高分辨卫星图像卷积神经网络分类模型

目的 卫星图像往往目标、背景复杂而且带有噪声,因此使用人工选取的特征进行卫星图像的分类就变得十分困难。提出一种新的使用卷积神经网络进行卫星图像分类的方案。使用卷积神经网络可以提取卫星图像的高层特征,进而提高卫星图像分类的识别率。方法 首先,提出一个包含六类图像的新的卫星图像数据集来解决卷积神经网络的有标签训练样本不足的问题。其次,使用了一种直接训练卷积神经网络模型和3种预训练卷积神经网络模型来进行卫星图像分类。直接训练模型直接在文章提出的数据集上进行训练,预训练模型先在ILSVRC（the ImageNet large scale visual recognition challenge）-2012数据集上进行预训练,然后在提出的卫星图像数据集上进行微调训练。完成微调的模型用于卫星图像分类。结果 提出的微调预训练卷积神经网络深层模型具有最高的分类正确率。在提出的数据集上,深层卷积神经网络模型达到了99.50%的识别率。在数据集UC Merced Land Use上,深层卷积神经网络模型达到了96.44%的识别率。结论 本文提出的数据集具有一般性和代表性,使用的深层卷积神经网络模型具有很强的特征提取能力和分类能力,且是一种端到端的分类模型,不需要堆叠其他模型或分类器。在高分辨卫星图像的分类上,本文模型和对比模型相比取得了更有说服力的结果。     

Semi-automatic detection and counting of oil palm trees from high spatial resolution airborne imagery

Plantation inventory and management require a range of fine-scale remote-sensing data. Remote-sensing images with high spatial and spectral resolution are an efficient source of such information. This article presents an approach to the extraction and counting of oil palm trees from high spatial resolution airborne imagery data. Counting oil palm trees is a crucial problem in specific agricultural areas, especially in Malaysia. The proposed scheme comprises six major parts: (1) discrimination of oil palms from non-oil palms using spectral analysis, (2) texture analysis, (3) edge enhancement, (4) segmentation process, (5) morphological analysis and (6) blob analysis. The average accuracy obtained was 95%, which indicates that high spatial resolution airborne imagery data with an appropriate assessment technique have the potential to provide us with vital information for oil palm plantation management. Information on the number of oil palm trees is crucial to the ability of plantation management to assess the value of the plantation and to monitor its production.     

Application of high spatial resolution satellite imagery for riparian and forest ecosystem classification

Terrestrial Ecosystem Mapping provides critical information to land and resource managers by incorporating information on climate, physiography, surficial material, soil, and vegetation structure. The main objective of this research was to determine the capacity of high spatial resolution satellite image data to discriminate vegetation structural stages in riparian and adjacent forested ecosystems as defined using the British Columbia Terrestrial Ecosystem Mapping (TEM) scheme. A high spatial resolution QuickBird image, captured in June 2005, and coincident field data covering the riparian area of Lost Shoe Creek and adjacent forests on Vancouver Island, British Columbia, was used in this analysis. Semi-variograms were calculated to assess the separability of vegetation structural stages and assess which spatial scales were most appropriate for calculation of grey-level co-occurrence texture measures to maximize structural class separation. The degree of spatial autocorrelation showed that most vegetation structural types in the TEM scheme could be differentiated and that window sizes of 3 × 3 pixels and 11 × 11 pixels were most appropriate for image texture calculations. Using these window sizes, the texture analysis showed that co-occurrence contrast, dissimilarity, and homogeneity texture measures, based on the bands in the visible part of the spectrum, provided the most significant statistical differentiation between vegetation structural classes. Subsequently, an object-oriented classification algorithm was applied to spectral and textural transformations of the QuickBird image data to map the vegetation structural classes. Using both spectral and textural image bands yielded the highest classification accuracy (overall accuracy = 78.95%). The inclusion of image texture increased the classification accuracies of vegetation structure by 2-19%. The results show that information on vegetation structure can be mapped effectively from high spatial resolution satellite image data, providing an additional tool to ongoing aerial photograph interpretation.     

