Urban Individual Tree Crown Detection And Segmentation Research Using Multispectral Images Dimensionality Reduction With Deep Learning

The dimensionality reduction of multispectral data can make it better applied,and it is also helpful for single tree crown detection and segmentation based on deep learning.However,how to use the appropriate dimensionality reduction method to improve the accuracy of single-tree detection is rarely discussed.In this work,a multispectral camera equipped with an UAV was used for aerial photography to collect multispectral images of ginkgo tree species in the research area.In this study,four dimensionality reduction methods(principal component analysis,independent component analysis,optimum index factor,standard false color composite)were used to reduce the dimensionality of multispectral images.The images after dimensionality reduction were made as dataset for network training.Three object detection networks(FPN-Faster-R-CNN,OLOv3,Faster R-CNN)were used to detect the ginkgo tree crowns of UAV multispectral images after dimensionality reduction in urban.Two instance segmentation networks(Blend Mask,Mask R-CNN)were used to delineate the ginkgo tree crowns of UAV multispectral images after dimensionality reduction in urban.The effect of dimensionality reduction methods was evaluated in detail.The result of experiments presented that the standard false color composite method obtained the best value in tree crown detection and segmentation tasks.The results of optimal exponential factor and independent component analysis are not good.The effect of principal component analysis method is in the middle.FPN-Faster-R-CNN can achieve the highest detection accuracy of ginkgo biloba crown,and Blend Mask can achieve the optimal segmentation result of the crown.The experimental results show that: the methods based on band selection are superior to the methods based on feature extraction when the appropriate bands are selected,and the band selection methods are also more stable.The effect of dimension reduction method of feature extraction is generally lower than that of band selection method.In the different dimensionality reduction methods,if the color and background of the target object in the image after dimensionality reduction are obviously different,and the contour is clear,the deep learning network can obtain better results in the detection of tree crown.However,the information content of the image itself has limited effect on the ability of deep learning network to detect tree crown.This article provides an useful reference for related researchers on the choice of dimensionality reduction methods.