一种汽车锁扣铆点的视觉检测方法

A Visual Detection Method for Rivets of Car Lock Catch

摘要:

汽车锁扣是保证汽车安全运行的重要部件,针对传统方法测量锁扣铆点不稳定、精度低的缺点,提出了一种基于机器视觉的检测方法；首先进行相机标定,通过预处理消除拍摄影响,接着提取锁扣的几何特征以建立测量坐标系,并对其进行仿射变换以消除图像几何失真；然后标定铆点所在区域,设置ROI(regions of interest)使铆点从检测背景中分离出来成为一个独立的图像单元,对其进行梯度锐化以增强图像边缘,并依据ROI自身特点对传统锐化方法进行了改进；最后,采用随机Hough变换分别提取改进前后图像铆点的轮廓,测量铆点直径并进行对比；实验表明,改进后图像铆点直径的测量值更加准确,精度达到0.035 mm,满足检测要求。该方法适用于铆点在线高精度检测,具有重要的应用价值。

Abstract:

The car lock catch is an important component for ensuring the safe operation of vehicles. Since traditional methods for measuring the riveting points of lock catch have the disadvantages of instability and low accuracy, a detection method based on machine vision is proposed. Firstly, doing the camera calibration, the shooting impacts are eliminated through preprocessing. And the measurement coordinate system is established by extracting geometric characteristics of the lock catch, utilizing affine transformation to eliminate the geometry distortion of the images. Then the ROI (Regions of Interest) where riveting points lie is set to separate the riveting points from the background to become an independent image unit, the gradient sharpening is adopted to enhance the image edge, and a traditional sharpening method is improved according to the characteristics of ROI. Finally, using randomized Hough transform, the riveting points’ contours are extracted respectively from the images before and after improvement, the riveting points’ diameters are measured and compared. Experiments show that the diameter of riveting points is more accurate in the improved images, and the detection accuracy can reach 0.032 mm, which meets the detection requirement. The proposed method is suitable for online riveting points measurement with high precision, thus has important application value.