摘要
针对轮毂生产过程中产生的气孔、缩松、缩孔、裂纹和夹杂缺陷进行识别检测研究。对经过预处理操作后的缺陷图像运用布谷鸟算法结合大津法的CS-Otsu算法,寻找最优阈值T以便缺陷部分与背景分割。通过提取不同类型缺陷的特征数据构建样本数据库。基于萤火虫算法改进BP神经网络的GSO-BP算法优化权值及阈值,并结合构建的缺陷数据库实现对轮毂缺陷类型的检测。通过结果可知运用基于改进遗传算法的正确识别率为95%,高于传统遗传算法,能够满足缺陷检测要求。
Identification and inspection of porosity,shrinkage porosity,shrinkage cavity,cracks and inclusions produced during the production of the wheel hub are conducted.The cuckoo algorithm combined with the CS-Otsu algorithm of the Otsu method is used for the defect image after the preprocessing operation to find the optimal threshold T,so that the defective part is separated from the background.A sample database is constructed by extracting characteristic data of different types of defects.The GSO-BP algorithm based on the firefly algorithm to improve the BP neural network optimizes the weights and thresholds,and combines the built defect database to detect the types of wheel defects.The results show that the correct recognition rate based on the improved genetic algorithm is 95%higher than that of the traditional genetic algorithm,which can meet the requirements of defect detection.
作者
张国胜
张帆
邹洵
张召颖
马保平
Zhang Guosheng;Zhang Fan;Zou Xun;Zhang Zhaoying;Ma Baoping(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;Intelligent Robot R&D Center,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《农业装备与车辆工程》
2021年第2期100-104,共5页
Agricultural Equipment & Vehicle Engineering
