基于机器视觉的服装尺寸在线测量系统

Measurement system of garment dimension based on Machine Vision

在服装尺寸在线测量过程中,针对传统人工测量所带来的误差率高、成本高、效率低等问题,提出了一种基于机器视觉的服装尺寸在线测量系统。服装尺寸在线测量系统从硬件和软件2个方面进行设计。系统硬件部分主要功能是通过CCD相机实现服装图像的采集;系统的软件部分是整个系统的核心,通过角点检测算法对特征点进行提取和定位,针对Forstner算法需要对图像中的每一个像素点进行扫描,从而导致检测速度比较慢的问题,采用SIFT算法先对图像进行快速的筛选,去除一些无关的点,然后运用Forstner算法在初选点集中进行角点提取。通过对提取出的关键角点进行坐标定位分析和比例尺寸测量,得到所测服装的真实值,并且运用友好的人机界面显示出尺寸测量的结果。所设计的系统用于对512×512的256级灰度图像进行检测,尺寸测量的标准误差均小于0.25 mm,重复性精度接近5 mm。实验误差和尺寸测量精度能够达到服装尺寸测量的标准。

In the process of clothes size measurement, aiming at the problems of error rate and low efficiency caused by traditional manual inspection, a new measurement system of garment dimension based on Machine Vision is presented. The hardware and software systems of the measuring system are designed. System hardware main function is to achieve clothes image acquisition by CCD camera. The core of the whole system is software part of the system, the feature point extraction and positioning by corner detection algorithm, aimed at Forstner algorithm need to scans for each image pixel, leading to slower detection problem, SIFT algorithm is used to quickly screening the image, removing some irrelevant point. Then,Forstner algorithm is used to corner extraction from the primaries centralized point. The key corner position is analyzed by the coordinate values positioning, and proportional calculation is used to measure clothes size. Using friendly interface shows the result of the dimension measurement. The system is designed for testing with 256 gray scale image of 512 x 512. The standard error of dimension measurement is less than 0. 25 mm, and the repeat ability of dimension measurement is closely 5 mm. Experimental error can be measured up the clothing standard.