基于RSM-S3VM判别模型的员工发型规范监测方法研究
|
摘要
随着新电改全面提速、竞争性电力市场的推进,工作人员的服务形象日益引起电力企业的重视。基于RSM-S3VM判别模型实现员工发型规范的审核与监测。首先基于现有的采集设备所提供的历史采集图像,采用HOG特征与纹理特征相结合的方式确定发质特征,以人员正侧面图像相结合的方式完善发型区域,并以形态学运算加以优化,采用Canny函数获取发型轮廓特征。随后,针对判别模型的构建,采用随机子空间算法解决样本类别不平衡问题,以半监督SVM算法构建判别模型,可实现对批量发型规范与否的判别,进而提高工作人员的服务形象。以国网某营业厅为试验环境,验证了所述方法的时效性与准确性。
With acceleration of the new round of power reform and propulsion of competitive power market,service image of the staff is attracting more and more attention of power enterprises. Based on the RSM-S3 VM discrimination model,this paper introduced an approach for examining and monitoring staff hairstyle standards. Firstly,based on the historical images provided by existing collection equipment,hair quality characteristics were determined by combining HOG characteristics with texture features,hairstyle areas were improved by combining front images with side images,optimization was achieved through morphological operation,and hairstyle contour features were obtained by using Canny function. Then,sample class imbalance was solved through the random subspace algorithm for the construction of the discrimination model in the semi-supervised SVM algorithm. This approach could realize discrimination of hairstyle standards in batch,thus improving service image of the staff. The timeliness and accuracy of the described approach was verified in the environment of a business hall of the state grid.
With acceleration of the new round of power reform and propulsion of competitive power market,service image of the staff is attracting more and more attention of power enterprises. Based on the RSM-S3 VM discrimination model,this paper introduced an approach for examining and monitoring staff hairstyle standards. Firstly,based on the historical images provided by existing collection equipment,hair quality characteristics were determined by combining HOG characteristics with texture features,hairstyle areas were improved by combining front images with side images,optimization was achieved through morphological operation,and hairstyle contour features were obtained by using Canny function. Then,sample class imbalance was solved through the random subspace algorithm for the construction of the discrimination model in the semi-supervised SVM algorithm. This approach could realize discrimination of hairstyle standards in batch,thus improving service image of the staff. The timeliness and accuracy of the described approach was verified in the environment of a business hall of the state grid.
引文
[1]袁立,田子茹.基于融合特征的行人再识别方法[J].模式识别与人工智能,2017,30(3):269-278.
[2]王志一,杨大利.人脸识别中发型遮挡检测方法研究[J].微型机与应用,2016,35(2):32-34.
[3]王思源,王金胜,严玺,等.输电线路状态检测图像的增强算法研究[J].电气自动化,2016(6):77-79.
[4]金遥力,王世亮,徐岗,等.复杂背景下发型轮廓的自动提取方法[J].系统仿真学报,2015,27(10):2305-2309.
[5] KAPIL JUNEJA,NASIB SINGH GILL. A four level aggregative weight scanned model facial quiz image identification[J]. Computer Graphics,Vision and Information Security(CGVIS),2015:214-219.
[6] ABDULLAH-AL-NAHID, YINAN KONG, MD NAZMUL HASAN.Performance analysis of Canny's edge detection method for modified threshold algorithms[C]∥Rajshahi, Bangladesh:2015 International Conference on Electrical&Electronic Engineering(ICEEE),2015:93-99.
[7]支力佳,张少敏.基于排序支持向量机的组合相似度图像检索[J].计算机应用,2017,37(S1):165-168,172.
[8] MARC-ANDRECARBONNEAU, ERIC GRANGER, GHYSLAIN GAGNON. Witness identification in multiple instance learning using random subspaces[C]∥Cancun,Mexico:2016 23rd International Conference on Pattern Recognition(ICPR),2016:3639-3644.
[9]杨帆,林珍.改进滑模观测器的永磁同步电机SVM-DTC系统[J].电气自动化,2015(1):4-6,30.
[10] PABLO QUESADA-BARRIUSO,FRANCISCO ARGELLO,DORA B.HERAS,et al. Wavelet-Based Classification of Hyperspectral Images Using Extended Morphological Profiles on Graphics Processing Units[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2015,8(6):2962-2970.
[11]金旭晖.基于区域颜色分割的交通标志检测和识别[J].电气自动化,2016,38(3):14-16,77.
[2]王志一,杨大利.人脸识别中发型遮挡检测方法研究[J].微型机与应用,2016,35(2):32-34.
[3]王思源,王金胜,严玺,等.输电线路状态检测图像的增强算法研究[J].电气自动化,2016(6):77-79.
[4]金遥力,王世亮,徐岗,等.复杂背景下发型轮廓的自动提取方法[J].系统仿真学报,2015,27(10):2305-2309.
[5] KAPIL JUNEJA,NASIB SINGH GILL. A four level aggregative weight scanned model facial quiz image identification[J]. Computer Graphics,Vision and Information Security(CGVIS),2015:214-219.
[6] ABDULLAH-AL-NAHID, YINAN KONG, MD NAZMUL HASAN.Performance analysis of Canny's edge detection method for modified threshold algorithms[C]∥Rajshahi, Bangladesh:2015 International Conference on Electrical&Electronic Engineering(ICEEE),2015:93-99.
[7]支力佳,张少敏.基于排序支持向量机的组合相似度图像检索[J].计算机应用,2017,37(S1):165-168,172.
[8] MARC-ANDRECARBONNEAU, ERIC GRANGER, GHYSLAIN GAGNON. Witness identification in multiple instance learning using random subspaces[C]∥Cancun,Mexico:2016 23rd International Conference on Pattern Recognition(ICPR),2016:3639-3644.
[9]杨帆,林珍.改进滑模观测器的永磁同步电机SVM-DTC系统[J].电气自动化,2015(1):4-6,30.
[10] PABLO QUESADA-BARRIUSO,FRANCISCO ARGELLO,DORA B.HERAS,et al. Wavelet-Based Classification of Hyperspectral Images Using Extended Morphological Profiles on Graphics Processing Units[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2015,8(6):2962-2970.
[11]金旭晖.基于区域颜色分割的交通标志检测和识别[J].电气自动化,2016,38(3):14-16,77.