• Korean Journal of Remote Sensing
• /
• v.24 no.5
• /
• pp.507-515
• /
• 2008

Many studies have generally used a large number of pure pixels as an approach to training set design. The training set are used, however, varies between classifiers. In the recent research, it was reported that small mixed pixels between classes are actually more useful than larger pure pixels of each class in Support Vector Machine (SVM) classification. We evaluated a usability of small mixed pixels as a training set for the classification of high-resolution satellite imagery. We presented an advanced approach to obtain a mixed pixel readily, and evaluated the appropriateness with the land cover classification from IKONOS satellite imagery. The results showed that the accuracy of the classification based on small mixed pixels is nearly identical to the accuracy of the classification based on large pure pixels. However, it also showed a limitation that small mixed pixels used may provide insufficient information to separate the classes. Small mixed pixels of the class border region provide cost-effective training sets, but its use with other pixels must be considered in use of high-resolution satellite imagery or relatively complex land cover situations.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.6
• /
• pp.2751-2756
• /
• 2012

Selection of training set influences the classification accuracy in supervised classification using satellite imagery. Generally, if pure pixels which character of training set is clear were selected, whole accuracy is high while if mixed pixels were selected, accuracy is decreased because of low-resolution imagery or unclear distinguishment. However, it is too difficult to choose the pure pixels as training set actually. Accordingly, this study should be suggested the suitable classification method in case of mixed pixels choice. To achieve this, a few pure pixels were chosen as training set and classification accuracy was calculated which was compared with classification result using an equal number of mixed pixels. As a result, accuracy of SVM was the highest among the classification method using mixed pixels and it was a relatively small difference with the result of classification using pure pixels. Therefore, imagery classification using SVM is most suitable in the mixed area of construction and green because it is high possibility to choose mixed pixels as training set.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.16 no.1
• /
• pp.75-84
• /
• 1998

To achieve more accurate information from satellite image data, a research on a technique for mixed pixel ex-traction has been produced. The mixed pixels with only two land covers have been experimented. By analyzing canonical vector in canonical correlation classification, the mixed pixels have been classified. The ratio of the two canonical weighted values-the elements of canonical vector have been used as a threshold to discriminate mixed pixels. In case of the classification for the mixed pixels of bridge and water class in TM data before or after the 1st of September, the threshold for the optimal classification of the mixed pixels is 4.0. That is, if the ratio of the two canonical weighted values is less than 4.0, the pixel is a mixed pixel. Also, using the distribution of canonical weighted values, the constitution percentages of land covers within one mixed pixel can be approximately deducted. The accuracy of mixed pixel extraction for experimental area is 90％ and quite acceptable. Conclusively, a technique for mixed pixel extraction by canonical vector analysis is effective.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.386-389
• /
• 2008

It is important that the training stage of a supervised classification is designed to provide the spectral information. On the design of the training stage of a classification typically calls for the use of a large sample of randomly selected pure pixels in order to characterize the classes. Such guidance is generally made without regard to the specific nature of the application in-hand, including the classifier to be used. An approach to the training of a support vector machine (SVM) classifier that is the opposite of that generally promoted for training set design is suggested. This approach uses a small sample of mixed spectral responses drawn from purposefully selected locations (geographical boundaries) in training. A sample of such data should, however, be easier and cheaper to acquire than that suggested by traditional approaches. In this research, we evaluated them against traditional approaches with high-resolution satellite data. The results proved that it can be used small mixed pixels to derive a classification with similar accuracy using a large number of pure pixels. The approach can also reduce substantial costs in training data acquisition because the sampling locations used are commonly easy to observe.

• Journal of Digital Contents Society
• /
• v.18 no.1
• /
• pp.175-182
• /
• 2017

Noises are usually generated by various external causes and low quality devices in image data acquisition and recording as well as by channel interference in image transmission. Since these noise signals result in the loss of information, subsequent image processing is subject to the corruption of the original image. In general, image processing is performed in the mixed noise environment where common types of noise, known to be Gaussian and impulse, are present. This study proposes an iterative weighted mean filter for reducing mixed type of noise. Impulse noise pixels are first turned off in the input image, then $3{\times}3$ sliding window regions are processed by replacing center pixel with the result of weighted mean mask operation. This filtering processes are iterated until all the impulse noise pixels are replaced. Applied to images corrupted by Gaussian noise with ${\sigma}=10$ and different levels of impulse noise, the proposed filtering method improved the PSNR by up to 12.98 dB, 1.97 dB, 1.97 dB respectively, compared to SAWF, AWMF, MMF when impulse noise desities are less than 60%.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10b
• /
• pp.514-519
• /
• 1992

To make an accurate retrieval of the proportion of each category among mixed pixels (Mixel's) of a remotely sensed imagery, a maximum likelihood estimation method of category proportion is proposed. In this method, the observed multispectral vector is considered as probability variables along with the approximation that the supervised data of each category can be characterized by normal distribution. The results show that this method can retrieve accurate proportion of each category among Mixel's. And a index that can estimate the degree of error in each category is proposed. AS one of the application of the proportion estimation, a method for image classification based on category proportion estimation is proposed. In this method all pixel in a remotely sensed imagery are assumed to be Mixel's, and are classified to most dominant category. Among the Mixel's, there exists unconfidential pixels which should be categorized as unclassified pixels. In order to discriminate them, two types of criteria, Chi square and AIC, are proposed for fitness test on pure pixel hypothesis. Experimental result with a simulated dataset show an usefulness of proposed classification criterion compared to the conventional maximum likelihood criterion and applicability of the fitness tests based on Chi square and AIC,

