基于智能学习的海量红外激光图像特征挖掘技术  

Feature mining technology of giant infrared laser image based on intelligent learning

作　　者：陆月然[1] 梁碧珍[1] LU Yueran;LIANG Bizhen(Baise University,Baise Guangxi 533000,China)

机构地区：[1]百色学院,广西百色533000

出　　处：《激光杂志》2019年第3期100-104,共5页Laser Journal

基　　金：国家自然科学基金(No.61063046);广西高校科学技术研究重点项目(No.KY2015ZD118)

摘　　要：传统基于FPGA的快速图像特征提取方法,未对图像实施轮廓构建,导致特征挖掘结果不理想,提出基于智能学习的海量红外激光图像特征挖掘方法。构建红外激光图像的活动轮廓模型,对图像实施小波降噪处理,对降噪后的海量红外激光图像进行活动轮廓线套索融合检索,基于检索结果采用SIFT算法实现海量红外激光图像特征挖掘。实验结果表明,所设计方法进行海量红外激光图像降噪的误差小于1%,特征挖掘平均用时约为8. 63 s,特征挖掘准确率高达98%以上,所设计方法能够用于海量红外激光图像特征的准确、高效挖掘。The traditional fast image feature extraction method based on FPGA does not construct the contour of the image,which leads to the unsatisfactory result of feature mining. The active contour model of infrared laser image is constructed,and the image is denoised by wavelet transform. The active contour of the denoised infrared laser image is fused and retrieved. Based on the retrieved results,the SIFT algorithm is used to mine the features of the massive infrared laser image. The experimental results show that the error of denoising is less than 1%,the average time of feature mining is about 8. 63 seconds,and the accuracy of feature mining is over 98%. The method can be used for accurate and efficient feature mining of massive infrared laser images.

关 键 词：海量 红外激光图像 轮廓构建 小波降噪 SIFT算法 特征挖掘 

分 类 号：TN391[电子电信—物理电子学]