基于矢量差分的未标定摄像机P5P问题的求解

针对PnP问题的方程建立与求解, 提出一种基于矢量差分的未标定摄像机P5P问题的线性求解算法. 将5个控制点组成非共面的4点集合, 按照成像过程, 通过两个集合间的控制点矢量差分逐步构建摄像机姿态及相机矩阵的线性约束方程. 再依据线性理论与旋转阵 R 的正交性化简约束方程, 利用矢量积运算给出未标定P5P问题摄像机姿态及相机矩阵的解析解. 模拟和真实实验都验证了该方法的有效性.     

Fluorescence detection in a micro flow cytometer without on-chip fibers

This paper describes a novel concept of integrated on-chip fiber free laser-induced fluorescence detection system. The poly-dimethylsiloxane (PDMS) chip was fabricated using soft lithography and was bonded with a glass substrate of 150 μm thickness that reduced the distance of channel-to-sidewall to less than 180 μm. The cells and particles detection was conducted by an external single fiber close to the glass substrate that transmitted laser light for simultaneous excitation and receipt of the emission light signals. The performance of the proposed device was demonstrated using fluorescence beads, stained white blood cells, and yeast cells. The experimental results showed the simplicity and flexibility of the proposed device configuration which can provide convenient on-chip integration interface for fast, high throughput, and low-cost laser-induced fluorescence detection micro flow cytometer.     

基于振荡型混沌神经网络的智能信息处理研究

提出了振荡型混沌神经网络的结构模型和动力方程,证明了该网络的稳定性和Hopf分支的必要条件,得到了在三维模式空间中的轨迹,利用功率谱密度分析了重迭函数的波形.试验结果表明输入模式在靠近和远离记忆模式两种情况时,网络具有不同的信息处理能力.该网络在时空模式的动力信息处理和智能信息处理系统中有很重要的应用价值.     

重大装备制造多机器人任务分配与运动规划技术研究综述

飞机蒙皮、船舶舱体、高铁车身等大型复杂部件高效高品质制造是航空航天、海洋舰船、轨道交通等领域重大装备发展的根基. 大型复杂部件具有尺寸超大、型面复杂等特点, 传统的人工、单机制造面临着效率低、一致性差、空间有限等问题, 多机器人具有高鲁棒性、高效性等优点, 为大型复杂部件制造提供了良好的制造基础. 任务分配与运动规划是多机器人制造系统的决策中枢, 其性能影响整个系统的运行效率. 考虑到重大装备部件制造任务分配与运动规划过程中任务工序多、冲突干涉多、精度需求高等挑战, 本文首先对复杂环境下多机器人任务分配与运动规划的重要性进行了说明; 然后阐述了目前主要的任务分配与运动规划方法, 包括其在智能制造领域下的应用; 在此基础上, 对现阶段复杂场景下任务分配和运动规划存在的问题进行了分析, 并使用强化学习与混合优化算法等方法提出了解决思路; 最后对重大装备大型复杂部件制造过程多机器人任务分配和动态规划技术及应用的发展进行了总结与展望.     

基于改进YOLOX的移动机器人目标跟随方法

针对移动机器人在复杂场景中难以稳定跟随目标的问题, 提出基于改进YOLOX的移动机器人目标跟随方法, 主要包括目标检测、目标跟踪以及目标跟随三个部分. 首先, 以 YOLOX 网络为基础, 在其框架下将主干网络采用轻量化网络 MobileNetV2X, 提高复杂场景中目标检测的实时性. 然后, 通过改进的卡尔曼滤波器获取目标跟踪状态并采用数据关联进行目标匹配, 同时通过深度直方图判定目标发生遮挡后, 采用深度概率信息约束及最大后验概率(Maximum a posteriori, MAP)进行匹配跟踪, 确保机器人在遮挡情况下稳定跟踪目标. 再采用基于视觉伺服控制的目标跟随算法, 当跟踪目标丢失时, 引入重识别特征主动搜寻目标实现目标跟随. 最后, 在公开数据集上与具有代表性的目标跟随方法进行了定性和定量实验, 同时在真实场景中完成了移动机器人目标跟随实验, 实验结果均验证了所提方法具有较好的鲁棒性和实时性.     

