珊瑚岛礁CBERS-02B与资源二号融合影像的质量评价

The quality evaluation of CBERS-02B and ZY-2 fused image of the coral reefs

南海海域分布着众多珊瑚岛礁,鉴于其蕴含着巨大的经济、军事和生态等价值,我国越来越重视对珊瑚岛礁的调查、监测和研究。在调查任务的推动下,选取国产中巴资源卫星-02B(CBERS-02B)多光谱影像与资源二号(ZY-2)全色影像数据,对其影像的质量进行评价研究,可验证其能否应用于珊瑚岛礁的监测。以西沙群岛中的珊瑚岛为例,应用主成分分析(PCA)、Brovey变换、乘积变换3种常用融合方法,对国产自主卫星CBERS-02B与资源二号影像进行了融合试验,并采用主观与客观相结合的评价方式,与国际主流的卫星SPOT-5自身融合影像进行对比分析。CBERS-02B、资源二号融合影像与SPOT-5融合影像质量对比分析结果表明:目视效果来看,SPOT-5卫星数据较高的分辨率使其融合影像的分辨能力明显高于CBERS-02B与资源二号融合影像。但在光谱信息方面,CBERS-02B与资源二号融合影像对绿色植被的分辨具有优势,有利于植被目视识别。客观统计指标显示,与SPOT-5融合影像相比,CBERS-02B与资源二号融合影像绿光波段灰度平均值较高,红光和近红外波段灰度平均值较低;两组融合影像的标准差相差不大;从影像的熵来看,CBERS-02B与资源二号融合影像的熵值较高;而平均梯度和相关系数低于SPOT-5融合影像。总体而言,由于受到空间分辨率的限制,CBERS-02B与资源二号融合影像目视效果逊于SPOT-5自身融合影像,但在空间细节分辨能力方面具有优势,能够反映出更多的光谱特征,尤其是对植被的光谱差异反映细腻。

There is a large amount of coral islands in the South China Sea （SCS）. Owing to their great economic, military, scientific and ecological value, China attaches more and more importance to the investigation, monitoring and study of them. Recently, remote sensing investigation and research of the coral islands in SCS are being carried out under the support of State Oceanic Administration （SOA） of China. With hyperspectral resolution, high spatial resolution and high temporal resolution, remote sensing has become an important monitoring technique of coral islands. Satellite data used for remote sensing investigation are mostly bought from abroad. But during remote sensing investigation of the coral islands, we find that the coverage of the high resolution data is very much limited, and the optical image is seriously influenced by cloud and mist. As a result, it is very difficult to get high quality high-resolution remote sensing images. Therefore, extension of choice scope of data should be considered to meet the investigation requirements. Homemade satellites have the following advantages： easy acquisition, low price, widespread coverage in Chinese territory and so on. Thus multi-spectral image of CBERS-02B and panchromatic image of ZY-2 are chosen to exploit the applicableness of these homemade satellites for the monitoring of the coral islands. This paper takes the Shanhu islands in Xisha archipelago as an example. Multi-spectral image of CBERS-02 and panchromatic image of ZY-2 are fused with 3 common fusion methods including PCA （Principle Component Analysis） transform, Brovey transform and multiplication transform, and using combined subjective and objective evaluation methods the results are compararatively analyzed with the self-fusion image of international popular SPOT-5. Subjective evaluation is mainly based on visual judgment of the fusion results, while objective evaluation mainly on statistics such as the average, standard deviation, entropy, average gradient and correlation coefficient. The comparison between different fusion methods applied to the image of CBERS-02B and ZY-2 manifests that PCA is the best in terms of its higher image definition, maximal entropy, and balanced ability of fusing spectral and spatial information. As for the visual effect, the quality analysis of the fused image of CBERS-02B and ZY-2 compared with SPOT-5 data shows that the fusion image resolution capacity of the higher-definition SPOT-5 satellite data is obviously higher than that of CBERS-02B and ZY-2. But the fusion image of CBERS-02B and ZY-2 has the advantage in resolving green vegetation and is favorable for visual recognition of vegetation over SPOT-5 image. The objective statistical indexes show that fusion image of CBERS-02B and ZY-2 has higher average in green band and lower average in red and near infrared bands than the fusion image of SPOT-5; the standard deviation of the two groups of fused images has few discrepancies; the entropy of fusion image of CBERS-02B and ZY-2 is higher than that of SPOT-5; and the average gradient and correlation coefficient of the combination of CBERS-02B and ZY-2 are lower than that of SPOT-5. Overall, limited by spatial resolution, the total visual effect of the fused image of CBERS-02B and ZY-2 is inferior to that of SPOT-5, but the former has the advantage in resolving spatial details, they can display more spectral characteristics, especially the spectral difference of vegetation in great details. The findings of this paper provide technical support and theoretical foundation for the implementation of remote sensing investigation of the coral islands, and also provide the reference for optimizing the remote sensors and the application and popularization of remote sensing data of the homemade satellite.