|
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
Marrying
Photogrammetry and LIDAR Abstract Much has been written over the last few decades about photogrammetric mapping, with an increased amount of attention in the past several years focused specifically on LIDAR mapping. However, less has been written about the decisions involved in combining these two approaches to complete a single project. To make informed map data purchasing decisions, one must compare and contrast the advantages and disadvantages of both methods. Introduction The mapping industry has regularly taken advantage of emerging technologies to produce products that are faster, less expensive, and offer higher levels of accuracy. New technology also opens up opportunities to introduce new products and applications, and to reach new markets. The foundation of most mapping products are digital terrain models (DTMs), which are required for the generation of contours, and digital orthophotos (DOs). The vast majority of these mapping products involve scales of 1:10,000 or greater. The process of deriving DTMs over the past twenty years or so has seen the mapping industry move from the brute force approach (ground surveying) to the passive-sensing approach (photogrammetry) and, more recently, to the active-sensing approach (LIDAR, IFSAR, etc.). As a result, producers, vendors and end-users have had to explore the most economical and appropriate mix of tools for satisfying various project requirements. The introduction of each new technology has also forced involved parties to deal with the approach of altering their data acquisition, the appropriate production processes and, perhaps more importantly, their collective frames of mind.
Project Requirements Working within a partnership framework, the customer and the data producer must agree on project requirements that produce the desired information in the most appropriate fashion. If only DOs are desired, a DTM can be generated at a relatively low cost that results in the required accuracy. Without further editing, this orthophoto DTM may not be the appropriate terrain model for applications such as hydrologic analyses, and it is rarely adequate for the generation of contours. The table below summarizes the general requirements for DTMs that support products at 1"= 400’ and greater, plus the relative difficulty (cost) of each. With relative costs in mind, it is not unusual for customers to cast project work into adjoining fiscal years to accommodate budgetary constraints. For example, it is common for digital orthophotos and a DTM model to be delivered to support engineering and planning, with contours following at a later date. The key factors of quality control and quality assurance are an important part of any such process. Many approaches commonly employ the use of control points, check points and QC points that can be used to assess the accuracy of all products generated on behalf of the customer. Ground control points required for aerotriangulation may be obtained through ground surveying, whether provided by the customer or collected by the data producer. Additional, manually measured QC points–collected as a result of the aerotriangulation or measured stereoscopically after the fact–are typically used for the QC/QA process. Kinematic GPS surveys performed along roads are an increasingly cost-effective means of providing an independent QC/QA approach for LIDAR data in open areas. In densely vegetated areas, additional ground surveying may be desired. Photogrammetry Photogrammetry has been a standard approach for several decades and has a great many advantages over other approaches, primarily that of leaving a reusable visual record. Over the last 10 years, the introduction of airborne GPS into flight systems has likely been the most dramatic change in the photogrammetric approach, reducing the amount of ground control required during the aerotriangulation process. The majority of photogrammetric production processes are still largely labor-intensive. DTM extraction via digital image-correlation techniques is in only moderate use and still requires interactive stereo editing. For larger map scales, the extraction of break lines is also required to ensure that derived contours behave properly. Digital orthophotos also commonly require specialized break lines for relief-displaced features such as buildings, bridges and overpasses. Planimetric data extraction is handled almost entirely by skilled operators on stereoscopic editing workstations. Photogrammetric operations are limited by weather, plus they require "leaf-off" conditions to allow collection of data through vegetation. Generally speaking, only 10,000 to 25,000 frames can be shot per year per camera system due to these constraints. Assuming that the typical medium-to-large project includes about a thousand frames, only 10 to 25 projects can be handled each year per system. LIDAR Since its introduction into the mapping marketplace around five years ago, the use of LIDAR has become increasingly common in mapping projects. Refinements in instrumentation technology, data processing, quality control/quality assurance mechanisms, and applicable accuracy standards have contributed to its increased acceptance within the mapping industry. As LIDAR instrumentation has matured through its increased dependability and sophistication, configurations appropriate for mapping efforts have similarly evolved. The newest instruments provide a capacity for multiple returns that supports canopy analysis and determination of the ground surface. In addition, the intensity data for each pulse is now captured and can be displayed as an intensity image to aid in data editing. Post-processing software, once largely proprietary among LIDAR system operators, is now more widely available. A standard data interchange format for LIDAR data is also on the market. Over the past several years, operators and customers have identified the proper means for quality control in generating truly effective and accountable results. To that end, progress has been made in developing vertical mapping accuracy standards that are applicable to LIDAR-derived surfaces. Integrated Systems Small-format metric digital cameras (generally 4K x 4K, panchromatic and/or color) typically accompany many of the integrated LIDAR systems in operation today. This imagery is used both as a visual record and as a basis for the stereo editing of LIDAR data. In some cases it can also be used as a source for digital orthophoto generation. Assuming the rigid mounting of the laser instrument, IMU and digital camera, aerotriangulation of the digital camera images may be unnecessary. This effectively creates a self-sufficient system, depending upon project requirements. However, reliance on images taken with small-format, metric digital cameras–captured simultaneously with LIDAR data–is not expected to be the norm. To limit the LIDAR capture to photographic-quality weather will hamstring the process. Use of a large-format-film camera allows photo