视频车辆检测中改进的阈值分割算法研究  

Improved threshold segmentation algorithm in video vehicle detection

作　　者：张延丽[1] 李春贵[2] 

机构地区：[1]广西科技大学电气与信息工程学院,广西柳州545006 [2]广西科技大学计算机科学与通信工程学院,广西柳州545006

出　　处：《广西科技大学学报》2017年第2期61-66,78,共7页Journal of Guangxi University of Science and Technology

基　　金：国家自然科学基金项目(61302178)资助

摘　　要：针对使用迭代式阈值分割算法检测车辆时,易造成目标特征信息丢失以及边缘模糊化的问题,提出了一种基于形态学权重自适应图像去噪的迭代式阈值分割算法:利用数学形态学原理设计了一种权重自适应形态学滤波器,采用由小到大的多结构元构造串、并联复合形态的滤波器对视频序列图像进行去噪;同时,对迭代式阈值分割法引入一个偏移系数,可以更加快速获取最优分割阈值,对图像作精确的分割.实验证明,该算法比迭代式阈值分割算法的抗噪性好且减少阈值分割中的寻优尝试次数,得到了比较理想的分割效果.In intelligent traffic control system based on video, segmentation of scene image has a great effect on the accuracy of vehicle detection system. Using an iterative threshold segmentation algorithm for detecting vehicles can not only lead to the loss of target information and edge blurring,but also take a long time.In response to this question,this paper proposes an iterative threshold segmentation algorithm based on morphological weight adaptive image denoising.On the one hand,a weighted adaptive morphological filter which is designed by using the principle of mathematical morphology constructs from small to large multi-structuring elements by serial and parallel mixed mode.On the other hand,an offset coefficient is introduced into the iterative threshold segmentation method,which can obtain the optimal segmentation threshold more quickly to segment the image exactly.Experimental results show that the proposed algorithm has better anti noise performance than the iterative threshold segmentation algorithm and reduces the number of attempts in the threshold segmentation,which has the ideal segmentation effect.

关 键 词：权重自适应 图像去噪 迭代式阈值分割 车辆检测 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]