基于活动轮廓的机器人视觉伺服控制

提出了基于活动轮廓的视觉伺服控制方法。将活动轮廓和光流模型相结合用于运动物体图像的实时跟踪 ,以抽取物体图像的边缘信息 ,并以此信息控制摄像机的运动 ,达到机器人定位、跟踪等目的。该方法跟踪精度高 ,鲁棒性好 ,且满足实时跟踪的要求。实验结果表明了该方法的正确性和有效性。     

分开合并算法的若干讨论和改进

分析了分开合并图像区域分割算法中两个影响处理结果正确性的因素 :区域合并顺序和区域合并谓词。针对这两个因素分别提出了改进的方法 ,并给出了具体改进算法。改进算法的实验结果和传统算法结果进行了对比 ,结果表明了改进的有效性。     

基于DSP的电视跟踪系统算法设计

研究基于ADSP21020高速DSP芯片构成的实时电视跟踪系统的图象采集。分割、跟踪定位和跟踪波门设计等算法,并给出算法实现的流程图。     

一种探雷声纳图像分割方法及应用

提出一种适用于探雷声纳图像的疑似水雷目标扫描分割方法。该方法以图像区域生长法为基础,根据探雷声纳的成像原理及水雷目标的图像特性,对区域生长法中的参数做了特别设定,分搜索和定位两步对图像中疑似目标进行扫描分割,讨论了搜索和定位两种扫描窗口的尺寸及扫描方法。将该方法应用于实际探雷声纳图像,可得到符合预定信混比要求的疑似目标区域,为基于探雷声纳图像的水雷自动判别奠定了基础。     

Determining newspaper marketing zones using contiguous clustering

We develop a heuristic procedure for partitioning graphs into clusters of nodes such that each cluster of nodes induces a connected subgraph with the objective of minimizing the differences within clusters as measured by the total differences between all pairs of nodes of a cluster. We apply our procedure to determine optimal delivery zones for community sections of a major newspaper while including a number of operational constraints. Our results demonstrate a 18–56% improvement in the total differences within the zones designed over the usual intuitive heuristics. Given the magnitude of the revenue generated by larger local newspapers through zoning, our method demonstrates how better zoning can significantly enhance the value of segmentation. © 2004 Wiley Periodicals, Inc. Naval Research Logistics, 2005.     

基于目标航迹分段的任务分配方案数学模型的建立

任务分配是进行计算机辅助决策的重要内容之一。本文从目标航迹分段着手,建立一种新的分配模型,解决了目标流的多次分配问题,并给出了具体应用实例。     

基于小波变换与数学形态学的红外图像分割方法

提出一种基于小波变换和形态学相结合的红外图像分割方法,该方法利用小波变换和形态学消除红外图像的混合噪声,抑制背景干扰和增强目标;通过选择适当的结构元素进行系列形态组合运算,自适应确定阈值,提取目标;利用形态学水线区域分割法对图像进行分割,分离目标区域。实验结果表明,该方法能有效检测和分割出低信噪比复杂背景红外图像中的目标。     

Dynamic pricing in a trade-in program with replacement and new customers

Trade-in programs have been widely adopted to enhance repeat purchase from replacement customers. Considering that a market consists of replacement and new segments, we study the joint and dynamic decisions on the selling price of new product (hereafter, “selling price”) and the trade-in price involved in the program. By adopting a vertical product differentiation choice model, we investigate two scenarios in this paper. In the base model, the manufacturer has sufficiently large production capacity to fulfill the customer demand. We characterize the structural properties of the joint pricing decisions and compare them with the optimal pricing policy under regular selling. We further propose a semi-dynamic trade-in program, under which the new product is sold at a fixed price and the trade-in price can be adjusted dynamically. Numerical experiments are conducted to evaluate the performance of the dynamic and semi-dynamic trade-in programs. In an extended model, we consider the scenario in which the manufacturer stocks a batch of new products in the beginning of the selling horizon and the inventory cannot be replenished. Following a revenue management framework, we characterize the structural properties with respect to time period and inventory level of new products.     

基于复杂背景的图像分割算法

针对圆柱形钢棒在线计数特定的复杂背景提出了基于双阈值结合数学形态学运算的分割算法。首先分别根据2个阈值进行分割得到2幅二值图像,然后根据目标是圆形以及其大致位置等先验知识分别对2幅图进行腐蚀,开启,闭合等数学形态学运算,最后将高阈值分割得到的图像中的边界信息叠加到低阈值分割得到的图像中。实验结果表明,本方法作为本系统的图像预处理是有效的。     

一种灰度图象中的小目标识别方法

介绍了一种基于目标灰度门限和目标之间灰度距离门限的区域自动阈值检测法,用该方法检测提取出灰度图像中包含目标的小区域。然后利用传统的门限自动选择方法找到合适的门限,利用该门限值对所获目标区域进行二值化以得到目标,然后采用改进不变矩方法提取目标特征并采用优化BP神经网络进行识别。该方法经实验验证,效果较好。