A 7.489Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE
Days to hours. New seamlessly integrated software for mobile laser scanning systems can take a company’s workflow and change time from days to mere hours.
The key factor to being successful in 2014 as a mobile mapping firm is scheduling. One of the places where scheduling falters is during the processing and alignment of the point cloud. The time it takes to extract features from the point cloud is determined and factored into each project as a known. The largest time variable is the processing and alignment of the point cloud.
Over five years ago, high-end mobile laser scanning was considered new, unknown technology that was half research experiment and half commercial product. This was the time when the main manufacturers started selling complete mobile laser scanning systems. At first, it took a few extra weeks or months to deliver bleeding edge technology data; it was considered part of the learning curve. Much has changed since the release of this technology. There are more systems and more competition in the marketplace. Time is highly critical, both for reputation and for meeting demanding project schedules that are required to keep the system profitable.
The greatest challenge in the mobile mapping industry is for precise point clouds to be delivered in a timely fashion. The accuracy of the final point cloud depends on three components that define the system’s integrated accuracy. The first component is the laser sensor accuracy of the high end scanner, such as RIEGL’s VQ-450. The accuracy of this scanner is 5mm. Other mobile scanners are as high as 2cm.
The next component to consider is the platform trajectory, which is the weakest link in the accuracy equation. The high-end solution used in the RIEGL VMX-450 system shows a comparably high accuracy of a few centimeters to meters and several mdeg of error, dependent on a variety of conditions. Other equipment used in various systems can have more error, especially in the drift of the inertial measurement unit. This can make them ill-equipped at dealing with any length of varying GNSS conditions.
Finally, the system calibration must be considered, which includes the lever arms and orientation of the sensors. A full calibration for a mobile laser scanning system is provided by some manufacturers while others provide the users with tools to calibrate the system themselves. Commonly, the units calibrated and seamlessly integrated have a higher accuracy.
Challenging environments, such as urban streets, cause the most issues with providing accurate trajectory information. A case study done at an international airport is useful for evaluating how time can be saved through several software approaches that enable users to provide quality point clouds. This area had low quality GNSS available, due to the arrivals area being located underneath the departure area.
The data was collected at the airport and the time intensive work then began. There are four tasks that must take place in any mobile LiDAR workflow after data acquisition. The first task is to copy the data from the acquisition computer to the processing computer. The second task is to make a backup of the raw data. The third task is to process the trajectory. The fourth task is to process the point cloud to the processed trajectory.
In this example, ten driven miles produced eight gigabytes of data to copy for processing and backup. The time to process both the trajectory and to create the unadjusted point cloud took twenty minutes. It should be noted that processing times are dependent on storage media speed, processor, and memory. The initial processed data was only assigned to the processed trajectory, which is much better than the real-time solution but not without issues in areas without strong GPS signal. Due to the multiple passes, the data offsets became visible. These offsets also varied from one area to another, which made adjustment of the data difficult. In some areas, the data was off by 53cm and in other areas, only by 21cm. This made alignment of the data relative to itself very difficult, not to mention to survey ground control.
A few years ago, the task of aligning mobile data was difficult, labor intensive and very time consuming. The typical method employed was the same as most current aerial mapping alignment algorithms, which was a least squares fit between two flight lines or in this case, drive lines. This was a constrained adjustment with only six degrees of freedom. When driving, there were not well defined start and stop points for every drive line and there were more than a few larger offsets between passes. This method was commonly used, but many were not happy with it, nor impressed with the results. This method was decoupled from the integrated system using only information from the laser scanners while relying on the platform trajectory to be accurate.
The next method used was to recouple the link between the trajectory quality and the scanner information. It used segments along the trajectory of the mobile laser scanning data to do a least squares fit between planar objects in each segment. This allowed for multiple, three degree of freedom adjustments to be performed. In the majority of cases, this worked well but there was always some residual that had to be chased out of the data or just accepted. Still, the data was much improved over the simple driveline adjustment method because this segment method allowed for more variation between two passes to exist and to be corrected. This, however, was limited to the amount of segments the data processor was willing to place and align. It greatly improved the point clouds collected in complex environments but was still time consuming and did not fully integrate the sensor information to create a seamless software complement to the hardware integration.
RiPRECISION, a mobile scan-data adjustment software tool, was released by RIEGL LMS to improve the time it takes to correct mobile point cloud data. This automated algorithm, by using all of the available sensor information, allowed for the complete resolution of the errors seen between multiple passes over the entire length of the collected point cloud. This new approach makes it possible to align the point cloud relativity and absolutely to ground control points. The data quality clearly speaks for itself, but there are even larger benefits to this new approach.
In regards to the airport case study data set it was clear after twenty minutes that the post-processed data was not ready for 3D feature extraction, even with the best workforce and software on the market. To resolve these areas, RiPRECISION was employed. The point cloud had over 753 million points and RiPRECISION fused all available sensor information to create a definite trajectory that resolved error, in only twenty-two minutes. This brought the total time to collect, process, and align the point cloud to about one hour of time worked. This created very quick results, considering the numerous corrections that were required.
The benefits of having an efficient way to improve end data quality also need to be considered. The cost associated with data processing is typically not appreciated when considering mobile laser scanning. Data processing for some jobs can consume most of the scheduled time and can really put the feature extraction process behind schedule. The time it took to correct the data with the latest RIEGL workflow, including RiPRECISION, only took one hour of work. Twenty to thirty minutes is all that is required of the system operator and data processor, as most data processing tasks are automated.
Looking back at some of the previous methods discussed, it would have likely taken close to four days of work or thirty-two hours. One of the reasons for this longer time frame is that there are many iterations to run when dealing with least squares adjustment methods. Each reprocessing takes twenty minutes, which is not bad but would still consume 1/3 of each hour of work. The other parts to consider is that in order to isolate sections of data, the collect record needs to be split at the time stamps before and after the some of the larger areas. This is never a fast task and usually would consume the better part of a day on a simple highway project. The airport terminal is much more complex and the errors are more numerous and larger.
RiPRECISION clearly is a time saver for the world of mobile laser scanning. The savings produced through the use of the seamlessly integrated software workflow can be seen through time, cost, and scheduling. By utilizing RiPRECISION to automate the post processing, mobile laser scanning system capacity can be better maintained which greatly improves a company’s bottom line.
Joshua France is the Mobile Systems Segment Manager at RIEGL USA. He has been working with the VMX since the first delivery in 2010 and has helped deploy each of the VMX’s in operation in North America.
A 7.489Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE