基于改进Mean Shift算法的实时视频目标跟踪

设计了一套嵌入式平台上实现的视频目标跟踪系统.该系统采用CMOS图像传感器获取视频信号,利用Z228多媒体芯片自带的ARM9处理器完成视频信号的控制,并通过MPEG-4硬件编码器实现视频信号的压缩.用Mean Shift算法跟踪运动目标,针对其收敛的局限性设置多个搜索点来提高其跟踪效果.通过减少采样点和标记已计算点来提高代码运行速度,增强了跟踪的实时性.实验结果表明,本系统能以27 fps速率连续稳定地实现视频目标的跟踪.     

激光等离子体推进中靶动量测量方法的改进

总结了激光等离子体推进中靶动量的测量方法,设计了双光束测量方法.该方法能够减小靶宽度和靶的运动速度对测量结果的影响.     

能量连续可调LCS光源SINAP-III

X/γ探测器在航天等国家战略需求领域具有非常广泛和重要的应用。 然而, X/γ探测器要在其有效能区进行精确的标定后才能发挥作用。 目前我国缺乏sub MeV～MeV能量连续可调、 单色性好的γ源, 而用于航天的X/γ探测器无法完成精确定标, 从而使航天探测的发展出现瓶颈问题。 提出了升级原有的激光康普顿散射(LCS)原理性实验装置的方案, 建立了一个通过改变激光入射角来连续调节散射光子能量、 准单色、 极化、 sub MeV～MeV LCS光源(SINAP-III), 从而开拓LCS光源在我国航天领域（如用于航天的X/γ探测器能量定标和抗辐射加固评估研究）的崭新应用前景, 并为将来建设一个基础和应用研究相结合的多功能的γ源实验平台打下基础。 The X/γ detectors in the field of national stratagem, such as astronautical technology, have very broad and important application. These detectors, however, will play their role properly only after accurate calibrations in effective energy region. For the shortage of continuously adjustable and quasi monochromatic γ source in China, it is impossible for the detector employed in aerospace to achieve an accurate calibration so that development of such detector has encountered a big obstacle (or a bottleneck). Therefore, we propose to upgrade the original LCS device to an adjustable photon energy by changing incident angle of laser beam, monochromatic, and polarized sub MeV~MeV LCS γ source(SIMAP III) , in order to explore the new applications of LCS γ source in aerospace as well as to establish a platform for a multifunctional of γ source.     

Prompt gamma detection for range verification in proton therapy

It is an on-going challenge to verify the proton range in situ during proton therapy. Since the protons stop in target tissue, measurement of gamma-rays emitted either promptly from nuclear de-excitation or in pair from positron annihilation is the feasible method to monitor the proton range in-vivo. Using the technique of gamma collimation, we empirically demonstrated that the proton range and prompt gamma distribution are well correlated in the therapy energy range, and that measuring prompt gammas is a viable method for the clinical application. However, this collimation technique appears not to be applicable to passively scattered proton beams. The device chosen for gamma imaging in 2D is an electron tracking Compton camera, which images single-emission photons employing a gas chamber to induce Compton scattering. Images of prompt gammas were attained at the proton beam energy of 140 MeV. Measurements showed that gamma image in the energy range of 800-2000 keV provides a better match with the proton range compared to the image by lower energy gammas.     

MS-KF融合算法用于锥套跟踪

针对无人机自主空中加油过程中锥套跟踪,提出一种均值漂移-卡尔曼滤波(mean shift-Kalman filter, MS-KF) 融合算法。分析了基于均值漂移算法的锥套目标模型、相似性度量、锥套目标定位的锥套定位原理;引入卡尔曼滤波器对锥套运动状态进行预测,将锥套运动信息融合到均值漂移算法中,以保证锥套跟踪算法的稳定性和鲁棒性;给出了MS-KF融合算法用于锥套识别跟踪的流程;搭建了锥套跟踪半物理实验验证系统,分别进行MS-KF融合算法用于锥套跟踪的半物理实验验证及数值仿真分析。实验结果表明:MS-KF融合算法可以对锥套精确定位跟踪,无人机3个轴向的跟踪误差保持在0.3 m的范围内,保证了无人机自主空中加油的顺利进行。     

