基于变步长采样的产品装配结构正确性自动检测技术分析-analysis of automatic inspection technology for correctness of product assembly structure based on variable step sampling.docxVIP

基于变步长采样的产品装配结构正确性自动检测技术研究摘 要随着人们对产品装配结构正确性要求的提高，检测技术也在不断改进，然而产品装 配结构的检测仍然是行业内迄今没有很好解决的难题。现有的对特定产品的检测系统大 多采用了针对最小待识别结构目标的固定旋转步长进行检测，造成了庞大的冗余数据 量，难以满足在线检测的实时性。工程上，对于一些装配结构复杂的精密机电产品，其外表及内部有多个尺寸不一、 形状各异的待检测目标。本文采用变步长采样机制在保证有效信息完整获取的情况下， 利用有限方位下不完全数据达到对结构体产品的快速识别，最终解决在线检测的实时性 问题。在各检测目标最小待识别尺寸确定的前提下，首先通过相关度计算确定各目标的 最优识别区间和旋转步长，并在上述各步长中选取最小值作为合格品的采样步长，保证 标准库中信息的完整性。其次采用改进的尺度不变特征变换（SIFT）算法将待检产品图 像与标准图像进行匹配；结合随机抽样一致算法（RANSAC）及折半查找法，根据精匹 配的点数确定折半查找的范围，最终准确快速确定随机摆放的待检产品在标准库中的位 置信息。最后对 4 种具有代表性的产品在最优识别区间内采用变步长机制进行检测，判 别产品装配结构是否正确。在保证准确率的前提下，对合格品及内部分别存在不同装配结构缺陷的三个不合格 产品进行检测。实验结果表明，基于最优识别区间的变步长检测方法比传统的固定步长 检测平均可节省 0.017s~1.003s 。关键词：装配结构检测，变步长采样，尺度不变特征变换算法，随机抽样一致算法， 折半查找法Research on Automatic Recognition Technology of Products Assembly Structures accuracy Based on Variable Step SamplingAbstractWith the improvement of requirements to products assembly structures accuracy, the detection technology is improving continually. However, the assembly quality testing is still a puzzle in industry has not been well solved. Most of the existing detection systems for specific products used fixed step method to test the objective, which caused enormous data and failed to meet the real-time problem of on-line test.In modern industry, for some assemble complex mechanical and electrical products that are of all shapes and sizes in the surface or internal to be tested. Variable-step sampling mechanism was employed in the case of ensuring the full acquirement of effective information, inadequate data at limited orientation were used to quickly recognize structural body, finally the real-time problem of online test was solved. In the case of minimum recognized size determined of every target to be tested, optimal recognized section and rotating step of every objective were determined according to relativity calculation at first. In order to assure the information integrity of standard libraries, minimum step was chosen as sampling step of qualified product from above-mentioned steps; Then, improve