Research on optimal fusion method of EOS/MOD l S

Zhuo Dengfeng Dong (Shaanxi Agricultural Remote Sensing Information Center An 7 100 15)

In order to improve the spatial resolution of MODIS images, this paper analyzes the commonly used his transform method, principal component analysis method, simple product method and Brovey method.

The advantages and disadvantages of image fusion algorithm and false color synthesis method are analyzed, and the fused image is analyzed according to the characteristics of EOS/MODIS remote sensing image.

After a simple evaluation, it is finally considered that the fusion effect using principal component analysis is the best.

Keywords: MODIS image fusion

1 Introduction

With the development of remote sensing technology, different sensors get the same data.

Multi-spectral, multi-resolution and multi-temporal images in a region are increasing.

It provides rich and valuable information for natural investigation and environmental monitoring.

Expensive information. However, few data are obtained by these single remote sensing methods.

There are some limitations and differences in spectrum and so on. how

It is very important that their respective advantages complement each other. Remote sensing signal

According to the data, integration is a good way to combine these advantages. Image fusion may

It is divided into the fusion of different sensors, such as the fusion of SPOT and TM.

Fusion between different resolutions of the same sensor, such as MODIS.

Data fusion. But no matter what kind of fusion, the spatial resolution will be high.

Multi-spectral images will be more abundant after fusion.

The spectral, texture and geometric information of the image are improved.

Explain the effect and reliability. There are many methods of image fusion,

Such as vegetation index algorithm, tassel change, principal component analysis, module-based

Types, methods and so on. From a technical point of view, data fusion mainly

Includes three layers: pixel-level fusion, feature-level fusion and decision-level fusion.

Second, different levels adopt different fusion methods. This paper mainly studies

HIS transform method, simple product method and principal component analysis method of feature level fusion

Analysis methods, Brovey image fusion algorithm and false color synthesis.

Advantages and disadvantages of this method in MODIS data fusion.

Data introduction of 2 MODI

MODIS (its full name is: medium resolution

Imaging spectro radioⅲeter was launched in the United States at the end of 1999.

Terra (EOS-AMI) is the only live broadcast.

A new generation of integrated spectral optical remote sensing instruments in the world.

***36 covering visible light, near infrared and far infrared can be obtained at the same time.

100 channel 100 meter remote sensing data. MODIS data has multi-spectral characteristics.

Resolution (* * * has 36 channels), high time resolution (every day

Top twice), multi-spatial resolution (250m, 500m and1000m)

Are there other features? . Its spatial resolution is higher than NOAA/AVHRR and spectrum.

The resolution is higher than TM, which just fills the gap in the middle. MODIS diagram

The spectral range of the image is from visible light to far infrared, and its data can be used.

In ecological environment monitoring, fire monitoring, drought monitoring and other aspects, such as

Can improve the spatial resolution of the low-resolution band, and can

Reflects a lot of details that NOAA can't do. For example, in a forest fire

In terms of monitoring, MODIS data has higher accuracy and higher saturation temperature.

It has high value for the identification and classification of forest fires. on a large scale

NDVI and classification of 250m resolution ecological environment monitoring.

The accuracy of the results is higher than that of NOAA, which is more valuable for reference. therefore

If we can fuse MODIS data with different resolutions well.

Combined, it can give full play to the advantages of multi-spectrum and medium resolution.

Wave.

3 image preprocessing

The data used in the study are high altitude and partly cloudy.

MODIS image of l 65438+20021October 26th. Will be 250m resolution.

Precise correction of image geometry The image is corrected in ENVI software.

Manually select the GCP point of the reference image to correct other resolutions respectively.

Images so that the correction error of each image is within one pixel, which may

So as to well ensure the coincidence degree between images with different resolutions, and can

Thereby improving the quality of the fused image.

4 image fusion method

4. 1 HIS transformation method

His transformation method is one of the most commonly used fusion methods, and

RGB color space, which describes the color attributes of objects.

