基于神经网络的人车识别技术在智能监控中的应用

文档编号:ZD858    文档字数:18842,页数:38

摘  要
 在现代智能交通中，随着传感器和计算机技术的高速发展，模式识别技术成为智能交通中的核心内容，人车图像识别在交通运输领域得到越来越多的重视和发展。
 本文介绍了图片的预处理，图像分割，提取出图像中的背景，从图像中检测、提取出运动目标，构建神经网络和分类树模型，对人车进行识别。预处理包括图像灰度化、图像二值化、图像的梯度锐化、图像的去噪处理等技术。由于背景是静态的，目标检测采用背景相减法。本文综合比较了神经网络和分类树这两种完全不同的算法，尽管分类树更易于理解，判决简单，但鉴于神经网络具有十分强的自适应、自学习功能，在精度上略胜一筹，最后采用基于神经网络的图像识别技术。
 本设计主要采用VS2005作为开发工具，以Opencv作为图像处理实施操作的运行平台，主要研究了固定某段道路下基于背景相减法的运动检测及其人车识别技术，实现了一个简单的智能监控系统。

关键词：模式识别、智能监控、神经网络

Abstract
 With the rapid development of the sensors and computer technology, pattern recognition becomes the core content of the modern intelligent transportation. And vehicle image recognition in the field of transport has been more and more attention and development.
 This article describes the image preprocessing, image segmentation, the extraction of the background image from the image detection and moving objects, building of a neural network and classification tree model, and identification of people and vehicles. The technology of preprocess includes gray image, image binarization, image gradient sharpening and image de-noising processing technology. As the background is static, object detection takes the method of background subtraction. This paper compares two completely different algorithms of the neural network and classification tree. The classification tree is easier to understand and simpler. But neural network has very strong self-adaptive, self-learning function and better accuracy. Finally, the neural network model is built to process image recognition.
 This project uses the VS2005 as its development instrument and the Opencv as its moving platform. The study builds a simple intelligent surveillance system, which is based on the motion detection under a fixed background subtraction and vehicle identification technology on a certain section of road.

Keywords：Pattern Recognition; Intelligent Surveillance; Neural Network

目 录
摘要...I
Abstract.II
目录...III
1 绪论...1
1.1  引言.1
1.2  视频监控系统的发展1
1.3  人车识别技术在智能监控的应用3
2 图像处理技术.5
2.1  图像预处理.5
 2.1.1  图像的灰度化.5
  2.1.2  图像的二值化.5
 2.1.3  图像的梯度锐化.5
 2.1.4  图像去噪处理.6
 2.1.5  基于提高精度的动量校正法.6
 2.1.6  图像的细化.6
2.2  阈值法图像分割.7
 2.2.1  阈值法的基本原理.7
  2.2.2  阈值法图像分割方法分类.7
 2.3  连通区域标记.8
3  运动检测...10
 3.1  背景相减法10
  3.1.1  初始背景帧的获取.11
  3.1.2  背景相减法检测运动物体.12
 3.2  时域差分法13
 3.3  光流法14
4  智能监控的实现..16
 4.1  设计步骤16
 4.2  样本图像16
 4.3  特征提取18
 4.4  OpenCV介绍.20
 4.5  图像识别21
 4.6  智能监控的效果图27
5  总结.29
致谢....30
参考文献...31
附录一...32

相关论文