Object-based urban detailed land cover classification with high spatial resolution IKONOS imagery

Improvement in remote sensing techniques in spatial/spectral resolution strengthens their applicability for urban environmental study. Unfortunately, high spatial resolution imagery also increases internal variability in land cover units and can cause a ‘salt-and-pepper’ effect, resulting in decreased accuracy using pixel-based classification results. Region-based classification techniques, using an image object (IO) rather than a pixel as a classification unit, appear to hold promise as a method for overcoming this problem. Using IKONOS high spatial resolution imagery, we examined whether the IO technique could significantly improve classification accuracy compared to the pixel-based method when applied to urban land cover mapping in Tampa Bay, FL, USA. We further compared the performance of an artificial neural network (ANN) and a minimum distance classifier (MDC) in urban detailed land cover classification and evaluated whether the classification accuracy was affected by the number of extracted IO features. Our analysis methods included IKONOS image data calibration, data fusion with the pansharpening (PS) process, Hue–Intensity–Saturation (HIS) transferred indices and textural feature extraction, and feature selection using a stepwise discriminant analysis (SDA). The classification results were evaluated with visually interpreted data from high-resolution (0.3 m) digital aerial photographs. Our results indicate a statistically significant difference in classification accuracy between pixel- and object-based techniques; ANN outperforms MDC as an object-based classifier; and the use of more features (27 vs. 9 features) increases the IO classification accuracy, although the increase is statistically significant for the MDC but not for the ANN.     

Procedures for correcting high resolution airborne video imagery

Airborne videography is a useful way of producing very high resolution remotely-sensed data. Before video data can be used for digital analysis, individual frames must be correcled for geometric distortion and brightness variation. The most obvious geometric distortion occurs because the odd and even fields in a video frame are collected sequentially and can be displaced relative to each other by aircraft roll and forward motion. We describe a procedure for correcting this distortion based on the cross-correla lion between individual fields at different spatial lags. Brightness variation across frames occurs because of dilTerences in viewing geometry, bi-directional reflectance variation and atmospheric scattering. We describe a method for removing much of this variation by calculating scattering angle across the image with an optional adjustment for sensor plane tilt. Relations between scattering angle and scene brightness statistics may then be calculated from sequences of images collected along a given transect. These relations can be used 10 normalise brightness. Tests of normalisation procedures based on mean, standard deviation, and median brightness show that the median gives the best results. This approach also produces better results than commonly-used band ratioing procedures.     

Scene classification based on a hierarchical convolutional sparse auto-encoder for high spatial resolution imagery

Efficiently representing and recognizing the semantic classes of the subregions of large-scale high spatial resolution (HSR) remote-sensing images are challenging and critical problems. Most of the existing scene classification methods concentrate on the feature coding approach with handcrafted low-level features or the low-level unsupervised feature learning approaches, which essentially prevent them from better recognizing the semantic categories of the scene due to their limited mid-level feature representation ability. In this article, to overcome the inadequate mid-level representation, a patch-based spatial-spectral hierarchical convolutional sparse auto-encoder (HCSAE) algorithm, based on deep learning, is proposed for HSR remote-sensing imagery scene classification. The HCSAE framework uses an unsupervised hierarchical network based on a sparse auto-encoder (SAE) model. In contrast to the single-level SAE, the HCSAE framework utilizes the significant features from the single-level algorithm in a feedforward and full connection approach to the maximum extent, which adequately represents the scene semantics in the high level of the HCSAE. To ensure robust feature learning and extraction during the SAE feature extraction procedure, a ‘dropout’ strategy is also introduced. The experimental results using the UC Merced data set with 21 classes and a Google Earth data set with 12 classes demonstrate that the proposed HCSAE framework can provide better accuracy than the traditional scene classification methods and the single-level convolutional sparse auto-encoder (CSAE) algorithm.     

Per-pixel vs. object-based classification of urban land cover extraction using high spatial resolution imagery

In using traditional digital classification algorithms, a researcher typically encounters serious issues in identifying urban land cover classes employing high resolution data. A normal approach is to use spectral information alone and ignore spatial information and a group of pixels that need to be considered together as an object. We used QuickBird image data over a central region in the city of Phoenix, Arizona to examine if an object-based classifier can accurately identify urban classes. To demonstrate if spectral information alone is practical in urban classification, we used spectra of the selected classes from randomly selected points to examine if they can be effectively discriminated. The overall accuracy based on spectral information alone reached only about 63.33%. We employed five different classification procedures with the object-based paradigm that separates spatially and spectrally similar pixels at different scales. The classifiers to assign land covers to segmented objects used in the study include membership functions and the nearest neighbor classifier. The object-based classifier achieved a high overall accuracy (90.40%), whereas the most commonly used decision rule, namely maximum likelihood classifier, produced a lower overall accuracy (67.60%). This study demonstrates that the object-based classifier is a significantly better approach than the classical per-pixel classifiers. Further, this study reviews application of different parameters for segmentation and classification, combined use of composite and original bands, selection of different scale levels, and choice of classifiers. Strengths and weaknesses of the object-based prototype are presented and we provide suggestions to avoid or minimize uncertainties and limitations associated with the approach.     

