Method Of Counting Construction Rebar Based On Digital Image Processing Technology

With rapid development of technologies,all traditional industries are eager to make some breakthrough by using smart terminal equipment to reduce investment in labor force.In this way,they can develop at a faster pace and improve their own competitiveness.In the traditional construction industry,workers are still required to count the quantity of steel bars in the carrier vehicle.Only in this way,can following works be carried out.However,this manual method is of many disadvantages such as low efficiency,unreliable accuracy and high initial investment.In order to solve these disadvantages,many solutions such as quasi-circular object-based segmentation,machine vision-based solution and neural network-based solution were proposed.However,all of these solutions cannot satisfy actual requirements for identifying bundled steel bars at construction site.Therefore,a digital image processing technology-based steel bar identification method for construction sites was proposed in this paper: at first,the computer vision library OpenCV was adopted to compile a complete bundled steel bar identification program in Microsoft Visual Studio.Then,collected images were used for experiment.According to the experiment results,the algorithm was gradually improved.After the identification program achieved excellent performance,it was adapted to Android system.At last,this program on the Android smartphone basically realized the function of identifying and counting bundled steel bars at construction site.Briefly,computer vision can be defined as using a machine to replace human eyes to study images,in order to realize the function of image identification.In this paper,based on digital image processing technologies,the images of cross section of steel bars were pre-processed at the initial stage;basic characteristics of the images such as brightness and color were analyzed;the integrated filtering method(i.e.median filtering and Gaussian filtering)was adopted for denoising;and the Maximum Between-Class Variance Method(Otsu's method)was modified,and the results showed that the modified Otsu's method was more suitable for the proposed system.The contour tracking method was used to label connected components of segmented images,and favorable experiment results were obtained.At last,the method,which used the area of connected components for counting bundled steel bars at construction site was proposed.In the compiling environment Microsoft Visual Studio,based on the computer vision library OpenCV,a complete steel bar identification & counting system program was compiled to realize steel bar counting.Then,a number of improvement measures were proposed.Innovations of this paper could be divided into two aspects: 1)application innovation: this system was a prototype application of digital image processing technologies in the construction industry.In addition,interaction between OpenCV and Android was realized;and 2)algorithm innovation: a modified Otsu's method was adopted for threshold segmentation,and a new area method was used to label connected components,so as to realize the function of counting bundled steel bars at construction site.In this paper,a lot of image processing technologies were contained.A complete Visual Studio & OpenCV-based steel bar counting system was introduced.Through testing the collected steel bar image samples,high accuracy of this system was proven.In addition,this system could be adapted to Android smartphones to realize basic functions.