基于轮廓特征的神经网络目标识别研究

王灿进; 孙涛; 王挺峰; 陈娟

doi:10.3788/YJYXS20132804.0641

您当前的位置：

首页 >

文章列表页 >

基于轮廓特征的神经网络目标识别研究

成像技术与图像处理 | 更新时间：2020-08-12

- 基于轮廓特征的神经网络目标识别研究
- Target Recognition Using Neural Network Based on Contour Features
- 液晶与显示 2013年28卷第4期页码：641-648
- 作者机构：
  
  1. 中国科学院长春光学精密机械与物理研究所, 吉林长春 130033
  2. 中国科学院大学北京,100049
- 作者简介：
- 基金信息：
  
  院地合作(长吉图专项基金)(No.2011CJT0006)()
- DOI：10.3788/YJYXS20132804.0641
  中图分类号： TN391
- 收稿日期：2012-11-01，
  
  修回日期：2012-12-29，
  
  纸质出版日期：2013-08-15
- 稿件说明：
移动端阅览
王灿进, 孙涛, 王挺峰, 陈娟. 基于轮廓特征的神经网络目标识别研究[J]. 液晶与显示, 2013,28(4): 641-648

WANG Can-jin, SUN Tao, WANG Tin-feng, CHEN Juan. Target Recognition Using Neural Network Based on Contour Features[J]. , 2013,28(4): 641-648
王灿进, 孙涛, 王挺峰, 陈娟. 基于轮廓特征的神经网络目标识别研究[J]. 液晶与显示, 2013,28(4): 641-648 DOI： 10.3788/YJYXS20132804.0641.

WANG Can-jin, SUN Tao, WANG Tin-feng, CHEN Juan. Target Recognition Using Neural Network Based on Contour Features[J]. , 2013,28(4): 641-648 DOI： 10.3788/YJYXS20132804.0641.

摘要

为了实现威胁源自动报警

使用BP神经网络构建自动报警系统。针对帧差法提取出的目标轮廓有重复和受变化背景影响的问题

提出了一种基于轮廓片段的目标特征提取方法。首先使用k-mediod聚类以剔除重复轮廓

再结合轮廓片段生长的方法

计算待识别轮廓和验证图片集的匹配代价以剔除背景轮廓

提取出匹配代价最小的轮廓生成轮廓片段字典。随后计算归一化的轮廓矩生成特征向量。最后将提取出的特征向量输入事先训练生成的BP神经网络进行分类。实验结果表明

算法适用于典型刚性目标识别

对于实验视频中枪支的平均识别率达到93.5%

单帧平均运算时间3.67 ms;对于Berkeley运动分割数据集中车辆的识别率达到98.2%

单帧平均运算时间5.26 ms。算法具有高实时性、高准确率的特点。

Abstract

In order to achieve automatic alarm for source of threats

an automatic alarming system is established using BP neural network. A feature extraction algorithm using contour fragments are proposed to solve the problem that the edges of the target extracted by frame subtraction is repeated and affected by changing background. Firstly

K-mediod clustering is used to remove repeated edge fragments. Then the matching cost between fragments to be recognized and validation set are computed combining with fragments growing method to eliminate the fragments of background.Then the fragments that have the minimal matching costs are extracted to obtain a codebook of edge fragments. Then improved normalized moments are calculated to generate the feature vector. At last the feature vector is inputted into the improved BP neural network for classification. Experimental results indicate that the algorithm is suitable for typical rigid target recognition. The averaging detection rate is up to 93.5% and averaging computing time of single frame is 3.67 ms for guns in experimental video. For cars in motion segmentation data set from Berkeley the averaging detection rate is up to 98.2% and averaging computing time of single frame is 5.26 ms. The algorithm proposed in this article has a high recognition rate and good real-time performance.

关键词

Keywords

references

刘翔,周桢.基于分块背景建模的运动目标检测技术 [J]. 液晶与显示,2011,26(6):831-835.[2] 靳永亮,王延杰,丁南南,等.改进的红外弱小目标检测方法 [J]. 液晶与显示,2011,26(4):555-560.[3] Chang S H, Cheng F H, Hsu W H.Fast algorithm for point pattern matching:invariant to translations,rotations and scale changes[J]. Pattern Recognition, 1997, 30(2): 311-320.[4] 舒宁,马洪超,孙和利. 模式识别的理论与方法[M].武汉:武汉大学出版社,2004.[5] 边肇祺. 模式识别 [M].北京:清华大学出版社,1998.[6] Ma S, Ji C. Performance and efficiency: recent advances in supervised learning [J]. Proceeding of the IEEE, 1999, 87(9):1519-1535.[7] Skitt P J C. Process Monitoring using auto-associative feedforward artificial neural networks [J]. J. Intell. Manuf., 1993, 4(1):79-94.[8] Lampinen J, Oja E. Distortion tolerant pattern recognition based on self-organized feature extraction [J]. IEEE Trans on Neural Network, 1995, 6:539-547.[9] Jenkins B K, Tanguay A R. Handbook of Neural Computing and Neural Networks[M]. Boston: MIT Press, 1995.[10] Xia M, Fang J N, Tang Y. Dynamic depression control of chaotic neural networks for associative memory [J]. Neurocomputing, 2010, 73(2):776-783.[11] McDuff R J, Simpson P K. An adaptive resonance diagnostics system [J]. Neural Network Comp., 1990, 2(1):19-29.[12] Leibe B, Schiel E B. Analyzing appearance and contour based methods for object categorization[C]//Proceedings of CVPR, Madison:IEEE Computer Society, 2003: 409-415.[13] Shotton J, Black A, Cipolla R. Countour-based learning for object detection//Proceedings of ICCV, Beijing: IEEE, 2005: 503-510.[14] Opelt A, Pinz A, Zisserman A. A boundary-fragment-model for object detection[C]//Proceedings of ECCV, Graz: Institute of Electrical Measurement and Measurement Signal Processing and the Institute for Computer Graphics and Vision, 2006: 575-588.[15] Huang Z. Extensions to the k-means algorithm for clustering large data sets with Categorical values [J]. Data Mining and Knowledge Discovery, 1998, 2(3): 283-304.[16] Brraw H G, Tenenbaum J M, Bolles R C. Parametric correspondence and chamfer matching: two new techniques for image matching[C]// Proc 5th Int Joint Conf Artificial Intelligence, San Francisco, USA:Morgan Kanfmann Publishers Inc., 1977: 659-663.[17] 崔潇潇, 王贵锦, 林行刚.基于边缘片段特征的目标检测 [J]. 清华大学学报, 2008,48(10):1636-1638.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于迭代传播网络的快速鲁棒的微光图像增强

基于侧窗滤波改进的神经网络非均匀性校正算法

FRKDNet：基于知识蒸馏的特征提炼语义分割网络

面向光场显示的虚拟视点生成技术进展

基于多种眼动行为的裸眼3D显示视觉疲劳评估方法