Selecting the spatial resolution of airborne MSS imagery for small-scale agricultural mapping

The spatial resolution determines the number of data and amount of information in a remotely sensed image of a given scene. The 'optimal' spatial resolution may be defined as that which maximizes the information per pixel, and this maximum is realized when the semivariance at a lag of one pixel (the average squared difference between neighbouring pixels) is maximized. For mapping, a spatial resolution should be chosen that is much finer than the 'optimal' spatial resolution as defined above. Airborne MSS images in both red and near-infrared wavelengths for three different dates and two sites were investigated to determine a spatial resolution suitable for mapping spatial variation in agricultural fields in the U.K. The spatial resolution most appropriate for mapping the spatial variation in the images was between 0.5 m and 3 m.     

Moment-based feature extraction from high spatial resolution hyperspectral images

In this article, we propose a method for extracting spatio-spectral features from high spatial resolution hyperspectral (HS) images. The method is based on extracting two-dimensional moments from neighbourhoods of pixels. Three different types of moments are considered: geometric, complex Zernike and Legendre. Moments of a given type are extracted from a few principal components (PC) of HS data, and are stacked on the original HS data to form a joint spatio-spectral feature space. These features are classified using a support vector machine (SVM) classifier. The influence of the moments orders and the size of the neighbourhood window on the quality of the extracted features are analysed. A few experiments are conducted on two widely used HS data sets, Pavia University and Salinas. The results demonstrate high capabilities of the proposed method in comparison with some state-of-the-art spatio-spectral HS classification methods.     

Applying image transformation and classification techniques to airborne hyperspectral imagery for mapping Ashe juniper infestations

Ashe juniper (Juniperus ashei Buchholz) in excessive coverage reduces forage production, interferes with livestock management, and degrades watersheds and wildlife habitat on infested rangelands. The objective of this study was to apply minimum noise fraction (MNF) transformation and different classification techniques to airborne hyperspectral imagery for mapping Ashe juniper infestations. Hyperspectral imagery with 98 usable bands covering a spectral range of 475–845 nm was acquired from two Ashe juniper infested sites in central Texas. MNF transformation was applied to the hyperspectral imagery and the transformed imagery with the first 10 and 20 MNF bands was classified using four hard classifiers: minimum distance, Mahalanobis distance, maximum likelihood and spectral angle mapper (SAM). For comparison, the 10‐ and 20‐band MNF imagery was inversely transformed to noise‐reduced 98‐band imagery in the original data space, which was also classified using the four classifiers. Accuracy assessment showed that the first 10 MNF bands were sufficient for distinguishing Ashe juniper from associated plant species (mixed woody species and mixed herbaceous species) and other cover types (bare soil and water). Although the 20‐band MNF imagery provided better results for some classifications, the increase in overall accuracy was not statistically significant. Overall accuracy on the 10‐band MNF imagery varied from 88% for SAM to 93% for minimum distance for site 1 and from 84% for SAM to 94% for maximum likelihood for site 2. The 98‐band imagery derived from the 10‐band MNF imagery resulted in overall accuracy ranging from 91% for both SAM and Mahalanobis distance to 97% for maximum likelihood for site 1 and from 87% for SAM to 93% for minimum distance for site 2. Although both approaches produced comparable classification results, the MNF imagery required smaller storage space and less computing time. These results indicate that airborne hyperspectral imagery incorporated with image transformation and classification techniques can be a useful tool for mapping Ashe juniper infestations.     

Building detection in high spatial resolution remote sensing imagery with the U-Rotation Detection Network

Building detection in high spatial resolution optical remote sensing images is important for city planning, navigation, population estimation and many other applications. Although many methods have been proposed, building detection is still a challenging problem due to complex scenes and small or arbitrarily orientated buildings. Moreover, most algorithms detect rotated buildings with horizontal bounding boxes leading to many background pixels being preserved in the final detection, which is not beneficial for post-processing. To address these problems, we present the U-Rotation Detection Network (U-RDN), which can effectively detect buildings with arbitrarily orientated detection bounding boxes. First, the U-Rotation Region Proposal Network (U-RRPN) is proposed to generate rotated proposals through rotated anchors. Then, a Rotation Fast-Region Convolutional Neural Network (RFast-RCNN) is performed, which extracts fixed-size features from rotated proposals and utilizes them to obtain fine-detections. For extracting fixed-size features from rotated proposals, we propose Auto Mask Region-Of-Interest Align (AM-ROI Align). The AM-ROI Align not only reduces abundant noise but also preserves the proper information of an object in ROI. Experimental results using the public building dataset, SpaceNet, show that our method can detect buildings with skewed bounding boxes and has a state-of-the-art performance compared with other algorithms.     

