自适应轮廓的变分水平集复杂背景多目标检测

发布时间：2018-02-15 01:46

本文关键词： 变分水平集帧间差分算法 K-means聚类算法形态学复杂背景　出处：《软件学报》2017年10期 　论文类型：期刊论文

【摘要】：无需重新初始化的变分水平集模型能够避免经典水平集模型的重复初始化步骤,进而简化计算,缩短检测所需时间,同时能够有效利用图像的边缘梯度信息,从而准确定位图像的局部结构.但该模型不能自适应地获得初始化曲线,水平集的拓扑结构也无法改变,不能解决多个目标的检测问题.针对以上问题,提出了一种基于自适应轮廓的变分水平集复杂背景多目标检测方法,该方法采用帧间差分算法与K-means聚类算法相结合,以获得多个运动目标的初始化曲线,通过形态学方法来降低图像噪声的干扰,从而快速自适应地估计复杂背景下运动目标的位置和轮廓大小.该算法进一步对无需初始化的变分水平集进行改进,将其由单目标检测模型扩展为多目标检测模型,并修正原模型难以处理图像灰度不均匀的问题,最终实现对复杂背景下多个目标的检测.在标准数据库和实际数据集上的测试结果表明,所提方法能够准确地定位不同尺度和灰度目标的轮廓,从而提高算法的演化迭代效率及准确性.
[Abstract]:Repeat initialization steps without re initialization of the variational level set model can avoid the classical level set model, so as to simplify calculation, shorten the detection time required, and can effectively use the edge information of the image gradient, local structure so as to accurately locate image. But the model can adaptively obtain the initial curve, the topological structure of the level set can not change, can not solve the detection problem of multiple targets. To solve the above problems, put forward a set of multi target detection method in complex background based on adaptive contour, the method of mining algorithm and K-means clustering algorithm combined with the frame difference, to initialize the curve of multiple moving targets to reduce. Image noise by morphological method, and fast adaptive estimation of moving target under complex background position and outline size. The algorithm of step on To initialize the variational level set is improved, which consists of a single target detection model is extended to multi target detection model, and modified the original model is difficult to deal with the problem of uneven gray image, finally realize the detection of multiple targets in complex background. In the standard database and the actual data set on the test results show that the the proposed method can accurately locate the different gray scale and the outline of the target, so as to improve the efficiency and accuracy of the iterative evolution algorithm.

【作者单位】：西安电子科技大学空间科学与技术学院;西安电子科技大学电子工程学院;
【基金】：国家自然科学基金(61501352,61503292,61203202) 陕西省自然科学基础研究计划-青年人才项目(S2015YFJQ0573) 中央高校基本科研业务费专项资金(JB151308,JB150228,JB161308,XJS16075)~~
【分类号】：TP391.41

【相似文献】