遥感影像在采集过程中,地面覆盖种类数量庞大且采集影像清晰度低、分辨率较差,关键像素特征之间的阈值衡量标准模糊,导致信息提取难度增大,从而降低信息利用率.为此,提出了基于像素紧密程度的多源遥感影像信息提取方法.利用Contourlet变换,实现遥感影像空间域、变换域的多角度增强,优化遥感影像整体清晰度.利用简单线性迭代聚类(Simple Linear Iterative Clustering,SLIC)超像素算法计算像素聚类中心与邻近像素紧密程度,摆脱固定阈值影响.引入灰度共生矩阵(Gray-level Co-occurrenceMa-trix,GLCM),提取主体特征信息;构建相关向量机分类模型,结合拉普拉斯二次逼近回归算法将提取问题转化为噪声回归问题,并对其展开求解,进而实现遥感影像的信息提取.实验结果表明:所提方法对遥感信息主体的分类与真实遥感信息主体分类基本一致,在信息提取过程中的错提取率和漏提取率低,总体提取精度保持在99％以上,且对道路信息提取清晰度高,表明该方法提高了遥感信息的解译水平.
A Method for Extracting Information from Multisource Remote Sensing Images Based on Pixel Tightness
During the collection process of remote sensing images,there are a large number of ground cover types with low clarity and resolution.The threshold measurement standards between key pixel features are blurry,which increases the difficulty of information extraction and reduces information utilization.Therefore,a multi-source remote sensing image in-formation extraction method based on pixel compactness is proposed.Utilizing Contourlet transformation to achieve multi angle enhancement in the spatial and transformation domains of remote sensing images,optimizing the overall clarity of re-mote sensing images.Using the SLIC superpixel algorithm to calculate the closeness between pixel clustering centers and neighboring pixels,eliminating the influence of fixed thresholds.Introducing GLCM gray level co-occurrence matrix to ex-tract subject feature information.The classification model of relevance vector machine is constructed,and the extraction problem is transformed into noise regression problem by combining with the Laplace quadratic approximation regression al-gorithm,and the solution is expanded to realize the information extraction of remote sensing images.The experimental re-sults show that the classification of remote sensing information subjects using the proposed method is basically consistent with the classification of real remote sensing information subjects.The error extraction rate and omission extraction rate are low during the information extraction process,and the overall extraction accuracy remains above 99％.Moreover,the clarity of road information extraction is high,indicating that the method improves the interpretation level of remote sensing infor-mation.
Contourlet transformationSLIC super pixel segmentation methodCIE LAB color spaceGLCM gray-level co-occurrence matrixcorrelation vector machine classification model
国家科技期刊平台
NSTL
万方数据