Extracting impervious surfaces from medium spatial resolution multispectral and hyperspectral imagery: a comparison

Remote sensing estimation of impervious surfaces is significant in monitoring urban development and determining the overall environmental health of a watershed, and has therefore recently attracted increasing interest. The main objective of this study was to develop a general approach to estimating and mapping impervious surfaces by using medium spatial resolution satellite imagery. We have applied spectral mixture analysis (SMA) to Earth Observing 1 (EO‐1) Advanced Land Imager (ALI) (multispectral) and Hyperion (hyperspectral) imagery in Marion County, Indiana, USA, to calculate the fraction images of vegetation, soil, high albedo and low albedo. The effectiveness of the two images was compared according to three criteria: (1) high‐quality fraction images for the urban landscape, (2) relatively low error, and (3) the distinction among typical land use and land cover (LULC) types in the study area. The fraction images were further used to estimate and map impervious surfaces. The accuracy of the estimated impervious surface was checked against Digital Orthophoto Quarter Quadrangle (DOQQ) images. The results indicate that both ALI and Hyperion sensors were effective in deriving the fraction images with SMA and in computing impervious surfaces. The SMA results for both ALI and Hyperion images using four endmembers were excellent, with a mean root mean square error (RMSE) less than 0.04 in both cases. The ALI‐derived impervious surface image yielded an RMSE of 15.3%, and the Hyperion‐derived impervious surface image yielded an RMSE of 17.5%. However, the Hyperion image was more powerful in discerning low‐albedo surface materials, which has been a major obstacle for impervious surface estimation with medium resolution multispectral images. A sensitivity analysis of the mapping of impervious surfaces using different scenarios of Hyperion band combinations suggests that the improvement in mapping accuracy in general and the better ability in discriminating low‐albedo surfaces came mainly from additional bands in the mid‐infrared region.     

Multi-temporal analysis of high spatial resolution imagery for disturbance monitoring

Mountain pine beetle red attack damage has been successfully detected and mapped using single-date high spatial resolution (< 4 m) satellite multi-spectral data. Forest managers; however, need to monitor locations for changes in beetle populations over time. Specifically, counts of individual trees attacked in successive years provide an indication of beetle population growth and dynamics. Surveys are typically used to estimate the ratio of green (current) attack trees to red (previous) attack trees or G:R. In this study, we estimate average stand-level G:R using a time-series of QuickBird multi-spectral and panchromatic satellite data, combined with field data for three forested stands near Merritt, British Columbia, Canada. Using a ratio of QuickBird red to green wavelengths (Red-Green Index or RGI), the change in RGI (ΔRGI) in successive image pairs is used to estimate red attack damage in 2004, 2005, and 2006, with true positive accuracies ranging from 89 to 93%. To overcome issues associated with differing viewing geometry and illumination angles that impair tracking of individual trees through time, segments are generated from the QuickBird multi-spectral data to identify small groups of trees. These segments then serve as the vehicle for monitoring changes in red attack damage over time. A local maxima filter is applied to the panchromatic data to estimate stem counts, thereby allowing an indication of the total stand population at risk of attack. By combining the red attack damage estimates with the local maxima stem counts, predictions are made of the number of attacked trees in a given year. Backcasting the current year's red attack damaged trees as the previous year's green attack facilitates the estimation of an average stand G:R. In this study area, these retrospective G:R values closely match those generated from field surveys. The results of this study indicate that a monitoring program using a time series of high spatial resolution remotely sensed data (multi-spectral and panchromatic) over select sample locations, could be used to estimate G:R over large areas, facilitating landscape level management strategies and/or providing a mechanism for assessing the efficacy of previously implemented strategies.     

