Research On The Circular Pipe Count Based On Image Processing

In the process of infrastructure construction and industrial production, the circular pipe count need to be done very frequently, but the work has always been very complex and burdensome. Traditional circular pipe count is done in manual way, and it will have a lot of problems in the implementation process, such as a huge amount of work, high error rate and low work efficiency, etc. Therefore, the research and development of automatic circular pipe count system is becoming increasingly important and urgent, and it is of great significance for solving the problem of automatic circular pipe count, and improving the efficiency of work.Digital image processing technique has the advantages in speediness, flexibility, being easy to operate and many other aspects. Therefore, automatic circular pipe count system based on image processing has a lot of implementation plans, and the main research methods have two kinds:(1) image segmentation processing model:the image segmented of the end face of the pipe as the basic unit for count;(2)edge algebraic mode:algebraic analysis result of image edge will be counted as a fundamental basis, such as fitting, Hough ellipse detection, and etc. However, the two processing modes have their drawbacks:it's difficult for image segmentation processing model to deal with keeping out each other between the pipes, the incomplete edge, large circular pipe clearance, and etc; while it's also difficult for edge algebraic mode algebraic model to deal with edge of adhesion, too much broken edges, overly complex background texture image, and etc. The process is not only time-consuming, but also at a low correct rate. Therefore, this paper, according to the characteristics of the circular pipe image, combines the two models, and proposes a new automatic circular pipe count system implementation strategy, which completes the automatic circular pipe count by the classification guidance filter of edge.In the process of the related study of the circular pipe count system based on image processing, this paper adopted the on-site experimental data taken by mobile phone. In this paper, the main contents and conclusions are as follows:1) In order to obtain high quality edge information of pipe image, in this paper, some commonly used image preprocessing methods are compared and analyzed by experiments. The results show that weighted average method is more suitable for gradation processing of color images, in order to reduce the image dimensions and the complexity of the data; Median filtering method is more suitable for filtering processing of gray image to minimize losses in the image edge information; Image edge objects detected by Canny operator are relatively more complete, which can benefit the subsequent classification and extraction of edge.2) In order to make the edge image can satisfy the requirements of the edge detection, the image edge has been carried on the edge thinning, multiple join points elimination and bend the interrupt processing. Edge thinning can eliminate redundant data at the same time to and improve the precision of the edge at the same time; Edge points to eliminate processing is used to solve the problems that it often arise which edge adhesion between different types of edges; A bending interrupt handling is used to solve the problems that two different edge objects often appear the phenomenon of connection in the edge image. Using these methods to the experimental results show that, The method combination can decompose the interference edge of the irregular and complicated in the background edge of the image, in order to eliminate; And it can interrupt connection of edge of different pipe end, in order to make different pipe the edge of the object to be independent.3)In order to realize circular pipe count by means of classification and recognition, this paper put edge divided into two classifications, closed edge and no-closed edge according to the connectivity of edge, and filtered the edge of the classification results according to the image segmentation properties and filtered edge fitting parameters. According to the way of segmentation screening in image segmentation processing model, closing edge is shifted according to features of image segmentation filled closing edge, and then it use the ellipse fitting statistical parameters of filtered closing edge to guide the non-closed edge filter. The process by experiment shows that:(1) Making easy to identify the closing of the edge be treated separately can reduced the indiscriminate fitting the amount of data, and reduces the complexity of the screening process;(2) Using high recognition rate sections to guide the low recognition rate of screening, has higher accuracy and stability than a single recognition model.4) The automatic circular pipe count system design and programming is completed, and its total recognition accuracy is higher.