Development of panchromatic image classification techniques for land-use applications

在许多土地利用应用中，图像分类是必要的。自从多光谱图像出现以来，人们在优化技术方面做出了重大努力，以使其光谱特性得到充分利用。但是，在对土地变化进行长期研究时，需要使用大量的全色图像（来自航空照相机和空间传感器）。此外，还有新的传感器，如机载激光雷达和SAR系统，也提供全色图像，需要充分利用。该项目的目标是发展改进的方法和技术，以便利用从不同遥感传感器捕获的全色图像进行图像分类应用。拟议的项目旨在实现四个主要目标。首先，评估来自不同传感器的全色图像的特征。然后，研究了多光谱图像分类方法在全色图像分类中的应用。其次，针对来自不同传感器的全色图像的特点，开发和测试新的全色图像分类方法。创新的新技术，如超级分类器算法也将成为目标。第三，针对激光雷达主动传感器技术，本项目提出开发激光雷达图像强度目录，研究传感器高度、激光束入射角和目标反射率对图像强度值的影响，将其纳入图像分类过程。第四，本项目将利用点/线性特征评估不同图像对图像配准模型在分类过程中对不同全色图像进行整合的适用性。总的来说，该项目的研究成果将有助于遥感图像分类的理论和实践，并将引起社会对全色图像在不同应用中的使用的关注。

Image classification is necessary in many land-use applications. Since the advent of multi-spectral imagery a significant effort has been made in optimising techniques to allow their spectral characteristics to be fully exploited. However, there is a vast volume of panchromatic imagery (from aerial cameras and space borne sensors) that would need to be used when long-term studies of land-change are to be undertaken. In addition, there are new sensors such as airborne LIDAR and SAR systems that also provide panchromatic imagery and need to be fully exploited. The goal of this project is to develop improved methods and technologies for the use of panchromatic imagery captured from different remote sensing sensors for image classification applications. The proposed project aims to achieve four major objectives. First, the characteristics of panchromatic imagery from different sensors will be evaluated. Then, methods used for classifying multi-spectral images will be examined for panchromatic image classification. Second, new methods for panchromatic image classification will be developed and tested with respect to the characteristics of panchromatic imagery from different sensors. Innovative new techniques, such as hyper classifier algorithms will also be targeted. Third, with regards to the active sensor technology of LIDAR, the project proposes the development of a LIDAR image intensity catalogue by investigating the effects of sensor altitude, incidence angle of the laser beam, and the target reflectivity on the image intensity values so they can be integrated into the image classification process. Fourth, the project will evaluate the applicability of different image-to-image registration models using point/linear features to integrate of different panchromatic images in the classification process. Generally, the results of the project will contribute to both theory and practice of remote sensing image classification and they will attract the society's attention to the use of panchromatic imagery for different applications.