• Korean Journal of Remote Sensing
• /
• v.19 no.5
• /
• pp.403-409
• /
• 2003

Since SPOT4 satellite contained VEGETATION 1 sensor launched, the noise in VEGETATION data was occasionally arisen a difficulty for the data traitement. Blind line noise types were studied in VEGETATION-l short wave infrared channel(SWIR). In order to provide a precis product, the procedure for removing this noise is strongly recommended. In the case that the blind values are clearly distinguished from contamination-free values a simple threshold method was applied, while a changeable threshold method was used for the blind value mixed with contamination-free values. New algorithm presented in this study is consists of two method for each type of SWIR blind. After removing blind line, there were again some residual pixels of blind, because the threshold is not determinated sufficiently low. Lower threshold could remove the blind line as well as the contamination-free pixels. Nevertheless, the results showed a good qualitative improvement as compared with other algorithm.

• Korean Journal of Remote Sensing
• /
• v.30 no.4
• /
• pp.511-524
• /
• 2014

In this study, we investigated the statistical occupations and interannual variations of land cover types over Northeast Asian region using the 12 years (2001-2012) MODerate Resolution Imaging Spectroradiometer(MODIS) land cover data sets. The spatial resolution and land cover types of MODIS land cover data sets are 500 m and 17, respectively. The 12-year average shows that more than 80% of the analysis region is covered by only 3 types of land cover, cropland (36.96%), grasslands (23.14%) and mixed forests (22.97%). Whereas, only minor portion is covered by cropland/natural vegetation mosaics (6.09%), deciduous broadleaf forests (4.26%), urban and built-up (2.46%) and savannas (1.54%). Although sampling period is small, the regression analysis showed that the occupations of evergreen needleleaf forests, deciduous broadleaf forests and mixed forests are increasing but the occupations of woody savannas and savannas are decreasing. In general, the pixels where the land cover types are classified differently with year are amount to more than 10%. And the interannual variations in the occupations of land cover types are most prominent in cropland (1.41%), mixed forests (0.82%) and grasslands (0.73%). In addition, the percentage of pixels classified as 1 type for 12 years is only 57% and the other pixels are classified as more than 2 types, even 9 types. The annual changes in the classification of land cover types are mainly occurred at the almost entire region, except for the eastern and northwestern parts of China, where the single type of land cover located. When we take into consider the time scale needed for the land cover changes, the results indicate that the MODIS land cover data sets over the Northeast Asian region should be used with caution.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.6
• /
• pp.2302-2316
• /
• 2015

We propose an efficient method of extracting targets within a region of interest in non-homogeneous infrared images by using a principal component analysis (PCA) plane and adaptive Gaussian kernel. Existing approaches for extracting targets have been limited to using only the intensity values of the pixels in a target region. However, it is difficult to extract the target regions effectively because the intensity values of the target region are mixed with the background intensity values. To overcome this problem, we propose a novel PCA based approach consisting of three steps. In the first step, we apply a PCA technique minimizing the total least-square errors of an IR image. In the second step, we generate a binary image that consists of pixels with higher values than the plane, and then calculate the second derivative of the sum of the square errors (SDSSE). In the final step, an iteration is performed until the convergence criteria is met, including the SDSSE, angle and labeling value. Therefore, a Gaussian kernel is weighted in addition to the PCA plane with the non-removed data from the previous step. Experimental results show that the proposed method achieves better segmentation performance than the existing method.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.20 no.1
• /
• pp.21-31
• /
• 2002

This study applied each Neural Networks theory and Fuzzy Set theory to improve accuracy in remotely sensed images. Remotely sensed data have been used to map land cover. The accuracy is dependent on a range of factors related to the data set and methods used. Thus, the accuracy of maps derived from conventional supervised image classification techniques is a function of factors related to the training, allocation, and testing stages of the classification. Conventional image classification techniques assume that all the pixels within the image are pure. That is, that they represent an area of homogeneous cover of a single land-cover class. But, this assumption is often untenable with pixels of mixed land-cover composition abundant in an image. Mixed pixels are a major problem in land-cover mapping applications. For each pixel, the strengths of class membership derived in the classification may be related to its land-cover composition. Fuzzy classification techniques are the concept of a pixel having a degree of membership to all classes is fundamental to fuzzy-sets-based techniques. A major problem with the fuzzy-sets and probabilistic methods is that they are slow and computational demanding. For analyzing large data sets and rapid processing, alterative techniques are required. One particularly attractive approach is the use of artificial neural networks. These are non-parametric techniques which have been shown to generally be capable of classifying data as or more accurately than conventional classifiers. An artificial neural networks, once trained, may classify data extremely rapidly as the classification process may be reduced to the solution of a large number of extremely simple calculations which may be performed in parallel.