Synergistic use of Landsat TM and SPOT5 imagery for object-based forest classification

Xiaoyan Sun,Ning Han,Guomo Zhou,Dengsheng Lu,Huaqiang Du,Lijuan Liu,Xiaojun Xu,Hongli Ge

doi:10.1117/1.jrs.8.083550

Xiaoyan Sun, Ning Han + Show 6 more

Open Access

https://doi.org/10.1117/1.jrs.8.083550

Copy DOI

Abstract

This study evaluated the synergistic use of Landsat5 TM and SPOT5 images for improving forest classification using an object-based image analysis approach. Three image segmentation schemes were examined: (1) segmentation based on both SPOT5 and Landsat5 TM; (2) segmentation based solely on SPOT5; and (3) segmentation based solely on Landsat5 TM. The optimal scale parameters based on TM/SPOT5 and SPOT5 were determined by measuring the topological similarity between segmented objects and reference objects at 10 different scales. Mean and standard deviation of the pixels within each object in each input layer were the classification metrics. Nearest neighbor classifier was performed for the three segmentation schemes. The results showed that (1) the optimal scales of TM/SPOT5, SPOT5, and TM were 70, 100, and 0.8, respectively and (2) classification results with medium spatial resolution images were not desirable, with overall accuracy of only 72.35%, while synergistic use of Landsat5 TM and SPOT5 greatly improved forest classification accuracy, with overall accuracy of 82.94%.

Highlights

Object-based image classification is carried out on the premise that adjacent pixels with similar spectral responses are aggregated into an object which segments the image into nonoverlapping regions
An object-based classification approach is applied, relying on the spectral information of remotely sensed data and making full use of the spatial information, including geometry, texture, and some topographic factors such as slope, aspect, and elevation.[1]. This technique can reduce the “salt and pepper” effect caused by variation of the spectral responses in the same entity,[2,3] especially for very high-spatial resolution (VHR) imagery, in which the same entity is usually represented by pixels with high spectral heterogeneity
The results showed that the optimal scale parameter for the segmentation of TM/SPOT5 was 70 [Fig. 4(a)]

Summary

Introduction

Object-based image classification is carried out on the premise that adjacent pixels with similar spectral responses are aggregated into an object which segments the image into nonoverlapping regions. An object-based classification approach is applied, relying on the spectral information of remotely sensed data and making full use of the spatial information, including geometry, texture, and some topographic factors such as slope, aspect, and elevation.[1] This technique can reduce the “salt and pepper” effect caused by variation of the spectral responses in the same entity,[2,3] especially for very high-spatial resolution (VHR) imagery, in which the same entity is usually represented by pixels with high spectral heterogeneity. Previous research has demonstrated that better results can be achieved using object-based classification than pixel-based classification,[4,5,6] which is widely applied in a complex forest ecosystem for species classification and information extraction.[7,8,9,10,11]

Objectives

Methods

Results

Discussion

Conclusion