基于FGD技术的公交车客流检测系统的开发

发布时间：2018-04-15 15:09

本文选题：公交车客流计数 + Hi-3515　；参考：《山东师范大学》2017年硕士论文

【摘要】：公交车作为城市居民出行的主要交通工具,犹如大动脉一般为城市的正常运作提供动力。然而由于经济的发展,城市交通情况变得越来越复杂,旧式公交系统的运作模式存在明显的不足,尤其体现在客流高峰期发车的时间,发车的次数等方面,而公交车又存在乘坐方式多,难以实名制的问题,所以统计公交车客流量并不容易。图像从古至今就是人类学习知识、增长认知以及获取信息的主要感知方式,随着时代的发展,科技的不断进步,在计算机网络日益发达,图像处理技术日趋完善的今天,利用计算机图像处理技术来帮助人们完成各种复杂的信息整合和处理变得更加简单,利用图像处理技术统计公交车的客流量便是其中之一。近年来,各个公共场合已经出现了许多检测客流的方法,其中基于红外对射统计的方法就得到了广泛认可和应用,但是红外对射法对于公交车车内环境比较复杂的现在不能很好地应对,如出现在检测区反复经过可能重复计数的情况,而且需要在车辆上外加红外对射的设备;而基于重力感应的统计技术也难以实现密集客流量统计,而且检测设备易损耗甚至损坏。目前,应用基于视频序列的图像处理方法来进行客流统计已经逐渐成为主要的手段,利用帧差法,背景差分法,光流法等手段锁定目标区域来检测运动目标,再通过对差分数据的分析来统计客流量的多少,然而,帧差法等方法存在算法自身固有的缺陷,对于公交车复杂的客流情况并不能生搬硬套,直接应用。本文针对帧间差分法的缺点提出了一种改进算法,即将帧间差分后的差分值量化后顺序转换为一维特征数组,通过非线性滤波平滑特征,在排除噪声干扰的同时保留有效突变点;另外在视频中指定检测区域,对不同区域的特征数组使用归一化处理与连续数值检测等方法,计算出目标通过检测区域的数量,具体方法如下:首先利用背景差分生成感兴趣区域,然后对感兴趣区域的Y分量进行差分,进行中值滤波和自适应阈值化,最后对波形检测与判决。本文将此算法运用于基于Linux系统的Hi-3515平台上,搭建了一个客流统计系统,并进行实验验证,完成了区域密集客流流量计数系统的实现。实验表明此算法可显著降低帧间差分的运算量,提高了密集人群流量计数的准确率。
[Abstract]:As the main means of transportation for urban residents, bus provides power for the normal operation of the city.However, due to the development of economy, the urban traffic situation becomes more and more complicated, and the operation mode of the old public transport system has obvious shortcomings, especially in the time of departure during the rush hour of passenger flow, the number of times of departure, and so on.But the bus has many riding ways, difficult to real-name system, so it is not easy to count the bus passenger flow.Image is the main way for human beings to learn knowledge, increase cognition and obtain information from ancient times. With the development of the times and the progress of science and technology, the computer network is increasingly developed and the image processing technology is becoming more and more perfect.The use of computer image processing technology to help people to complete a variety of complex information integration and processing become easier, the use of image processing technology to calculate the bus passenger flow is one of them.In recent years, there have been many methods for detecting passenger flow in various public places, among which the method based on infrared photogrammetry has been widely recognized and applied.However, the infrared shooting method is not able to deal with the complex environment in the bus, such as the repeated repeated counting in the detection area, and the need to add infrared shooting equipment to the vehicle.The statistical technique based on gravity induction is also difficult to realize the dense passenger flow statistics, and the equipment is easy to wear and even damage.At present, the method of image processing based on video sequence for passenger flow statistics has gradually become the main means, using frame difference method, background difference method, optical flow method and other means to lock the target area to detect moving targets.Then through the analysis of the differential data to count the number of passenger flow, however, the frame difference method and other methods have their own inherent defects, for the bus complex passenger flow situation can not be mechanically applied.In this paper, an improved algorithm is proposed to solve the shortcoming of the inter-frame difference method, which is to transform the difference between frames into a one-dimensional feature array after quantization, and to smooth the feature by nonlinear filtering.In addition, the detection region is designated in the video, and the number of target detection regions is calculated by using normalization processing and continuous numerical detection for the characteristic array of different regions.The specific methods are as follows: firstly, the region of interest is generated by background difference, then the Y component of the region of interest is differential, median filtering and adaptive thresholding are carried out, and finally the waveform is detected and determined.In this paper, the algorithm is applied to the Hi-3515 platform based on Linux system, a passenger flow statistics system is set up, and the experiment is carried out to verify the realization of the system of the area intensive passenger Flowmeter.Experiments show that the algorithm can significantly reduce the computation of the difference between frames and improve the accuracy of the Flowmeter.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;TP391.41

【参考文献】