多分辨批量古典建筑图像深度学习检索算法
|
摘要
针对传统的模糊BP分类识别方法进行多分辨建筑图像检索误分率较高的问题,提出了一种基于深度学习神经网络分类和多特征融合的多分辨古典建筑图像检索算法。采用小波降噪方法对模糊图像进行降噪处理,对降噪后的图像采用LGB向量量化算法进行特征分解,采用颜色分量融合方法进行图像的信息增强处理,提取图像的灰度不变矩特征量。将提取的特征量输入BP神经网络分类器中,在检索器的隐含层采用深度学习算法进行图像特征聚类的自适应寻优,进行多特征融合处理,避免聚类中心扰动,实现了对批量多分辨古典建筑图像检索的优化。仿真结果表明,采用该算法进行多分辨古典建筑图像检索的准确性较好,抗类间属性扰动能力较强,图像输出的查全率较高,图像检索的时间开销较小。
To solve the problem of high misclassification rate in the traditional fuzzy BP classification method, a multi-resolution classical architectural image retrieval algorithm based on deep learning neural network classification and multi-feature fusion is proposed. The wavelet denoising method is used to reduce the noise of fuzzy image, the LGB vector quantization algorithm is used to decompose the feature of the image, and the color component fusion method is used to enhance the information of the image. The gray invariant moment feature of the image is extracted. The extracted feature quantity is input into the BP neural network classifier, and the depth learning algorithm is used in the hidden layer of the retrieval to self-adaptively search the image feature clustering, and the multi-feature fusion processing is carried out to avoid the clustering center disturbance. The optimization of batch multi-resolution classical architectural image retrieval is realized. The simulation results show that the proposed algorithm has good accuracy, strong ability to resist inter-class attribute disturbance, high recall rate of image output and less time cost for image retrieval.
To solve the problem of high misclassification rate in the traditional fuzzy BP classification method, a multi-resolution classical architectural image retrieval algorithm based on deep learning neural network classification and multi-feature fusion is proposed. The wavelet denoising method is used to reduce the noise of fuzzy image, the LGB vector quantization algorithm is used to decompose the feature of the image, and the color component fusion method is used to enhance the information of the image. The gray invariant moment feature of the image is extracted. The extracted feature quantity is input into the BP neural network classifier, and the depth learning algorithm is used in the hidden layer of the retrieval to self-adaptively search the image feature clustering, and the multi-feature fusion processing is carried out to avoid the clustering center disturbance. The optimization of batch multi-resolution classical architectural image retrieval is realized. The simulation results show that the proposed algorithm has good accuracy, strong ability to resist inter-class attribute disturbance, high recall rate of image output and less time cost for image retrieval.
引文
[1] 代具亭,汤心溢,王世勇,等.扫描型红外焦平面探测器图像实时传输系统[J].激光与红外,2016,46(4):476-480.
[2] 王小玉,张亚洲,陈德运.基于多块局部二值模式特征和人眼定位的人脸检测[J].仪器仪表学报,2014,35(12):2739-2745.
[3] 朱贺,李臣明,张丽丽,等.联合灰度阈值分割及轮廓形态识别的河道提取[J].电子测量与仪器学报,2014,28(11):1288-1296.
[4] 李武周,余锋,王冰,等.基于形态学滤波的红外图像背景补偿[J].红外技术,2016,38(4):333-336.
[5] PASQUALE F,TIZIANO B.Image forgery localization via fine-grained analysis of CFA artifacts[J].IEEE Transactions on Information Forensics and Security,2012,7(5):1566-1577.
[6] TENGFEI L,WEILI J.Automatic line segment registration using Gaussian mixture model and expectation-maximization algorithm[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014,7(5):1688-1699.
[7] SIWEI L,XUNYU P,XING Z.Exposing region splicing forgeries with blind local noise estimation[J].International Journal of Computer Vision,2014,110(2):202-221.
[8] 施晓东,刘格.一种光学遥感图像海陆分割方法[J].国外电子测量技术,2014,33(11):29-32.
[9] GLAS D F,BRI D,MIYASHITA T,et al.SNAPCAT-3D:Calibrating networks of 3D range sensors for pedestrian tracking[C]// IEEE International Conference on Robotics and Automation.At seattle,WA,USA,IEEE,2015:712-719.
[10] TAN D J,TOMBARI F,ILIC S,et al.A versatile learning-based 3D temporal tracker:Scalable,robust,online[C]// IEEE International Conference on Computer Vision.Santiago,Chile,IEEE Computer Society,2015:693-701.
[11] MODER T,HAFNER P,WISIOL K,et al.3D indoor positioning with pedestrian dead reckoning and activity recognition based on Bayes filtering[C]// International Conference on Indoor Positioning and Indoor Navigation.Banff,Canada,IEEE,2015:717-720.
[12] 傅天宇,金柳颀,雷震,等.基于关键点逐层重建的人脸图像超分辨率方法[J].信号处理,2016,32(7):834-841.
[2] 王小玉,张亚洲,陈德运.基于多块局部二值模式特征和人眼定位的人脸检测[J].仪器仪表学报,2014,35(12):2739-2745.
[3] 朱贺,李臣明,张丽丽,等.联合灰度阈值分割及轮廓形态识别的河道提取[J].电子测量与仪器学报,2014,28(11):1288-1296.
[4] 李武周,余锋,王冰,等.基于形态学滤波的红外图像背景补偿[J].红外技术,2016,38(4):333-336.
[5] PASQUALE F,TIZIANO B.Image forgery localization via fine-grained analysis of CFA artifacts[J].IEEE Transactions on Information Forensics and Security,2012,7(5):1566-1577.
[6] TENGFEI L,WEILI J.Automatic line segment registration using Gaussian mixture model and expectation-maximization algorithm[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014,7(5):1688-1699.
[7] SIWEI L,XUNYU P,XING Z.Exposing region splicing forgeries with blind local noise estimation[J].International Journal of Computer Vision,2014,110(2):202-221.
[8] 施晓东,刘格.一种光学遥感图像海陆分割方法[J].国外电子测量技术,2014,33(11):29-32.
[9] GLAS D F,BRI D,MIYASHITA T,et al.SNAPCAT-3D:Calibrating networks of 3D range sensors for pedestrian tracking[C]// IEEE International Conference on Robotics and Automation.At seattle,WA,USA,IEEE,2015:712-719.
[10] TAN D J,TOMBARI F,ILIC S,et al.A versatile learning-based 3D temporal tracker:Scalable,robust,online[C]// IEEE International Conference on Computer Vision.Santiago,Chile,IEEE Computer Society,2015:693-701.
[11] MODER T,HAFNER P,WISIOL K,et al.3D indoor positioning with pedestrian dead reckoning and activity recognition based on Bayes filtering[C]// International Conference on Indoor Positioning and Indoor Navigation.Banff,Canada,IEEE,2015:717-720.
[12] 傅天宇,金柳颀,雷震,等.基于关键点逐层重建的人脸图像超分辨率方法[J].信号处理,2016,32(7):834-841.