Geostatistical and local cluster analysis of high resolution hyperspectral imagery for detection of anomalies

This paper describes a new methodology to detect small anomalies in high resolution hyperspectral imagery, which involves successively: (1) a multivariate statistical analysis (principal component analysis, PCA) of all spectral bands; (2) a geostatistical filtering of noise and regional background in the first principal components using factorial kriging; and finally (3) the computation of a local indicator of spatial autocorrelation to detect local clusters of high or low reflectance values and anomalies. The approach is illustrated using 1 m resolution data collected in and near northeastern Yellowstone National Park. Ground validation data for tarps and for disturbed soils on mine tailings demonstrate the ability of the filtering procedure to reduce the proportion of false alarms (i.e., pixels wrongly classified as target), and its robustness under low signal to noise ratios. In almost all scenarios, the proposed approach outperforms traditional anomaly detectors (i.e., RX detector which computes the Mahalanobis distance between the vector of spectral values and the vector of global means), and fewer false alarms are obtained when using a novel statistic S₂ (average absolute deviation of p-values from 0.5 through all spectral bands) to summarize information across bands. Image degradation through addition of noise or reduction of spectral resolution tends to blur the detection of anomalies, increasing false alarms, in particular for the identification of the least pure pixels. Results from a mine tailings site demonstrate the approach performs reasonably well for highly complex landscape with multiple targets of various sizes and shapes. By leveraging both spectral and spatial information, the technique requires little or no input from the user, and hence can be readily automated.     

Spatial-spectral-combined sparse representation-based classification for hyperspectral imagery

Recently, sparse representation-based classification (SRC), which assigns a test sample to the class with minimum representation error via a sparse linear combination of all the training samples, has successfully been applied to hyperspectral imagery. Alternatively, spatial information, which means the adjacent pixels belong to the same class with a high probability, is a valuable complement to the spectral information. In this paper, we have presented a new spectral-spatial-combined SRC method, abbreviated as SSSRC or \(\mathrm{S}^{3}\mathrm{RC}\), to jointly consider the spectral and spatial neighborhood information of each pixel to explore the spectral and spatial coherence by the SRC method. Furthermore, a fast interference-cancelation operation is adopted to accelerate the classification procedure of \(\mathrm{S}^{3}\mathrm{RC}\), named \(\mathrm{FS}^{3}\mathrm{RC}\). Experimental results have shown that both the proposed SRC-based approaches, \(\mathrm{S}^{3}\mathrm{RC}\) and \(\mathrm{FS}^{3}\mathrm{RC}\), could achieve better performance than the other state-of-the-art methods.     

Correction of atmospheric and topographic effects for high spatial resolution satellite imagery

A method for the combined correction of atmospheric and topographic effects has been developed. It accounts for horizontally varying atmospheric conditions and also includes the height dependence of the atmospheric radiance and transmittance functions to simulate the simplified properties of a threedimensional atmosphere. A Digital Elevation Model (DEM) is used to obtain information about surface elevation, slope, and orientation. Based on the Lambertian assumption the surface reflectance in rugged terrain is calculated. The method is restricted to high spatial resolution satellite sensors like Landsat TM and SPOT HRV, since some simplifying assumptions are being made to reduce the required image processing time. The possibilities and limitations of the method are critically discussed.     

The effect of spectral and spatial degradation of hyperspectral imagery for the Sclerophyll tree species classification

Hyperspectral imaging can be a useful remote-sensing technology for classifying tree species. Prior to the image classification stage, effective mapping endeavours must first identify the optimal spectral and spatial resolutions for discriminating the species of interest. Such a procedure may contribute to improving the classification accuracy, as well as the image acquisition planning. In this work, we address the effect of degrading the original bandwidth and pixel size of a hyperspectral and hyperspatial image for the classification of Sclerophyll forest tree species. A HySpex-VNIR 1600 airborne-based hyperspectral image with submetric spatial resolution was acquired in December 2009 for a native forest located in the foothills of the Andes of central Chile. The main tree species of this forest were then sampled in the field between January and February 2010. The original image spectral and spatial resolutions (160 bands with a width of 3.7 nm and pixel sizes of 0.3 m) were systematically degraded by resampling using a Gaussian model and a nearest neighbour method, respectively (until reaching 39 bands with a width of 14.8 nm and pixel sizes of 2.4 m). As a result, 12 images with different spectral and spatial resolution combinations were created. Subsequently, these images were noise-reduced using the minimum noise fraction procedure and 12 additional images were created. Statistical class separabilities from the spectral divergence measure and an assessment of classification accuracy of two supervised hyperspectral classifiers (spectral angle mapper (SAM) and spectral information divergence (SID)) were applied for each of the 24 images. The best overall and per-class classification accuracies (>80%) were observed when the SAM classifier was applied on the noise-reduced reflectance image at its original spectral and spatial resolutions. This result indicates that pixels somewhat smaller than the tree canopy diameters were the most appropriate to represent the spatial variability of the tree species of interest. On the other hand, it suggests that noise-reduced bands derived from the full image spectral resolution rendered the best discrimination of the spectral properties of the tree species of interest. Meanwhile, the better performance of SAM over SID may result from the ability of the former to classify tree species regardless of the illumination differences in the image. This technical approach can be particularly useful in native forest environments, where the irregular surface of the uppermost canopy is subject to a differentiated illumination.     

