非均匀光照下静态手势识别研究  

Static Gesture Recognition Research Under Uneven Illumination

作　　者：卜明龙 黄家海 辛文斌 郝惠敏 BU Ming-long;HUANG Jia-hai;XIN Wen-bin;HAO Hui-min(College of Mechanical and Vehicle Engineering of Taiyuan University of Technology,Shanxi Taiyuan 030024,China;Key Laboratory of Advanced Transducers and Intelligent Control System of Ministry of Education,Taiyuan University of Technology,Shanxi Taiyuan 030024,China)

机构地区：[1]太原理工大学机械与运载工程学院,山西太原030024 [2]新型传感器与智能控制山西省重点实验室,山西太原030024

出　　处：《机械设计与制造》2023年第1期273-277,共5页Machinery Design & Manufacture

基　　金：山西省关键核心技术和共性技术研发攻关专项项目(2020XXX009);山西省重点研发计划(国际合作)(201603D421009)。

摘　　要：在工业生产任务中,实际操作环境通常比较复杂,常常导致手势识别的准确率降低。为提高手势识别对环境的适应能力并提高识别的实时性,分别在预处理,特征提取及分类识别三个方面进行研究。首先采用改进的同态滤波算法进行图像的增强预处理,然后提取增强后图像的梯度方向直方图(HOG)特征,采用主元分析(PCA)方法对其进行降维,并将降维后的特征输入到支持向量机(SVM)中进行分类。结果表明,改进后的同态滤波算法能较好克服光照不均导致的手势分割困难问题,使识别率从93.2%提高到了95.6%。而PCA结合HOG使每张图像的分类时间从18.07ms缩短到降维后的1.43ms,在大幅提高识别速度的同时,识别精度几乎不受影响。In industrial production tasks,the actual operating environment is usually complicated,resulting in the greatly reduced accuracy of gesture recognition. In order to improve the ability of adapting to the environment and recognition accuracy,preprocessing,features extraction and classification aspects are studied. First,an improved homomorphic filtering algorithm is introduced to enhance the input images,then the features of histogram of oriented gradient(HOG)are extracted,and the principal component analysis(PCA)method is used to reduce the features dimensions,and the reduced dimension features are input to the support vector machine(SVM)for classification. The results show that the homomorphic filtering algorithm can better overcome the difficulty of gesture segmentation caused by uneven illumination and improves recognition rate from 93.2% to 95.6%.The combination of PCA and HOG shortens the classification time of each image from 18.07ms to 1.43ms after dimensionality reduction. While greatly improving the recognition speed,the recognition accuracy is almost not affected.

关 键 词：工业生产 手势识别 同态滤波 主元分析 

分 类 号：TH16[机械工程—机械制造及自动化] TH164[自动化与计算机技术—计算机应用技术] TP391.4[自动化与计算机技术—计算机科学与技术]