System, where I stands for brightness, H stands for chroma, and S stands for.

The saturation of color represents the average radiation of three bands respectively.

Intensity, data vector and equivalent size. How to do this method.

The method is: first, I, H,

S component, and then use 250m 2 band data (choose 2 band original

Because it is introduced below) instead of 1 component, and then convert it into

RGB color space, forming a new image. This new image has not only maintained

The brightness index with high resolution is obtained, and the chromaticity sum of multi-spectrum is retained.

Saturation index. However, the gray values of the fused multispectral images are the same.

The original multispectral images are very different, that is, the spectral characteristics are distorted, basically

The hue of the multispectral image is maintained.

4.2 Simple product method

Crippen used four methods of addition, subtraction, multiplication and division in 1989.

Methods A brightness image was converted into a color image, and the results showed that only one color image was available.

Multiplication has a minimum color distortion, and the calculation formula is as follows:

DNf multispectral image): l: dn (high resolution image) = Dⅳ Ⅳ (new image)

Weipu information

1O quarterly journal of surveying and mapping technology and equipment for academic research, volume 0, issue 9 1 No.2007

Among the three commonly used fusion methods, this method is the simplest and time-consuming.

The least method, but this method changes the radiance of multi-spectral data.

The brightness component of information and images increases, so in cities and suburbs.

In the study of urban areas, this method is often used to place urban roads and artificial features.

The contour reflection feature is highlighted.

4.3 Brovey image fusion algorithm

Brovey image fusion is also called color standardization.

(Colornormalized) transformation fusion, by American scientists.

R.L.Brovey established the model and extended it, hence the name. Its algorithm is to combine many

The image space of spectral image is decomposed into color and brightness components and processed.

Do the math. Its characteristics are: simplifying the image conversion process and maximizing the coefficient.

The information of multi-spectral data is preserved to the maximum extent. This method mainly uses

Calculate the three bands of the image, and the calculation formula is as follows:

(DN port 1 I DN port 1+DN port 2+DN port 3) Wood W high-resolution image =DN port 1 a new image.

(DN port 2 IDN port1+dⅣ B2+DN port 3) Wood dⅣ high-resolution image =DN port 2 A new image.

(3IDNB 1+ 2+ 3)*DN~ m，mm= image。

B= band

The result of this method is that the color tone is very good and almost complete.

Tone information of the original image is preserved. And it will increase.

The ratio of the high and low parts of the image histogram provides the shadows and water in the city.

Compared with highly reflective ground objects, this can produce a higher picture.

RGB image, the degree of which reflects the contrast between the high part and the low part of the image histogram. but

However, this method has some limitations on the radiation information of multi-spectral images.

Change, if the radiation information of multi-spectral data is very in the future research.

Important, then you can't use this method.

4.4 Principal Component Analysis Fusion Method

Principal component analysis (PCA) is a feature selection method.

Each new feature is a linear function of the original feature. It mainly

It aims to fuse images with more than three bands. His transformation method and

Brovey method is limited when it exceeds three bands and can only be extracted.

Select three bands in the multi-spectral image to participate in the change, so that it will not appear.

Doubt will make the information in other bands lose, which is not conducive to the comprehensive benefits of the image.

Use. The image synthesized by this method has prominent hue and clear vegetation coverage.

Different vegetation types and growth conditions can be distinguished by colors and water bodies.

The performance is good, but the depth reflection is poor. It is based on the following assumptions:

The first principal component of multispectral images contains almost all optical bands.

Degree information, and other components include spectral information; Multispectral data

The brightness information of is similar to that of high-resolution data. In the fusion

In this process, the histogram shape of high-resolution image remains unchanged.

The first principal component of the spectral image is replaced and inversely transformed.

The method of fusion. Therefore, the fused image not only maintains multi-spectrum.

