Information Extraction Of Qianxi Chestnut Tree Based On Multi-feature Optimization

Chestnut is native to my country and has a long history of planting.Qianxi County is the country’s largest chestnut base.Fast and effective extraction of chestnut tree information is of great significance to the development of Qianxi chestnut industry.Remote sensing technology is an important method for large-area crop extraction.The traditional pixel crop extraction leads to the "salt and pepper phenomenon".Multi-temporal remote sensing images and multiple features provide rich information for crop extraction.Therefore,in order to grasp the chestnut tree planting information in Qianxi County in time,the research is based on multi-temporal images,combined with multiple algorithms to optimize multiple features,and the method of object-oriented combined with SVM is used to extract chestnut tree information.Sort out Landsat8-OLI images of 7 time phases in different months,determine the best time phase through the spectral difference between main features,and determine the best segmentation scale through the area ratio average method;construct a multi-feature space(NDVI time series features,Spectral features,texture features);the optimal feature set is calculated by PCA transformation,ABS method,OIF method,and JM distance method;SVM classification method is used to extract chestnut tree information based on the best time phase,primary feature set,and preferred feature set.,After the accuracy evaluation,the extraction effect of the preferred features is tested and the results are analyzed.The study came to the following conclusions:1)Through calculation and analysis of the spectral values of chestnut trees,pear trees,and evergreen trees on the 6-phase Landsat8 remote sensing image,the results show that the month with the largest spectral difference is June.The best time for chestnut tree identification in the study area is determined to be the chestnut tree Flowering.2)The area ratio average method is used to determine the best segmentation scale of the features.Different features have different changing laws.The smaller the area of the features,the more sensitive the impact of the segmentation scale.3)Through the comparison of the chestnut tree extraction results of the optimal feature set with the spectral features of the optimal time phase,the overall accuracy of the extraction of the optimal feature set was increased by 8.19%,and the production accuracy of chestnut trees was increased by 16.99%,indicating that the research constructs a multifeature space set to extract chestnut trees The effect is good.4)Through PCA,ABS,OIF,J-M algorithm combination of optimal features and PCA selected features of chestnut tree extraction results,the overall extraction accuracy of multiple algorithms is increased by 5.07%,and the production accuracy of chestnut trees is increased by 9.88%.It shows that the research using multiple algorithms to combine the optimal features has a good effect on chestnut tree extraction.Figure 23;Table 14;Reference 79...