Multivariate forest structure modelling and mapping using high resolution airborne imagery and topographic information

Remote sensing has been widely used for modelling and mapping individual forest structural attributes, such as LAI and stem density, however the development and evaluation of methods for simultaneously modelling and mapping multivariate aspects of forest structure are comparatively limited. Multivariate representation of forest structure can be used as a means to infer other environmental attributes such as biodiversity and habitat, which have often been shown to be enhanced in more structurally diverse or complex forests. Image-based modelling of multivariate forest structure is useful in developing an understanding of the associations between different aspects of vertical and horizontal structure and image characteristics. Models can also be applied spatially to all image pixels to produce maps of multivariate forest structure as an alternative to sample-based field assessment. This research used high spatial resolution multispectral airborne imagery to provide spectral, spatial, and object-based information in the development of a model of multivariate forest structure as represented by twenty-four field variables measured in plots within a temperate hardwood forest in southern Quebec, Canada. Redundancy Analysis (RDA) was used to develop a model that explained a statistically significant proportion of the variance of these structural attributes. The resulting model included image variables representing mostly within-crown and within-shadow brightness variance (texture) as well as elevation, taken from a DEM of the study area. It was applied spatially across the entire study area to produce a continuous map of predicted multivariate forest structure. Bootstrapping validation of the model provided an RMSE of 19.9%, while independent field validation of mapped areas of complex and simple structure showed accuracies of 89% and 69%, respectively. Multiscale testing using resampled imagery suggested that the methods could possibly be used with current pan-sharpened, or future sub-metre resolution, multispectral satellite imagery, which would provide much greater spatial coverage and reduced image processing compared to airborne imagery.     

Detecting Tamarisk species (Tamarix spp.) in riparian habitats of Southern California using high spatial resolution hyperspectral imagery

Exotic plant invasion is a major environmental and ecological concern and is a particular issue for Mediterranean-type ecosystems. Early detection of invasive plants is crucial for effective weed management. Several studies have explored hyperspectral imagery for mapping invasive plants with promising results. However, only a few extensive or comparative studies about image processing techniques for invasive plant detection have been reported, and even fewer studies have involved very high spatial and spectral resolution imagery. The primary goal of this study was to investigate the utility of very high spatial (0.5 m) and spectral (4 nm) resolution imagery and several classification approaches for detecting tamarisk (Tamarix spp.) infestations, the most problematic invasive plant species in the riparian habitats of southern California.Hierarchical clustering was a particularly effective and efficient statistical method for identifying wavebands and spectral transforms having the greatest discriminatory power. Products resulting from the classification of airborne hyperspectral image data varied by scene, input data type, classifier, and minimum patch size. Overall accuracy of image classification accuracy of products co-varied with commission error rates, such that products having strong agreement with reference data also had a high number of false detections. Integrating the findings from qualitative map analysis, areal proportion statistics, and object-based accuracy assessment indicates that the parallelepiped classifier with several narrow wavebands selected through hierarchical clustering yielded the most accurate and reliable tamarisk classification products.     

Haze detection,perfection and removal for high spatial resolution satellite imagery

We present a technique to remove spatially varying haze contamination for high spatial resolution satellite imagery. This technique comprises three steps: haze detection, haze perfection and haze removal. Background Suppressed Haze Thickness Index (BSHTI) in haze detection is used to indicate relative haze thickness. ‘Fill sink’ and ‘flatten peak’ routines in haze perfection are applied to correct some spurious background effects. Virtual Cloud Point (VCP) method based on BSHTI is used in haze removal. Case study using two QuickBird images (hazy and clear) of Shenyang City in China proves the effectiveness of this technique except for those regions where haze is too thick. Comparison of the overlapped region between hazy and clear images using 76 paired polygon samples shows that squared correlation coefficient of each band between the two images becomes larger than 0.7. The advantages of this technique are that aerosol transparent bands are not needed and the technique is suitable for urban remote sensing.