A Study of Network Defects of a Phosphated Si-containing Low Alloy Steel

This study was to investigate the formation of network defects that appeared on a phosphated low alloy steel rod containing 1.1% Si. Laboratory oxidation, pickling, annealing, and phosphating were performed. A scanning electron microscope (SEM) with energy dispersive spectrometer (EDS) was employed to analyze the surface morphology and the elemental distribution of the specimen from each stage. From the results, a mechanism of formation of network defects has been proposed. Fast finish rolling produced fragments of Si-containing iron oxides, which were not removed completely through primary pickling → annealing → secondary pickling. Then, small pits occurred on the substrate near the interface of the substrate and the residual scale. After phosphating, small pits grew to large pits and propagated along the grain boundaries. Some grains spalled off eventually and looked like a network by human naked eyes.     

基于CT心脏图像的腔体区域分割新算法

在心脏图像处理中,为解决心脏四腔过分割问题,使用基于直方图的多阈值分割和标记分水岭分割相结合的新方法.由于CT心脏腔体区域灰度十分相似,组织相互连通,采取单一阈值分割无法得到目标区域,新算法使用多阈值分割,根据图像对应的直方图选择合适的阈值,并将分割结果作为基于标记的分水岭分割算法的输入图像,最终获得心脏的腔体区域,最后,通过重建得到左心房心室的三维结构图.而与其他分水岭改进方法相比,该算法以基于直方图的多阈值分割为前提,标记提取更为准确,目标组织边界清晰,能够获得更加理想的分割结果,有效降低了仅使用分水岭算法存在的过分割现象.     

Rapid prototyping of glass-based microfluidic chips utilizing two-pass defocused CO2 laser beam method

A method is proposed for the scribing of glass substrates utilizing a commercial CO₂ laser system. In the proposed approach, the substrate is placed on a hotplate and the microchannel is then ablated using two passes of a defocused laser beam. The aspect ratio and surface quality of the microchannels formed after the first and second laser passes are examined using scanning electron microscopy and atomic force microscopy. The observation results show that the second laser pass yields an effective reduction in the surface roughness. The practicality of the proposed approach is demonstrated by fabricating a microfluidic chip for formaldehyde concentration detection. It is shown that the detection results obtained for five Chinese herbs with formaldehyde concentrations ranging from 5 to 55 ppm deviate by no more than 5.5 % from those obtained using a commercial macroscale device. In other words, the results confirm that the proposed defocused ablation technique represents a viable solution for the rapid and low-cost fabrication of a wide variety of glass-based microfluidic chips.     

Optical microflow cytometer for particle counting, sizing and fluorescence detection

A novel on-chip optical flow cytometer concept is reported for fluorescence detection, enumeration, and sizing of microparticles in a poly-dimethylsiloxane (PDMS) microchip. The detection system integrates a pair of external optical fibers and other optical components for particle counting, sizing and fluorescence analysis in each measurement simultaneously. The scattered light signal indicates the total number and the size of the particles passing through the detection window, whereas the concurrent backward fluorescence signal shows only the number of fluorescence particles. In the experiments, microparticles of four different sizes with diameters ranging from 3.2 to 10.2 μm were discriminated and counted based on the fluorescence and scattered light intensity. The relative percentage of the fluorescence-labeled particles can be analyzed by the ratio of the events of fluorescence signals to forward scattered signals.     

Rapid glucose concentration detection utilizing disposable integrated microfluidic chip

A novel three-dimensional (3D) disposable glucose concentration detection chip is presented. The chip comprises a four-layer polymethyl methacrylate (PMMA) structure and is fabricated using a commercial CO₂ laser and a hot-press bonding technique. In the proposed device, the glucose solution is injected into a double parallel connection micromixer (DPCM) and is mixed with DNS reagent by means of a self-rotation effect. An experimental platform has been created for multiple reaction process by integrating chip and micro-heater. The fluid streams exiting the two circular mixing chambers of the DPCM are then combined and mixed further at a T-type microchannel outlet before passing to a collection chamber. Numerical simulations are performed to analyze the vortex streamline distribution within the DPCM and to estimate the mixing performance. The numerical results show that a mixing efficiency as high as 92.5% can be obtained at low Reynolds numbers (Re = 12). It is found a good linear relation of R ² = 0.9953 from the chip detection method comparing to the traditional method of R ² = 0.9976 at DNS reagent and glucose solution volume ratio of 1:1. In addition, the experimental results show that the accuracy of the glucose concentration measurements obtained using the proposed microfluidic chip is comparable with that of the measurements obtained using a conventional large-scale detection method. Overall, the results presented in this study indicate that the DPCM chip provides a rapid and low-cost means of detecting the concentration of glucose solutions.