交通事故现场视觉图像优化识别仿真研究  

Simulation Research on Visual Image Recognition of Traffic Accident Scene

作　　者：李毕祥[1] 方兰[1] LI Bi - xiang FANG Lan(City College of Wuhan University, Science and Technology, Wuhan Hubei 430083, Chin)

机构地区：[1]武汉科技大学城市学院,湖北武汉430083

出　　处：《计算机仿真》2017年第4期387-390,395,共5页Computer Simulation

基　　金：湖北省教育厅科学技术研究项目(B2015359)

摘　　要：对交通事故现场视觉图像的优化识别,可有效提高交通事故处理效率。对视觉图像进行识别,需要获取现场图像形态学特征,建立交通事故现场图像扇形区域特征向量,完成对图像的优化识别。传统方法通过提取图像关键点的邻域图像特征,设定合理阈值,但忽略了建立图像扇形区域特征向量,导致识别精度低。提出基于智能视觉的交通事故现场视觉图像识别方法。利用均值背景建模理论提取事故现场背景图像,对背景图像进行归一化处理,同时将事故现场的轮廓图像转换为二值图像,对交通事故现场图像过滤处理,获取现场图像不同尺寸的形态学特征,建立交通事故现场图像扇形区域特征向量,结合局部辨别映射理论对交通事故现场图像特征向量进行维数约简,完成交通事故现场视觉图像识别。实验结果表明,所提方法能够有效提升交通事故现场视觉图像识别率,且识别精度较高。This article proposes a method for recognition of visual image in scene of traffic accident based on in- telligent vision. Firstly, the theory of mean background modeling is used to extract background image of scene of accident and carried out normalization processing for background image. In the meantime, the outline image of scene of accident is converted into binary image and the image of scene is filtered, then morphological feature of image with different size is acquired to built feature vector of image sector domain. Moreover, the dimensionality reduction for the feature vector is carried out. Thus, the recognition is completed. Experimental results show that the proposed method can improve recognition rate and recognition precision of visual image.

关 键 词：交通事故现场 视觉图像 优化识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]