微型齿轮的机器视觉检测系统设计

MICRO-GEAR MACHINE VISION INSPECTION SYSTEM

摘要:

微型齿轮在精密仪器设备中经常使用，其齿长的误差大小与整个仪器的精密准确性密切相关。传统微型齿轮检测以人工检测为主，人工检测存在检测精度低、量化不准确的问题。为了能够精确的计算齿长误差，并给出量化结果，文章提出一种基于机器视觉的微型齿轮齿长误差检测系统，先通过小波变换去除图像噪声，然后使用Radon变换算法矫正零件图像,再使用一种基于局部区域特征的三次曲线模型提取感兴趣区域亚像素边缘信息，并通过投影映射精确计算边界位置，最后计算齿轮中心点的动态极差并以此数据作为判断齿长是否合格的标准。实验结果表明，该方法精度可达到2 μm,准确率可达到99%，单帧检测时间平均18 ms，一个零件大约5 s可以给出可靠的结论，该方法效率高，准确性好，能够满足工业检测的要求。

Abstract:

Micro gears are often used in precise instruments and devices, and the error of their tooth length is closely related to the precision of an entire instrument. Traditional gear detection is mainly performed manually,however, low precision and inaccurate quantification exist in manual detection. To accurately calculate the tooth length error and obtain quantified results, this paper proposed a system to detect the tooth length error of micro gears based on machine vision. In this system, firstly, the image noise was removed by wavelet transform; then Radon transform method was adopted to correct the part image; a cubic curve model based on local characteristics was employed to extract the sub-pixel edge information of an interested area, and the boundary location was accurately calculated through projection mapping; finally, the dynamic range of the gear center was calculated to judge the tooth length. The experimental results show that this method can reach a precision up to 2 μm, an accuracy rate up to 99%, and an average single-frame detection time of 18 ms, as well as obtain reliable conclusions of a part about 5s. With efficiency and accuracy, this method can meet the requirements of industrial detection.