公路隧道太阳光直接增强照明系统的设计

为了利用绿色能源有效解决公路隧道进出口区段的黑洞效应和白洞效应,在阳光输送机研究的基础上,研制一种利用平行太阳光直接实现公路隧道进出口区段增强照明功能的新系统。该系统通过光路变换和反射等措施实现了在无光纤等传输介质条件下高效率、远距离的太阳光输送功能,利用太阳光本身时有时无的性能缺陷改善了隧道增强照明系统亮度自适应隧道外光强的技术要求,针对受光体位置固定不变的系统专门设计了极轴式自动跟踪系统,保证了对太阳光的实时跟踪。实验结果证明,该增强照明系统不但无须光导纤维或者导光管,而且使得传输效率提高40%左右,为隧道等白天也需要增强照明的地下工程直接利用太阳能开辟了一条新路。     

红外双波段目标模拟器方案与光学系统设计

通过对传统双波段目标模拟器的工作原理进行分析,针对其结构复杂以及生产成本较大的问题,提出一种只使用一套景象生成器和投影系统的双波段目标模拟器设计方案。利用分振幅的方法,通过滤光片和遮光片的遮挡滤光作用使2个波段的能量按特定的比例透过,在特定的积分时间下,探测器探测到2个波段的能量比不同,以此模拟不同红外目标。对设计方案中的投影系统进行光学设计,光学系统属于透射式,工作波段在3.3 m~3.8 m和4.4 m~4.8 m,F数为3.2,半视场角为1.1,系统分辨率达到10 lp/mm,对300 K~3 000 K的黑体、灰体或选择体均可实现红外目标的模拟,同时模拟器产生的辐射能量也大于探测器探测到的最小能量,设计方案可行。     

机载侦察设备目标定位数据处理方法

机载侦察设备对目标定位时受到的影响因素较多,目标定位数据中存在的系统误差和随机误差会影响定位结果。提出采用圆概率误差处理目标定位数据时,需分析判断圆概率误差的不同适用条件。根据圆概率误差的定义和圆概率误差方程式的一般形式,推导了6种不同适用条件情况下的圆概率误差简化计算公式。用两种圆概率误差简化计算公式对某目标定位数据进行了对比计算分析,在采用适用条件下计算公式的数据处理精度优于0.95,达到了CEP定义50％的数据概率要求,能够准确分析处理机载侦察设备的目标定位结果。     

特定边界跟踪中角点检测研究

针对边界跟踪算法中在角点附近丢失边界信息的问题,对待检测边界中角点的判别和检测方法进行了研究。分析了边界搜索中所扩展的邻域的半径和角点位置的关系,提出了角点存在判据,给出了角点存在区间。提出了寻区间法角点检测算法,用5个已知边界点之间的关系,判断出角点存在的邻域,通过选取合适的邻域半径,使角点存在区间向角点收敛。对测试图像和人侧面轮廓图像的边界利用所提出的算法进行跟踪,测试图像的边界点从19个增长为37个,人侧面轮廓图像边界点从13个增长为21个,新算法准确地检测到轮廓角点和后续种子点,使边界跟踪能够反映所有的边界信息。     

Multi-objective optimization of gradient coil for benchtop magnetic resonance imaging system with high-resolution

Significant high magnetic gradient field strength is essential to obtaining high-resolution images in a benchtop mag- netic resonance imaging （BT-MRI） system with permanent magnet. Extending minimum wire spacing and maximum wire width of gradient coils is one of the key solutions to minimize the maximum current density so as to reduce the local heating and generate higher magnetic field gradient strength. However, maximum current density is hard to optimize together with field linearity, stored magnetic energy, and power dissipation by the traditional target field method. In this paper, a new multi-objective method is proposed to optimize the maximum current density, field linearity, stored magnetic energy, and power dissipation in MRI gradient coils. The simulation and experimental results show that the minimum wire spacings are improved by 159% and 62% for the transverse and longitudinal gradient coil respectively. The maximum wire width increases from 0.5 mm to 1.5 mm. Maximum gradient field strengths of 157 mT/m and 405 mT/m for transverse and lon- gitudinal coil are achieved, respectively. The experimental results in BT-MRI instrument demonstrate that the MRI images with in-plane resolution of 50 ~tm can be obtained by using the designed coils.