Bughole Detection And Evaluation Of Fairfaced Concrete Surface Based On Convolutional Neural Network

In order to promote the sustainable development of the construction industry and meet the national requirements for green environmental protection,green building materials came into being.Among them,fair-faced concrete is widely used as a kind of green concrete in construction projects.Since the use of fair-faced concrete does not require external decoration,the appearance quality is an important indicator to measure the quality of the fair-faced concrete project.At present,the domestic appearance quality of fair-faced concrete mainly adopts human eye visual evaluation and manual measurement,which is easily subjectively affected and causes unnecessary errors.Therefore,the development of objective automatic detection and evaluation methods based on computer vision has important research value and practical significance for promoting the development of fair-faced concrete technology in China.This paper is based on a large number of research and development status and deep learning development history and research progress of existing concrete appearance quality defect detection.A main quality defect on the surface of fair-faced concrete bugholes as the research object.The application of deep learning based object detection algorithm and image segmentation algorithm in the detection,location,quantitative identification and quality evaluation of bughole defects on fair-faced concrete surface has been deeply studied.Some progress has been made.The main research contents and innovative achievements of this paper are as follows:(1)In view of the current lack of objectivity and low detection efficiency of the quality assessment method for fair-faced concrete,the object detection algorithm based on convolutional neural network was studied.The bughole detection model of fair-faced concrete surface based on deep convolutional neural network was proposed for the first time.The model can quickly and accurately detect and locate the bugholes on the concrete surface,from the results of image test,the proposed DCNN had an excellent bughole detection performance and the recognition accuracy reached 96.43%.By adding the Inception modules into the traditional convolution network structure to solve the problem of the relatively small size of input image(28×28 pixels)and the limited number of labeled examples in training set(less than 10 K).The effects of noise such as illumination,shadows and combinations of several different surface imperfections in real-world environments were considered.By the comparative study with the Laplacian of Gaussian(Lo G)algorithm and the Otsu method,the proposed DCNN had good robustness which can avoid the interference of cracks,color-differences and non-uniform illumination on the concrete surface.(2)By studying the image segmentation algorithm based on convolutional neural networks,the bughole recognition and quantification model of fair-faced concrete surface based on the instance segmentation framework(Mask R-CNN)was proposed for the first time.The experimental results show that the proposed model can quickly and accurately identify the bugholes on the concrete surface image and output quantitative information such as the area and maximum diameter of the bughole.By comparing the actual size of the measured bughole with the quantification information identified by the method proposed in this paper,it is found that the area error rate of more than 68%of the bughos is less than 10%,and the maximum diameter error rate of more than 74%of the bugholes is less than 10%,the error rate of the quantitative information of most bugholes is distributed in a lower interval,showing the good recognition and quantization performance of the proposed method.(3)By analyzing the influencing factors of bughole level evaluation on fair-faced concrete surface,combined with considering the bughole level reference scale recommended by CIB and domestic related technical regulations,a method for evaluating the surface bughole of fair-faced concrete based on convolutional neural network was proposed.The method takes the bughole area percentage A_B and the bubble maximum diameter D_m as the main evaluation parameters,and proposes a bughole level evaluation table based on these two evaluation parameters.By automatically calculating the ratio of the sum of the pixels of all the detected bugholes to the total pixels of the input image,the evaluation of the area percentage of the bughole area A_B,through the quantitative analysis,can directly output the evaluation parameter of the maximum diameter D_m of the bughole.Thus,the bughole level of the input image is evaluated based on the bughole level evaluation table of the two evaluation parameters.