Abstract
Abstract. A novel algorithm for forest road identification and extraction was developed. The algorithm utilized Laplacian of Gaussian (LoG) filter and slope calculation on high resolution multispectral imagery and LiDAR data respectively to extract both primary road and secondary road segments in the forest area. The proposed method used road shape feature to extract the road segments, which have been further processed as objects with orientation preserved. The road network was generated after post processing with tensor voting. The proposed method was tested on Hearst forest, located in central Ontario, Canada. Based on visual examination against manually digitized roads, the majority of roads from the test area have been identified and extracted from the process.
Highlights
The characterization of forest roads is important for forest management and wildlife habitat mapping
A number of studies were carried out to automatically extract roads from remotely sensed data, most of them were focused on extract road centerline from the urban area using object based classification or edge detection focused morphological methods. (Shi et al, 2014) (Beck et al, 2015) The classification methods utilized the spectral signature of the road segment to separate it from other objects
Song and Civco used support vector machine (SVM) to classify the imagery and further detect road networks. (Song and Civco, 2004) Yuan et al proposed a multistage process based on robabilistic SVMs and salient features to extract road networks from satellite imagery. (Yuan et al, 2011) Huang and Zhang proposed the detection of road centerlines from high-resolution images by integrating multiscale structural features and SVM. (Huang and Zhang, 2009) One the other hand, the edge detection focused approached utilize the road as line feature where the shape feature was used to distinguish it from the background
Summary
The characterization of forest roads is important for forest management and wildlife habitat mapping. (Huang and Zhang, 2009) One the other hand, the edge detection focused approached utilize the road as line feature where the shape feature was used to distinguish it from the background In this aspect, Shi and Zhu proposed a line segment match rule to extract urban road networks from morphological processing. (Chaudhuri et al.,2012) The existing methods were based on the facts that roads are linear features with edges and have different spectral signatures from surrounding objects Such approaches may work well with major roads built with concrete or pavement, but fail to detect roads with vegetation coverage or under tree canopies shadows. Such separation is important especially for wildlife habitat mapping, as lifted roads representing different ecology with more herbaceous. (Fahrign and Rytwinski, 2009) The development of airborne LiDAR (Light Detection And Ranging) technology provides good opportunity to improve forest roads identification due to its ability to capture the vertical structures of surface objects complementary to the spectral information from passive optical imagery
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