missions to be flown higher, at faster air speeds, and with more ground coverage. As a result, some aircraft come already configured with a second porthole for just such a camera. Note that using a single porthole for both photogrammetric and LIDAR instruments, "swapping" them as required, is generally not tenable. This swapping process is time-consuming, and both photogrammetric cameras and LIDAR systems suffer if continually removed and re-installed. LIDAR and photogrammetric cameras are generally not used simultaneously since the photography oftentimes requires a different flying height to get the proper photographic scale. However, daylight photography can be augmented by nighttime LIDAR acquisition, or by LIDAR data acquired on days not amenable to photography. General Process Comparison The photogrammetric approach has the advantage of supporting a self-contained process flow. All products can be generated without reliance upon other data sources, assuming that heavy vegetation cover is not present. Skilled operators collect data–elevations, position and attribution–directly into the format that is required for completion and/or delivery. The net effect is the collection of an optimal amount of data required, thus reducing data storage and resulting in an efficient and simplified characterization of the terrain. There is an abundance of production software on the market for compiling photogrammetric DTMs. LIDAR indiscriminately collects highly dense sets of discrete elevation points that represent a sampling of the surface. Since LIDAR collects points at this surface level, the data must be reduced to bare earth levels by removing data points that fall on man-made structures. In vegetated areas, multiple returns can be analyzed to determine the ground elevation. For contour generation, filtering is often performed to remove elevations that would otherwise result in rugged, thereby misleading contours. For map scales greater than 1" = 400’, imagery is often required for final editing and/or break line introduction. In some cases this effect is handled using a small-format digital camera; in others it is handled by film photography as required by other project work. While LIDAR can provide a very dense and detailed surface model, the large number of generated points can be a concern in production. Most of the standard mapping industry software packages do not easily deal with this large amount of data. As a result, a number of companies and vendors are now providing their own filtering of the LIDAR data as well as reducing the data points to a minimal number. Comparative Strengths and Weaknesses The table below summarizes the general strengths and weaknesses of LIDAR versus photogrammetric approaches: The primary advantages of the photogrammetric approach are the visual record it produces, and the ability to generate storage-efficient DTMs, delineate breaks in the terrain, and perform feature extraction. The extended operation times (both time-of-day and seasonality), the ability to "see through the trees," and its rapid response capability are the primary advantages of LIDAR. As a consequence, LIDAR can provide a project-scheduling advantage in some cases. Aerial photography must nearly always be done during leaf-off conditions, thus limiting the days of acquisition each year. On the other hand, LIDAR can be flown many days during the season and can occur either before or after the aerial photography. If LIDAR acquisition and basic processing is involved, the DTMs can be edited and finalized shortly after the aerotriangulation of the photography is complete. In the case of an integrated LIDAR system accompanied by a small-format digital camera, a final DTM can be available by the time the photography is complete. Combined Approaches One question remains: "When and where do we use LIDAR?" With both benefits and shortcomings in mind, it is possible to design a project that takes advantage of the strengths of each process. Some product types make for simple designs. If planimetric data extraction is required for a project, photogrammetric acquisition (at a minimum) is required so that skilled stereo compilers can collect the position and attributes of features. Pure LIDAR products are normally used when schedules outstrip the need for a traditional large-scale map product. LIDAR data filtered to a bare-earth model, or in its raw form, can be gridded to create DEM files for viewing as shaded relief or color-coded elevation data. These forms of data can be useful in an emergency situation such as a natural disaster. The table below lists a number of project scenarios and possible solutions: These associated figures show how obscured areas in traditional stereo compilation can be supplemented with LIDAR data. This project involved both plan- and surface collection with areas of very dense vegetation. Break lines were collected stereoscopically in open areas, and bare-earth LIDAR data was collected to fill in the obscured regions. Summary LIDAR technology has reached an important milestone through its wide acceptance and support by many vendors. However, LIDAR has not proven to be the universal, speedy and low-cost replacement for photogrammetry that was originally expected. Instead, mapping firms should leverage the strength of both technologies to create a product that meets accuracy, data structure, cost, and schedule requirements in order to maximize benefits to their customers. Careful project planning is necessary to ensure the appropriate use of each respective technology. Capital investment in LIDAR can reach more than a million dollars, with a life span of from five to seven years. When purchasing a map product it should be clear that a project’s requirements, not the provider’s need for return on investment, should drive the use of any technology. For example, it is currently not possible to exceed the approximate 15-centimeter accuracy level due to GPS error. There is potential for further advancement in processing and display of LIDAR. The fusion of LIDAR and imagery in automated processes will continue to improve, and it may yield levels of automation that provide vector data at significantly lower costs. About the Authors: Scott Merritt is chief technologist of the Surdex R&D staff and is charged with implementing technologies and tools at all levels of production within the company. He was awarded a Bachelor’s degree in Photography from the Rhode Island School of Design and has 10 years of experience in image processing and map production. Angelo Corrubia is the sensor specialist on the Surdex R&D staff, responsible for integrating new sensor technologies such as LIDAR into the production environment within the company. He holds a Master’s degree in Electrical Engineering from Washington University and has nearly 20 years experience in both sensor technology and system development for mapping applications. |
Left:
Photogrammetric compilation showing obscured areas in heavy vegetation.
Right: LIDAR data used to fill in the obscured areas.