基于数字图像处理技术建筑工地钢筋计数的方法

发布时间：2018-11-09 20:22

【摘要】：随着科技的发达,传统行业均想打破传统的桎梏,利用智能终端设备来减少劳动力的投入,从而获得更快的发展,增强自身竞争力。传统建筑行业在钢筋成车来料时,仍然需要人工清点数量,然后才能开展后续工作。这样效率低、准确率无法保证、增加前期投资。针对这一问题的解决办法,主要提出过基于类圆的分割、基于机器视觉和基于神经网络等解决方案,但都仍难以达到建筑工地识别成捆钢筋的实际要求。本文提出一种基于数字图像处理技术建筑工地识别钢筋的方法:首先在VS上利用OPENCV库编写了一套完整的成捆钢筋识别程序,对采集到的相片进行实验,利用实验结果一步步指导算法的改进。然后当识别程序效果良好时,并移植到安卓系统中,在安卓手机上基本实现了建筑工地成捆钢筋识别计数的功能。计算机视觉简要概括:用机器代替人类的眼睛,研究图像,从而实现识别的功能。本文以数字图像处理为基石,通过对钢筋横截面图片进行前期预处理,分析了原始图像在亮度、色泽等方面的基本特征,采用组合滤波(中值滤波和高斯滤波)的方式去噪,对最大类间方差法进行改进,改进后的最大类内方差法更适合本系统。利用轮廓跟踪法完成了对分割后图像的连通区域标定,得出良好的实验结果。并提出利用连通邻域的面积对建筑工地成捆钢筋进行计数。在Microsoft Visual Studio编译环境下结合计算机视觉库OPENCV编写了完整的钢筋识别计数系统程序,实现钢筋计数。并提出了一些改进措施等。本文的创新点分两部分:一是应用创新:该系统是数字图像处理技术在建筑行业上的原型应用,并实现了OPENCV与安卓之间的交互。二是算法创新:通过改进的最大类内方差法进行阈值分割,运用一种全新的面积法对连通区域进行标定,从而实现建筑工地成捆钢筋的计数。本文包涵了众多图像处理的技术,是基于VS和OPENCV的一套完整的钢筋计数系统。通过采集的图像样本测试,结果准确率高。并移植到安卓系统手机实现基本功能。
[Abstract]:With the development of science and technology, the traditional industries want to break the traditional shackles, use intelligent terminal equipment to reduce labor input, so as to obtain faster development and enhance their competitiveness. The traditional construction industry still needs to count the quantity manually before it can carry out the follow-up work. This efficiency is low, accuracy can not be guaranteed, increase investment in the early stage. In order to solve this problem, we mainly put forward some solutions, such as segmentation based on circle, machine vision and neural network, but it is still difficult to meet the practical requirement of identifying the bundles of steel bar on construction site. In this paper, a method based on digital image processing technology to identify steel bar in construction site is proposed. Firstly, a complete program for identifying steel bars in bundles is compiled by using OPENCV library on VS, and the collected photos are tested. The experimental results are used to guide the improvement of the algorithm step by step. Then, when the recognition program works well, it is transplanted to Android system, and it basically realizes the function of recognizing and counting the construction sites in bundles of steel bars on the Android phone. Computer vision is a brief summary: machine instead of human eyes, study images, so as to achieve the function of recognition. Based on the digital image processing, this paper analyzes the basic features of the original image in brightness and color by pre-processing the cross-section image of steel bar, and adopts the combined filter (median filter and Gao Si filter) to de-noise. The method of maximum inter-class variance is improved, and the improved method is more suitable for the system. The contour tracking method is used to calibrate the connected region of the segmented image, and good experimental results are obtained. In addition, it is proposed to count the bundles of steel bar in the construction site by using the area of the connected neighborhood. In the environment of Microsoft Visual Studio compilation, a complete program of steel bar recognition and counting system is written in combination with computer vision library OPENCV. Some improvement measures are also put forward. The innovation of this paper is divided into two parts: one is application innovation: this system is a prototype application of digital image processing technology in the construction industry, and realizes the interaction between OPENCV and Android. The second is the innovation of the algorithm: the threshold is segmented by the improved maximum intra-class variance method, and a new area method is used to calibrate the connected area, so as to realize the counting of the bundles of steel bars on the construction site. This paper includes many image processing techniques, is a complete steel bar counting system based on VS and OPENCV. Through the image sample test, the accuracy of the result is high. And transplanted to Android mobile phone to achieve basic functions.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TU755.3

【相似文献】