圆形和类圆形的管道广泛应用于地质石油勘探，建筑等行业中，由于大量无序的管道受光照、摆放位置、角度、直径大小以及作业环境复杂等多方面因素的影响，它的识别和计数一直以来是一项复杂的任务。传统的人工计数方法存在着明显的缺点，劳动强度大，并且效率低下。为解决这一问题，迫切需要研究出高精度的自动统计设备。本文利用数字图像处理及计算机视觉等方面的算法对钢管进行计数。 针对大量的钢管摆放无序，光照不均匀且多阴影的现象提出了改进的局部动态阈值算法，利用该算法对钢管图像进行分割，得到了较精确的二值图像；然后对该二值图像用高斯-拉普拉斯算子进行边缘检测，该算法不仅检测了边缘，而且在一定程度上消除了噪声；由于一幅钢管图像中钢管半径相差不大，为精确对钢管进行计数，本文利用距离变换先获取半径的大致范围，而后再利用改进的霍夫变换对钢管进行识别计数，这样不仅可降低霍夫变换参数空间的维数和运行时间，还可以提高钢管的识别率。 本文首先在MATLAB平台上对各个算法的性能进行测试，实验结果表明本文提出的算法可以对大量无序钢管进行准确识别计数；然后在VS2013平台及OpenCV计算机视觉库的基础上实现钢管识别计数软件，利用该软件对钢管图像进行检测，准确率达到97%。

Circular and classes of circular pipe is widely used in geological oil exploration, construction and other industries, due to a large amount of disordered pipe by the light, placement, angle, diameter size and the influence of complex factors such as the working environment, identify and count it has been a complex task. The traditional manual counting method exists obvious shortcomings, the intensity of labor is big, and inefficient. In order to solve this problem, an urgent need to work out the statistical precision automatic equipment. This paper, by using digital image processing and computer vision algorithms for steel pipes to count. For a large number of steel placed disorderly, uneven illumination and multi-shadow phenomenon, an improved local dynamic threshold algorithm is put forward, using the algorithm for image segmentation to obtain a more exact binary image; then the binary image with Gauss-Laplace edge detection operator, the algorithm not only to detect the edge, and eliminated noise to a certain extent; due to the radius of each steel pipe in steel pipe image were similar, for counting precision of steel tube, this paper, by using distance transformation to get approximate scope radius, and then using the hough transform to identify and count steel pipe, so not only can reduce the dimension of hough transform parameter space and running time, also can improve the recognition rate of steel pipe. Firstly, on MATLAB R2010b platform for testing the performance of each algorithm in this paper, experimental results show that the algorithm can identify a large number of steel disorderly counting, the counting system is implemented in a complete basis vs2013 platform and OpenCV computer vision library, the use of the system for detecting steel pipe image, accurate rate of 97%.