Doorkeeper

Seminar_Geoinformatics Unit _Talk by Dr.Nadeem FAREED.

Wed, 15 Dec 2021 10:00 - 10:30 JST
Online Link visible to participants
Register
Free admission
Registration closes 15 Dec 8:00

Description

This is an online seminar. Registration is required.
【Geoinformatics Unit】
【Date】2021/Dec/15(Wed) 1000am-1030am (JST)
【Speaker】 Dr. Nadeem FAREED

【Zoom】https://zoom.us/j/91406604351?pwd=aWtwZFhRL21XM25yMzl4VWtmeUE4UT09

Title:
Development and Evaluation of LiDAR Based Drainage Structures Mapping Algorithm (DSMA) for Culvert-modified DEM Generation
Abstract:
Bridges and culverts are collectively known as drainage structures (DS). Globally, most of the DSs are completing their designed age and posing a potential threat of failure under extreme weather conditions. Therefore, wide-area drainage structure (DS) mapping is essential, though wide-area DS locations are frequently unavailable making their inspection and management on time more challenging. Remotely sensed data along with Geographical Information systems (GIS), and Global Navigational Satellite Systems (GNSS) have been reported for mapping DS data through manual methods. In recent times, the increased use of airborne laser scanning (ALS) data has been witnessed for DSs mapping using ALS-based digital elevation models (DEMs) and hill shade images, respectively. However, a unified, automated algorithmic-based DS mapping solution is not developed and yet the need for such a method is frequently reported in the past. The manual methods were reported to be slower, constrained by weather conditions and funding sources for limited study sites as investigated in the past. To address the aforementioned challenges in the context of DS mapping, the present research primarily focusing the development of an automated DS mapping algorithm (DSMA) to map DSs in wide-area capacity. The ALS three-dimensional (3D) portrayal of topography owing to the success of DS mapping in the past, therefore, classified ALS point clouds along with road centerlines were used for the development of DSMA. In the context of DS mapping using ALS data, the roads were found to be elevated than neighboring ALS ground points posing a potential challenge in drainage network (DN) and afterward DS mapping, respectively. Thus, automated DSMA initiates by eliminating ALS ground points of roads first. The number of different buffers values representing the road widths of highways, freeways, private roads, residential streets, etc., were estimated using the ALS ground points. Then a combined road mask is generated using buffers values and road centerlines, respectively. Then the combined road mask is used to remove all the ground points belonging to roads in the ALS data. The remaining ground points are then interpolated to create a newly developed ALS-modified DEM (ALS-mDEM). A drainage network (DN) is then derived from the ALS-mDEM automatically using the flow-direction, flow-accumulation, and Strahler stream-order algorithms, respectively. Several stream order thresholds were then tested to map candidate DSs by intersecting the DN with the road centerlines. In the final step of the DSMA development, different DS refinement buffers were generated and tested to clean the mapped DSs from duplicate DS records. A refinement buffer of 15 m is then finally selected through DS refinement analysis. The performance of DSMA was assessed in wide-area capacity under two different geographical settings for a total area of 50 km² in Vermont, USA, including an urban site and a rural site, respectively. The classified ALS point clouds were acquired from Unite State Geological Survey (USGS), while the road functional classification scheme of the Federal Highway Administration (FHWA), and was obtained from a public data portal of Vermont. The non-FHWA roads are comprised of private roads and residential streets that were out of the jurisdiction of FHWA i.e., unavailable from a public data portal, therefore, were automatically mapped using impervious surface (ISA) of land use land cover (LULC), respectively. Finally, the DS dataset comprised of bridges and culverts was gathered from the Vermont agency of transportation along with FHWA roads, and the DS dataset along with non-FHWA roads was digitized from Google Earth Street View (GE-SV) images, respectively. The benchmark DS dataset was used to assess the positional and prediction accuracies of the mapped DS of the DSMA method. Based on the one-to-one correspondence between a benchmark DS and corresponding mapped DS, the Euclidean distances were computed to assess the positional accuracies of the mapped DS compared to the benchmark DS dataset, respectively. The mean positional accuracy for the urban site and rural sites were 13.5 m and 15.8 m were reported for both geographical settings. The prediction accuracies of the mapped DS in terms of F1 scores were calculated along with FHWA and non-FHWA roads separately. The F1 scores were 0.87 and 0.94 for the urban site and rural site respectively, for the FHWA administrated roads. The F1 scores were found to be 0.72 and 0.74 for the urban site and rural site respectively, for the non-FHWA roads.
ALS ground points are in great use to create High-resolution digital elevation models (HR-DEMs). However, in the application setting of hydrology and geomorphology, HR-DEMs require post-processing to create culvert-modified DEMs suitable for hydrological and geomorphological investigations, respectively. A culvert-modified DEM is processed by incorporating mapped DSs in an HR-DEM at the post-processing stage. Nevertheless, in the absence of a DS dataset in wide-area capacity, the creation of culvert-modified DEMs becomes challenging. Instead, the breach algorithm (BA) method is a standard procedure to obtain culvert-modified DEM when DSs data is not available to solve the problem to some extent. In the second part, the present research assesses the availability of DSs data in wide-area capacity originating from the DSMA to process a culvert-modified DEM comprised of an area of 36 km² in Vermont, USA. Benchmark DS dataset is used as a standard reference to assess the performance of the DSMA method. Furthermore, the research is extended to compare the DSMA method performance with the Breaching Algorithm (BA) method, which is a standard procedure to obtain a culvert-modified DEM when DSs data is unavailable. A new and automated methodological framework is then developed to generate culvert modified DEM using mapped DS of the DSMA method. Furthermore, the performance of the DSMA method compared to the BA method is assessed in the context of producing culvert-modified DEMs, respectively. The DS found from the culvert-modified DEMs of the DSMA and BA methods were analyzed to quantify the performance of both methods. To assess the performance of both methods, DS found from culvert-modified DEM were classified as true positive (TP) for the correct solution, false positive (FP) for the wrong solution, and false-negative (FN) no solution found by both methods, respectively. Using TP, FP, and FN originating from the DSMA and BA method, the commission error i.e. false positive rate (FPR), and omission error i.e., false-negative rate (FNR) were found. The reported evaluation matrices indicate that the newly developed DSMA-based DS data showed an FPR of 0.05 along with FHWA, and 0.12 with non-FHWA roads, respectively. In terms of FNR, The DSMA showed 0.07 along with FHWA, and 0.38 along with non-FHWA roads respectively. Compared to the DSMA, The BA method showed an FPR of 0.28 along with FHWA, and 0.62 along with non-FHWA roads, respectively. In terms of FNR, the FNR the BA method scores 0.32 along with FHWA, and 0.61 with non-FHWA roads, respectively. Based on the evaluation matrices of FNR, and FPR, the DSMA method has proven to be more accurate than the BA method, respectively. Thus, the formulated framework for producing culvert-modified DEM using DSMA was robust compared to the BA method.
Keywords: ALS point clouds; Bridge; Culvert; Algorithms; High-resolution DEM

About this community

RIKEN AIP Public

RIKEN AIP Public

Public events of RIKEN Center for Advanced Intelligence Project (AIP)

Join community