Inversion Of Apple Tree Canopies Reflectance In Hilly Region Based On Multispectral Remote Sensing Image

Apple tree canopies reflectance is one of the essential attribute of the apple tree. It plays an important role in its growth monitoring, nutrition inversion and yield estimation. Due to the effects of atmospheric, topography, soil, and the apple tree canopies density factors, result that apple tree canopies actual spectral reflectance and the sensor measurements are not consistent and remote sensing image does not reflect the true information of apple tree canopies in the use of satellite multispectral remote sensing inversion of apple tree canopies reflectance. Therefore, removing or weakening the atmosphere, terrain and the non-target objects in the mixed pixel is important for the accurate and efficient retrieval of the true reflectance of the apple tree canopies.The study area is Qixia City, Shandong Province, which is called "the capital of Chinese apple". Firstly, the Landsat8 and SPOT-5 multi spectral remote sensing images were processed by normal radiation correction, geometric correction, radiometric calibration and atmospheric correction, and combined with the ground measured data and DEM data. Secondly, the Topographic Radiation Correction of Landsat8 and SPOT-5 remote sensing images were carried out, which were based on cosine correction model, C correction model, SCS correction model, SCS+C correction model and C+ slope correction model and SCS+C+ slope correction model. Thirdly, the inversion reflectance of apple tree canopies was extracted after the mixed pixel decomposition. The main results are as follows:(1) Picked out the optimal model of Topographic Radiation CorrectionBecause the study area is hilly region, the terrain has a significant impact on Landsat8 and SPOT-5 image. Sunny side receives more sunlight illumination, reflectance of the objects is higher; shaded aspect receiving the less light, low reflectance of ground objects, so the reflectivity of the slope of sun and shade have a large difference. The spatial texture information of Landsat8 and SPOT-5 was improved, especially the surface information of the shadow region is greatly restored by cosine correction model, C correction model, SCS correction model, SCS+C correction model, C+ slope correction model and SCS+C+ slope correction model. Surface retrieved reflectance obtained after the correction of topographic radiation is generally higher than that obtained by atmospheric correction.SCS+C+ slope correction effect was the best, and the cosine correction and SCS correction were the phenomenon of excessive correction, from the visual effect, histogram, 6 kinds of Topographic Radiation Correction of the terrain after the correction of the image, the surface reflectance and the measured reflectance of the canopies were compared.(2)The characteristics and accuracy of Landsat8 and SPOT-5 mixed pixel decomposition in different bands were clarified.Transformed soil adjusted vegetation index(TSAVI) can reduce the effect of the soil background on the reflectance of vegetation. Combined with the soil hyperspectral data, the TSAVI were used to extract vegetation fraction. For landsat8, canopies reflectance inversion of band1, band2, band3, band4, band6, band7 were reduced 5.405%, 7.692%, 12.284%, 9.324%, 20.401% and 8.923% than surface inversion reflectance, while canopies reflectance inversion of band 5 was improved 26.543% than surface inversion reflectance by the process of mixed pixel decomposition. For SPOT-5, canopies reflectance inversion of band1, band2 and band4 were reduced 2.357%、5.287% and 5.753% than surface inversion reflectance, while canopies reflectance inversion of band 3 was improved 36.208% than surface inversion reflectance by the process of mixed pixel decomposition.(3)The inversion reflectance of SPOT-5 was better than that of Landsat8Because of the band 3, band 4, band 5, band 6 of Landsat8 were corresponding to the band 1, band 2, band 3 and band 4 of SPOT-5, their reflectance inversion accuracy were compared and analyzed. Comparative analysis showed that the landsat8 and SPOT-5 canopies reflectance inversion in the samples showed a similar trend. SPOT-5 the 1, 2, 3, 4 band correlation coefficient were respectively higher 0.218, 0.121, 0.172, 0.158, than that of the band 3, band 4, band 5 and band 6 of Landsat8. It indicated that the correlation between the inversion reflectivity and the measured value based on the SPOT-5 was higher than that of the Landsat8. The root mean square error of SPOT-5 was 0.003, 0.001, 0.014, 0.042 less than that of Landsat8. Therefore, the high spatial resolution of the SPOT-5 image reflectance retrieval effect was better than Landsat8. For the Landsat8, the relative error of the fifth band reflectance was only 5.010%, and the fifth band reflectance inversion effect was best. For the SPOT-5, the relative error of the forth band reflectance was only 4.778%, and the forth band reflectance inversion effect was best.