Analysis of LIDAR Data Fused with Co-Registered Bands

In the past decade, LIght Detection And Ranging (LIDAR) has been recognised by both the commercial and public sector as a reliable and accurate source for land surveying. Object classification in LIDAR data tends towards data fusion by employing additional simultaneously recorded bands. In this paper, a rule-based approach is presented for improving classification accuracy obtained in a supervised Maximum Likelihood classification. Simultaneously recorded co-registered bands are used such as high resolution LIDAR first, last echo and intensity data, aerial and near infra-red photos. Issues regarding feature and class selection and differentiated accuracy assessment are addressed. Furthermore, the individual influence of each band on the classification is investigated. The results show that incorporating additional knowledge and considering contextual relationships among classes is beneficial for improving classification accuracy in fused LIDAR datasets.