Research On Classification Of Forest Vegetation Types Based On UAV Remote Sensing

Forest resource is one of the important subjects of terrestrial ecosystem,it’s also a multifunctional and multi-resource complex.It is the most abundant ecosystem on earth.How to quickly and accurately grasp the state and change of forest resources plays a vital role in the protection of forest resources and species diversity.The correct classification of forest types provides important data support.Selecting effective features and training appropriate classifiers are the keys to improve the classification accuracy.With the rapid development of unmanned aerial vehicle(UAV)platform in recent years,UAV remote sensing provides a new method for remote sensing data acquisition by virtue of its flexible mode and low cost.Therefore,in order to obtain a high-precision and efficient classification method,this paper mainly studies the methods and technologies of forest type classification based on UAV remote sensing data,and carries out the following work.(1)Acquisition and processing of UAV remote sensing images.Using unmanned aerial vehicle(uav)in the greater hinggan mountains to get root river forest forest aerial photo,first using the 3d reconstruction system Agisoft Photoscan to get photos of distortion correction and Mosaic and so on processing,get the Numbers are projective images of study area,then the orthogonal projection of the acquired by cutting like figure according to certain size,used for classification of data sets are obtained.(2)An improved ant colony algorithm combined with support vector machine classification model(IACO-SVM)was proposed.First model of ant colony algorithm in ant colony search to add some limited search,avoid the local extremum,in the process of ant colony to update pheromone marching into time-varying function,the number of iterations,the objective function of the related dynamic update strategy and the combination of support vector machine(SVM),optimize the parameters of RBF kernel function of support vector machine(SVM).Then the validity of the algorithm model was evaluated on the UCI standard dataset.The results show that the IACO-SVM model has higher classification accuracy than the genetic algorithm optimized support vector machine(GA-SVM),bee colony algorithm optimized support vector machine(ABC-SVM)and the ant colony algorithm optimized support vector machine(ACO-SVM)on five different data sets,which proves the effectiveness of the model.(3)Extract different texture features from UAV remote sensing images and use the above model to identify forest types.Firstly,the traditional gray level co-occurrence matrix was used to extract the features from the data set,and then the feature data set was used to identify the forest types.Then Gabor texture features,which are commonly used in the field of machine vision,are introduced to extract features from data sets.Due to the high Gabor texture feature dimension,principal component analysis is used in this paper to reduce dimension and forest type recognition is carried out on the data set after dimension-reduction.The results show that adding texture features to IACO-SVM model can further improve the classification accuracy of forest types in remote sensing images.In addition,the Gabor texture features introduced in this paper have higher classification accuracy in the IACO-SVM model compared with the traditional texture features extracted by gray level co-occurrence matrix,which proves the effectiveness of the application of Gabor texture features in the classification of forest types.There are 20 figures,7 tables and 98 references in this paper.There are 25 figures,4tables and 87 references in this paper.