基于机器视觉的燃烧碳化面积测量方法研究

Research on measurement method of combustion carbonization area based on machine vision

摘要:

燃烧试验碳化面积测量是标准要求对于带电作业屏蔽服装耐燃性评估的必检项目。针对传统面积测量方法存在人为误差因素多，测量误差大，效率低等问题，提出基于机器视觉的亚像素面积测量新方法。本研究先进行像素级轮廓提取与面积测量，然后在获得像素级轮廓的前提下，提出一种梯度方向确定方法，采用基于灰度值的反正切函数拟合法获取亚像素轮廓，最后利用格林公式计算面积。采用标准图形块和实际碳化样品进行面积测量试验。结果表明，亚像素级面积测量方法误差可达到2%以下，较像素级方法精度提高4倍；亚像素级重复测量标准差为5.02mm2，为传统方法的1/4；机器视觉测量方法一次测量耗时3.6s，较传统方法效率提高20倍以上。为碳化面积测量提供了一种可满足实际需求的简单、准确、高效的新方法。

Abstract:

The measurement of carbonization area in combustion test is a necessary item for the evaluation of flammability of live working shielding clothing. Aiming at the problems of traditional area measurement methods, such as many human error factors, large measurement error and low efficiency, a new sub-pixel area measurement method based on machine vision is proposed. Firstly, the pixel level contour extraction and area measurement are carried out. Then, on the premise of obtaining the pixel level contour, a gradient direction determination method is proposed. The sub-pixel contour is obtained by the arc tangent function fitting method based on the gray value. Finally, the area is calculated by Green's formula. The area measurement experiment was carried out with standard figure block and actual carbonization sample. The results show that the error of sub-pixel area meas-urement method can reach less than2%, which is 4 times higher than that of pixel level method; the standard deviation of sub-pixel level repeated measurement is 5.02 mm2, which is 1/4 of the traditional method; the machine vision measurement method takes 3.6 s for one measurement, which is more than 20 times higher than the traditional method. It provides a simple, accurate and efficient new method for carbonation area measure-ment.