It also has the characteristics of high resolution. This method is the most

This is a common image fusion method, but its operation time is long.

4.5 False color synthesis

Pseudo-color synthesis is based on the principle of tricolor imaging and will match accurately.

Quasi-three-band gray-scale images are given R, G and B colors respectively,

Then combine them to form a new color image. Due to people

The ability of eyes to distinguish color images is much higher than that of gray images, so

This method is widely used in visual interpretation of images and thematic mapping.

Medium. In thematic mapping, if there is one more band, there will be more detailed information.

For MODIS image, in order to improve its resolution, this method

250m images are mainly used, but only 250m images are used.

There are two bands, which requires adding a third band to the image to fill it.

This method is enough. There are two ways to choose the third channel.

Method: One is to repeat one of the 250m bands and give it to the third one.

Band, such as RGB channel, is given to 1 channel, 2 channels, 1 channel respectively.

Tao, this combination is the simplest and does not need to be integrated.

But also beautiful and convenient when making thematic maps; Another method

Is to resample the 500m channel and add it to the.

250m image. This can increase the amount of information, such as adding waves.

The near infrared channel of segment 7, so the 3-band data formed is false color.

Images, such images can show the information of forest land well. And 7

This channel is sensitive to open fire and its energy, and can be used to monitor fire. this

Although this method is simple and convenient, it will lose 500m other bands.

Spectral information, so it is generally only used when making thematic maps.

5 Analysis and comparison of results

5. Band selection of1MODIS high resolution image

As we all know, MODIS250m resolution image has two waves.

Segment: one is a red band; The other is near infrared band. And in

Among the above three methods, only high-resolution images are generally used.

A band, such as the panchromatic band of SPOT. Therefore, it is necessary to pay attention to these two.

Comparing the amount of information in two bands, which band has more information,

Which band is used to fuse the letters of the fused image?

The interest content is also relatively large. When judging whether the amount of information is large or small, make

By looking at the histogram distribution and comparing the mean square deviation.

The histogram has a wide distribution range, which shows that it contains a lot of information. mean square

The difference reflects the dispersion of gray level relative to the average value of gray level, and the greater the variance.

The larger, the more dispersed the gray distribution. At this time, all gray levels in the image

The more equal the probability of occurrence, the more information it contains. mean square

The difference is calculated as follows:

Pressure one

O= 1f (Qj-Q) Ik- 1

i= 1

Is the first sample value; Q is the mean; K is the number of samples.

Through this formula, the mean square error of 1 band is 1696.794.

The mean square error of band 2 is 2708.608, indicating the signal of band 2.

Information is greater than 1 band.

The wave is also explained from the histogram shown in Figure 1 and Figure 2.

Weipu information

Surveying and Mapping Technology and Equipment Quarterly, Volume 9, 2007 1 Academic Research 1 1

The amount of information in segment 2 is greater than that with 1

Figure 2: 250 meters wide and 2 straight lines

5.2 Analysis of results

In the study, the above five methods are used for fusion respectively.

Fig. 3 is the MODIS image of 250m band 2, and fig. 4 is the image of 500m.

For example, fig. 5 is a false color combination repeatedly allocated in 250m band 1.

Imaging (band combination is band 1, band 2 and band 1), as shown in Figure 6.

It is the result image of his transformation method, and Figure 7 is the fusion of simple product method.

After image, fig. 8 is the image fused by Brovey method, and fig. 9 is the principal component.

Fractional analysis of fused images.

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Compare that fused image, the high-resolution image and the multispectral image,

False color synthesis method does not divide the space of 500m resolution image.

The resolution is improved without using the hyperspectral resolution of MODIS images.

Advantages; The fused image of HIS is very different from the original image in spectrum.

Not the same; The simple product method has the worst fusion effect and has spectral characteristics.

Weipu information

12 quarterly of surveying and mapping technology and equipment for academic research, Volume 9,No. 1, 2007.

