基于图像识别的管材计数系统方案研究.docVIP

　 　  基于图像识别的管材计数系统方案研究 专业：测控技术与仪器专业 摘要：目前国内对于管材的计数主要采用人工计数的方法，为减轻人工劳动强度和提高管材计数精度，论文基于图像识别对管材的计数系统进行了研究，通过识别管材的截面特征实现管材计数。软件部分主要包括图像的预处理、图像分割和目标计数部分。具体的研究内容包括：若管材截面图像内容较多，先截取对于图像中感兴趣的区域，再经过格式转换、图像增强、去噪等预处理得到边缘平滑的灰度图像，采用大津法自动获取阈值，将目标圆或椭圆与背景分离出来，由于管材之间存在一定空隙，可采用数学形态学中的闭运算以消除部分噪声和狭小间隙，之后通过对目标圆或椭圆的边缘进行计数，因为可能存在较大噪声在闭运算过程中并未消除，因此可通过设置边缘的最小外接圆的面积和边缘本身的面积的比率进行条件限制，以去除部分噪声。若结果存在明显误差，可通过人工修改计数结果以提高精度，系统可保存计数结果和源图像以供日后查证；最后设计软件实现管材计数系统。 关键词：图像识别，管材，计数，图像分割  Pipe Counting System Scheme Studies Based on Image Recognition Major:Measure and Control Technology and Instrument Abstract:In China,the method for pipes counting mainly used manual counting. To alleviate the labor intensity and improve the accuracy,the paper makes a research on pipes counting based on image recognition. The system counts pipes through recognising the feature of the pipes cross-section. The software part mainly include the image preprocessing,the image segmentation and pipes counting. And the paper mainly includes the following content:If the pipe cross-section image include too much content,clipping the region of interest in the image. Then preprocess the image,including converting the image format,the image enhancement,removing noise and so on,and then there will be a grayscale image with smooth edges. Next,use Otsu method to obtain threshold automatically,and get the image isolated target circle or ellipse and background. Since there are some gaps between the pipes,use closed operation of morphology mathematical to remove part of the noise and small gaps. Then count the edge of the circle or ellipse.Because there may be some large area noise that has not been eliminated in the process of closed operation,set the ratio of the area of the minimum circumscribed circle to the edge and its area to remove part of noise.If the result has obvious errors，it can be manually modified in order to improve the accuracy.The source image and the result can be saved for verification i