自适应烟花寻优的烟丝杂质图像多窗口检测  

Multi Window Detection of Cut Tobacco Impurity Image Based on Adaptive Fireworks Optimization

作　　者：辜志茂 陈加坤 刘银初 吴松涛 GU Zhi-mao;CHEN Jia-kun;LIU Yin-chu;WU Song-tao(Jiangxi Zhongyan Industry Co.,Ltd.Jinggangshan Cigarette Factory,Ji'an Jiangxi 343100,China)

机构地区：[1]江西中烟工业有限责任井冈山卷烟厂,江西吉安343100

出　　处：《计算机仿真》2025年第2期457-461,共5页Computer Simulation

基　　金：江西中烟井冈山卷烟厂科研项目(井烟科2022-05)。

摘　　要：烟丝生产现场环境复杂,所存在的干扰因素较多,使得烟丝杂质区域识别存在偏差,影响了检测精度。因此,提出基于图像识别的烟丝杂质在线检测方法。基于动态时间弯曲的核心理念,获得多维相似距离,并利用自适应烟花寻优算法确定最优聚类中心,实现对烟丝图像杂质区域和正常区域的精确聚类分割。基于分割的杂质区域,运用基于梯度能量的图像识别技术对烟丝杂质展开初检测,以克服环境干扰对识别结果的影响,进一步细化杂质区域。然后,再将HOG特征、LBP特征与级联Adaboost分类器相结合,提出多窗口检测方法完成在线烟丝杂质的二次检测。实验结果表明,所提方法可以获取精准的烟丝杂质在线检测结果,具有良好的检测性能。The field environment of cut tobacco production is complex,with many interference factors,which leads to biases in the identification of tobacco impurity areas and affects the detection accuracy.Therefore,an online detection method of tobacco impurities based on image recognition is proposed.Based on the core idea of dynamic time warping,multi-dimensional similarity distance is obtained,and the optimal clustering center is determined by an adaptive fireworks optimization algorithm,so as to realize accurate clustering segmentation of impurity areas and normal areas of tobacco images.Based on the segmented impurity region,the image recognition technology based on gradient energy is used to detect the tobacco impurities,so as to overcome the influence of environmental interference on the recognition results and further refine the impurity region.Then,combining the HOG feature,LBP feature and cascade Adaboost classifier,a multi-window detection method is proposed to complete the online secondary detection of tobacco impurities.The experimental results show that the proposed method can obtain accurate online detection results of tobacco impurities and has good detection performance.

关 键 词：图像识别 烟丝杂质 在线检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]