Research On Forest Type Recognition Method Of UAV Remote Sensing Image Based On Bee Colony Algorithm

Reasonable protection and utilization of forest resources is of great significance to the stability and sustainable development of the Earth's ecosystem.The correct identification of forest types is an important basis.The development of drones with low cost,high flexibility and high resolution has laid the foundation for obtaining resource information and orderly development of forests.Domestic and foreign scholars use hyperspectral remote sensing data and different classification methods to identify forest types in order to find more efficient and reasonable ways and more accurate results.Considering the practical application,there were two kinds of unlabeled sample information and labeled samples.In the case of data sets,the paper studied the use of unsupervised classification and supervised classification to identify forest types.The bee colony k-means unsupervised classification method and the support vector machine(SVM)supervised classification algorithm were used to identify the forest types in the forest farm to meet the forest type identification requirements of different data samples.The main work of the thesis includes:(1)Data preprocessingThe image of the drone image in the study area was preprocessed by digital orthophoto processing and cropping,which formed the basic data that can be used for classification.The paper used the LPS(Leica Photogrammetry Suite)digital photogrammetry system in ERDAS IMAGINE to produce digital orthophoto maps.The digital orthophoto image was then cropped.(2)Research on forest type clustering model of UAV remote sensing image and application research of UAV remote sensing image clusteringThe perfonnance of the clustering model directly affected the forest type clustering accuracy of remote sensing images.To this end,the paper proposed a clustering model based on the swarm group k-means(IKABC)and applied it to the clustering application research of forest type identification of remote sensing images.Firstly,the model firstly used the maximum and minimum distance product neighborhood mean method to initialize the bee colony,and combined the inter-class and intra-class relationship of the cluster to improve the fitness function of the bee colony algorithm.Then,the bee colony algorithm was executed once and the new one will be executed.The position was used as the initial point of k-means and a k-means clustering was performed,and the new cluster center obtained by clustering was used to update the bee colony,so that the bee colony algorithm and the k-means algorithm were alternately executed until the end of the algorithm.Then,the IAKBC model was validated on four sets of UCI public data sets.Finally,IKABC was applied to the clustering application research of forest type identification of UAV remote sensing images.(3)Research on forest type classification algorithm based on LABC-SVM for remote sensing image of UAVThis paper studied the classification application of the classification model based on swarm colony algorithm SVM(LABC-SVM)in forest type recognition of remote sensing images.Firstly,the paper extracted the SIFI features,and clustered the SIFT features to obtain the visual word table and visual dictionary,and then combineed the spatial pyramid matching model to obtain the final image representation feature data set,and used the feature data set as the data input of LABC-SVM.LABC-SVM used the chaotic operator to initialize the bee colony algorithm,and used the chaotic local search operator to improve the bee colony search strategy,and selected the ten-fold cross-validation result of SVM as the fitness value.When applying the LACC-SVM to identify the forest type,the parameters of the LABC-SVM(the kernel function and the pyramid level L)were compared and optimized.Then,five sets of UCI public data sets and three sets of common image data sets were selected to verify the validity of the LABC-SVM classification model.Finally,LABC-SVM was used for forest type identification in broadleaf,coniferous,mixed forest and non-forest(water and construction).