Considering there is great similarity between the wave–band design and spatial resolution of the China HJ1A/1B satellite CCD Camera and Landsat TM data, this paper plans to conduct comparison and analysis to the data mentioned above from the aspects of track parameter, spectral response characteristics and imaging quality.
可以较为详细的介绍 tab2 The sensor of HJ1A and 1B satellite CCD cameras does not have shortwave infrared (SWIR) waveband, and its infrared imaging is completed by the infrared camera of HJ1B satellite. Of the 1~4 wavebands of the CCD camera, except there is a difference of 0.02μm between the first waveband and the TM data, the range of the other three wavebands is consistent (Table2). This has provided a consistent data foundation to conduct comparative research of image qualities of the two sensors.
From Figure 1 we can see that in the range of the 1~4 (blue, green, red and near-infrared) wavebands, these two sensors have similar spectral response characteristics, thus having obvious consistent comparability from the spectral perspective.
In order to comparatively analyze images in the research zone, selection of image pair should consider similar viewing angles of sensors and cloudlessness, and the imaging time should also be close to obtain images with similar solar altitudes and solar azimuths to compare two different sensor images. In addition, in order to objectively analyze the relation of two different remote sensing images in the spectral range of the whole design, the research zone should have rich surface features; there should be highly reflective areas and there should also be lowly reflective areas, which should to a great extent be averagely distributed in the whole grayscale range of the remote sensing images (Figure2). Figure2. Overlaid area of TM and HJ1A-CCD1 synchronization scenes within China
But the physical significance of the image’s grayscale is not determined, which is closely related to the calibration parameter. This paper has adopted maximum, minimum, mean value (Formula1) and variance of radiances of different wavebands to measure the radiation precision (Formula2).
This paper adopts the normalized Gray-level Co-occurrence Matrix to calculate the entropy (Formula3); the improved Point Sharpness Method is used to evaluate definition of the image (Formula4).
Radiation precision evaluation results show that in the 4 wavebands of blue, green, red and near-infrared, HJ1A-CCD1 is more sensitive to low radiance, and has a stronger capability to receive low radiance value than TM; in the blue and red wavebands, it also has a stronger capability to receive high radiance value than TM, from which we can conclude that HJ1A-CCD1 has a wider threshold value range than TM in the blue and red wavebands; it has a weaker capability to receive high radiance than TM in the green and near-infrared wavebands; from the aspects of radiance mean value and variance we can see that the radiance level of HJ1A-CCD1 is not as high as that of TM(Table3.). In particular, the fourth waveband of HJ has a much lower radiance mean value and variance than TM, which is closely related to the spectral response characteristics of the fourth waveband of HJ1A-CCD1; HJ1A-CCD1 presents an obvious decrease after the wave length of 0.83μm, which indicates that under the same solar radiation, after the wave length exceeds 0.83μm, HJ1A-CCD1’s capability to receive radiation energy will rapidly decrease (Figure2)
Coefficient of correlation between wavebands also indicates that image data of various wavebands of HJ1A-CCD1 have a higher independence than Landsat TM, and also a smaller information redundancy. (Table4.)
Texture and definition evaluation results show that both HJ1A-CCD1 and Landsat TM data are of the spatial resolution of 30m (actual sub-stellar point resolution of TM is 28.5m); however, from the perspective of imaging effect, the information amount, texture feature and definition of various wavebands of HJ1A-CCD1 data are not as good as Landsat TM data(Table5.),so when conducting discrimination and classification based on image texture and definition, the application effect of HJ1A-CCD1 data needs further analysis.
CROSS-COMPARISON BETWEEN CHINA HJ1A-CCD AND LANDSAT TM DATA(Guoqing Li).ppt
Cross-Comparison Between China HJ1A-CCD and Landsat TM Data Guoqing Li, Xiaobing Li*, Hong Wang, Lihong Chen, Wanyu Wen State Key Laboratory of Earth Surface Processes and Resource Ecology, College of Resources Science and Technology, Beijing Normal University, Beijing, China, 100875
<ul><li>1. Introduction </li></ul><ul><li>2. Selection of Image Test Zone </li></ul><ul><li>3. Methodology </li></ul><ul><li>4. Results and Discussion </li></ul>
1. Introduction <ul><li>With the increasing service life of Landsat 5, the stability of over-life operation of sensor is facing important challenges. In addition, many currently used data outcomes directly or indirectly come from TM or ETM+ data. It is especially necessary to choose a remote sensing data source similar to Landsat TM data which can replace TM in a certain degree. </li></ul>
<ul><li>The HJ-1 satellite was launched successfully in 2008, and carried two satellites: HJ-1A and HJ-1B. HJ-1A carries two CCD sensors with 30 m spatial resolution and a hyper-spectral sensor with 100 m spatial resolution. HJ-1B carries two identical CCD sensors and an infrared sensor with two kinds of spatial resolution (150 m at near, short-wave and middle-infrared band scope and 300 m at far-infrared band). The return period of the HJ-1 satellite is two days, with synergistic operation of HJ-1A and HJ-1B. The scan width exceeds 700 km with the two satellite CCD sensors working together. This enables HJ-1 CCD remote sensing images covering all lands of China to be captured every two to three days. </li></ul>
<ul><li>3.1 Radiation Precision Analysis </li></ul><ul><li>Radiation precision is the index which reflects the information richness of the image. Many scholars have adopted grayscale mean-value and grayscale variance to evaluate radiation precision, and they believe that for different images of the area, the bigger is the distribution range of grayscale, the bigger the variance, and the richer the image information ( Franke2006 ， LI Shi-hua 2009 ). </li></ul>3. Methodology
3.1 Radiation Precision Analysis ( m, n are the height and width of the image; f(i, j) is the radiances of the image; v is the mean value of radiance.)
3.2 Calculation of Texture and Definition <ul><li>Information entropy is a measurement of the information amount that the image possesses. The higher complexity the texture has, the bigger the image information amount is, and the bigger its information entropy is . </li></ul>
is the probability of pixel gray value of “ i ”, MAX is the max valve of grayscale. ( m, n are the height and width of the image; df is the amplitude of the image grayscale; dx is the distance increments between pixels; “ a ” is the number of pixels around pixel “ i ”.)
4. Results and Discussion Table3.Landsat TM and HJ1A-CCD1 mean value and variance of radiance TM/HJ1A-CCD1 Min Max Mean Variance Band1(B) 40.99/ 9.32 254.97/ 373.71 59.97/ 6.11 107.16/ 83.11 Band2(G) 33.92/ 9.18 618.17/ 454.65 76.56/ 1.50 401.48/ 138.32 Band3(R) 14.99/ 7.51 347.67/ 369.46 49.31/ 48.53 333.21/ 177.11 Band4(N-I) 21.01/ 4.15 670.38/ 314.87 192.52/ 85.91 1579.84/ 189.06