translational invariant wavelet without threshold adjust factor is proposed to get a better de-noising performance for FBG sensor signal and overcome the drawbacks of soft or hard threshold wavelets. In order to get a high wavelength demodulation precision of FBG sensor signal, this de-noising method is designed to combine with Gaussian fitting peak searching algorithm. The simulation results show that the wavelength maximum measurement error is lower than 1 pm, and can get a much higher accuracy.     

Further results on large sets of disjoint group‐divisible designs with block size three and type 2n41

Large sets of disjoint group‐divisible designs with block size three and type 2ⁿ4¹ were first studied by Schellenberg and Stinson because of their connection with perfect threshold schemes. It is known that such large sets can exist only for n ≡0 (mod 3) and do exist for all odd n ≡ (mod 3) and for even n=24m, where m odd ≥ 1. In this paper, we show that such large sets exist also for n=2^k(3m), where m odd≥ 1 and k≥ 5. To accomplish this, we present two quadrupling constructions and two tripling constructions for a special large set called ^*LS(2ⁿ). © 2002 Wiley Periodicals, Inc. J Combin Designs 11: 24–35, 2003; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/jcd.10032     

The influence of space irradiated PIN photodetector on BER in satellite laser communication system

By analyzing the radiation dose on PIN photodetectors in space radiation environment, the variation of photocurrent and dark current after radiation is obtained. On the basis, the bit error rate of satellite laser communication based on space radiation dose of PIN photodetectors is established. According to simulation, when radiation dose is about 1.6 × 10³ Gy and 7 × 10⁴ Gy, bit error rate reaches 10⁻⁶ induced by 50 MeV and 10 MeV protons separately; and when radiation dose is within the range of 5 × 10 Gy–6 × 10⁵ Gy, electrons and gamma-ray irradiation also cause increase in bit error rate to 10⁻⁶. The principle of damage dose on bit error rate is investigated, and the influence of decision threshold on bit error rate is further discussed. The result shows that when radiation dose is 1 MGy, if decision threshold is increased from 4.3 × 10⁻⁷ A to 5.5 × 10⁻⁷ A, bit error rate will decrease about 4 orders of magnitude. Hence, a proper decision threshold can improve system bit error rate efficiently.     

基于相似测度算法的激光诱导薄膜损伤表面特征研究

鉴于薄膜激光损伤性能评价是增强抗激光红外观察窗口性能的重要保证,给出薄膜在脉冲激光诱导作用下的损伤表面特性及其机理。实验采用YAG脉冲激光器对TiO2薄膜样片进行1-on-1方式的激光诱导。通过CCD采集TiO2薄膜激光辐照前后2幅图像,将这2幅图像进行匹配,建立差异图像测度算法;实验得出TiO2薄膜样片的差异能量测度可判别出损伤情况,即测度值M＜0.1为未发生损伤,0.1＜M＜0.2为轻度损伤,0.2＜M＜0.5为中度损伤,M＞0.5为严重损伤。薄膜样片经过能量密度为0.5 J/cm2的激光辐照后粗糙度明显增大。研究结果表明,采用激光辐照前后图像匹配的测试方法可实现薄膜激光损伤与否的判别。     

Percolation Thresholds in Ternary Polymer Melt Blends with Separate Dispersions of the Inner Phases: Mathematical Model

A mathematical model based on a straightforward geometrical background is developed which enables predictions of a transition of one dispersed phase to a cocontinuous one (i.e., the percolation threshold) on addition of another dispersed phase during melt mixing in ternary polymer blends. The present work concerns only ternary blends with two separate dispersions of the inner phases in which no encapsulation takes place. In addition, in order to simplify the model, one of the inner phases was represented by hard, nondeformable microspheres The expression developed describes well an experimental relationship between the percolation threshold, the concentration above which the former dispersed phase transforms to a continuous one, and concentrations of both inner phases. The results agree well with the experimental data obtained in a previous work.