结合混沌鸟群阈值分割的铜板表面结瘤缺陷图像检测

【作者】 王卓；

【导师】 张长胜；

【作者基本信息】 昆明理工大学 ， 测试计量技术及仪器， 2021， 硕士

【摘要】 铜电解是铜冶炼过程中的重要工艺,常因多方面电解工艺因素的影响,致使阴极铜板表面出现结瘤缺陷,严重影响其表面质量。考虑到在人工识别该缺陷的过程中,受内、外多方面因素的干扰,致使操作人员对铜板表面结瘤缺陷结果产生误判,影响最终决策合理性。针对上述问题,本文提出一种结合混沌鸟群算法的铜板结瘤缺陷图像识别方案,旨在提高企业生产的智能化,同时降低生产成本。主要工作如下:(1)分析了不同视角下采集铜板图像的差异及其对铜板表面缺陷检测精度的影响,在考虑铜板尺寸、采集图像尺寸的基础上定位匹配特征点,再通过视角变换矫正铜板图像。(2)本文针对传统鸟群算法固有缺陷,选取种群中寻优位置、路径最差个体进行混沌扰动及变步长位置更新,提高算法跳出局部最优位置的概率;针对混沌理论遍历性不足的问题进行改进,并将其引入鸟群算法以增强算法的全局搜索能力;用6种测试函数对本文算法及GA、CSO、FA、BSA等五种算法进行测试,结果表明,本文算法具有较好的寻优性能。(3)选择KSW熵作为各算法的目标函数,利用本文算法及GA、CSO、FA、BSA五种算法分别对多幅铜板图像对进行预分割;针对光照、铜板条纹等因素对铜板图像分割效果的影响,提出8邻域搜索滤波矫正误分类像素,设计基点生长法剔除铜板图像表面纹理,以提高算法对铜板表面结瘤缺陷的检测精度。将分割结果分别在时间、适应度值和结构相似度(SSIM)三个指标下分析比较。结果表明,本文算法适应度平均值可提高0.007～1.707,SSIM值可提高0.0034～0.168。最后,通过计算铜板图像瑕疵类别像素占比,对铜板表面的合格与否做出决策。更多还原

【Abstract】 Copper electrolysis is an important process in copper smelting process.Due to the influence of various electrolytic process factors,nodular defects appear on the surface of cathode copper plate,which seriously affects its surface quality.Considering that in the process of manual identification of the defect,the interference of internal and external factors causes the operator to misjudge the result of the nodule on the surface of the copper plate,which affects the rationality of the final decision.Aiming at the above problems,this paper proposes a nodule defect image recognition scheme of copper plate combined with chaotic bird flock algorithm,aiming at improving the intelligence of enterprise production and reducing the production cost.The main work is as follows:(1)The difference of copper plate images collected from different perspectives and their influence on the detection accuracy of copper plate surface defects are analyzed.The matching feature points are located on the basis of considering the size of copper plate and the size of the collected images,and then the copper plate images are corrected by perspective transformation.(2)Aiming at the inherent defects of the traditional bird flock algorithm,this paper selects the individuals with the optimal location and the worst path in the population to carry out chaotic perturbation and position update with variable step size,so as to improve the probability of the algorithm jumping out of the local optimal position.Aiming at the lack of ergodic property of chaos theory,this paper improves the ergodic property of chaos theory and introduces the bird flock algorithm to enhance the global searching ability of the algorithm.Six test functions are used to test the proposed algorithm and its comparison algorithm,and the results show that the proposed algorithm has better performance.(3)KSW entropy is selected as the objective function of each algorithm,GA,CSO,FA,BSA and the proposed algorithm are used to pre-segment multiple copper plate images.In view of the influence of illumination and copper stripe on the segmentation effect of copper image,an 8-neighborhood search filter is proposed to correct the misclassified pixels,and a base point growth method is designed to remove the surface texture of copper image,so as to improve the detection accuracy of the algorithm for nodular defects on the surface of copper plate.The segmentation results are analyzed and compared under three indexes: time,fitness value and structural similarity(SSIM).The results show that the average fitness of the proposed algorithm can be increased by0.007 ～ 1.707,and the SSIM value can be increased by 0.0034 ～ 0.168.Finally,the quality of the copper surface is determined by calculating the pixel proportion of the copper image defect category.更多还原