Change; After Brovey fusion, the contrast between light and dark of the image increases; Major achievements

The fusion effect of fractional analysis is the best, and the spectral characteristics and the original image are the most

For similarity.

6 conclusion

For MODIS images, its resolution and multispectral characteristics are decisive.

It is mainly used for large-scale remote sensing monitoring, such as drought monitoring,

Fog monitoring, sandstorm monitoring, etc. And the application of spectrum in these applications.

Features are more important. Therefore, spectral characteristics and fusion are considered comprehensively.

Effect and other factors, the fusion result in Figure 8 is the most ideal, that is

Under the principle of maintaining multi-spectral characteristics, principal component analysis is used.

Line fusion works best.

refer to

1 Qin Xianlin, Yi Haoruo. Study on combustion monitoring method based on MODIS data J. Remote sensing technology and application, 2002,4 (2): 66 ~ 69

2 Niu Zhichun and Ni. Research progress of data fusion method in land use dynamic remote sensing monitoring J. Journal of Nanjing Normal University (Natural Science Edition), 2002,25 (3);

12~ 17

3 Wang Jian, Lu Anxin, Guo Tingtian, Miao Chun, et al. Application. Application of Brovey image fusion in land cover investigation of Gongyi irrigation area J. Application of remote sensing technology, 200 1, 9 (3);

173~ 177

4ERDAS company ERDAS field guide. June 1998

5 sheets, Zhang Jixian,, et al. Comparative study on multi-source remote sensing image fusion methods in dynamic remote sensing monitoring of land use. Surveying and Mapping Science, 2000,9 (3): 46-51

(Continued from page 22)

refer to

1 National Natural Science Foundation. Geospatial information science. Beijing: Higher Education Press, springer Press, 200 1

Liu Guangyun, Han Libin. Electronic map technology and its application. Beijing: Surveying and Mapping Press, 1996.

3 Sarah. Research on theory and technology of seamless geographic information system. Doctoral thesis of PLA Information Engineering University, 2004.

4 Tian Desen. Modern cartographic theory. Beijing: Surveying and Mapping Press, 199 1.

People's Republic of China (PRC) national standard. 1: 5000, 1: 10000 topographic map schema. Beijing: China Standards Publishing House, 1995.

6 Li Deren. On the development of remote sensing and geographic information system in 2 1 century J. Journal of Wuhan University (Information Science Edition), 2003,28 (2):127-131.

7 Wu Fan. Research on multi-scale processing and expression of geospatial data. Doctoral thesis of Wuhan University, 2002.

Qi Qingwen, Liang Yajuan, Jing He, et al. Discussion on the theory, method and technical system of digital map. Surveying and Mapping Science, 2005,30 (6):15-18.

(Continued from page 45)

Sequence setting, operation rules, appraisal scheme and appraisal environment all need to be entered.

Line establishment and compilation.

In a word, the introduction and application of digital camera ADS40 in China,

The realization of all-digital aerial photography is a milestone in order to speed up the aerial survey figures.

Another powerful production and construction tool, but ADS40 camera, after all,

It's new things, new equipment, high technology and supporting software, right.

Its understanding, understanding, learning, digestion, absorption and application need a

Process, only by continuous testing, exploration and research, can we speed up production and application.

The development of system and interface can make ADS40 camera play the most important role.

Put into production and application as soon as possible.

refer to

1Ra iner team 3. Design principle of ADS40 airborne digital sensor for LH system. International Society for Photogrammetry and Remote Sensing, 2000

Liu Jun, Zhang Yongsheng and Fan Yonghong. Photogrammetry processing and application of ADS40 airborne digital sensor. Zhengzhou: Journal of Surveying and Mapping Technology, 2002.

3Udo Tempelmann。 Photogrammetry software of ADS40 airborne Digita 1 sensor in LH system. International Society for Photogrammetry and Remote Sensing, 2000

Weipu information