3D laser scanning and surveying adjustment in traffic infrastructure management

In view of difficulties in automatic and refined management of traffic infrastructure, road lamps and signboard in the test road section are taken as examples, an application framework for automatic and refined management of traffic infrastructure is proposed by using 3D laser scanning technology combined with GPS (Global Positioning System) and QGIS (Quantum Geographic Information System). The advantages of 3D lidar sensor, such as high precision, strong anti-interference ability and good real-time performance are fully used. Through data fitting by using total least squares method, digital management of infrastructure is carried out and its precision is high. Combined with GPS and QGIS, the spatial information of infrastructure is developed. In this way, manpower and efficiency of management of infrastructure are saved and improved.


Introduction
At present, low degree of automation and refinement are problems of traffic infrastructure management. The traditional manual detection is time-consuming and laborious, and the degree of automation and refinement is not enough, which can not meet the development needs of traffic infrastructure management in the future. As a new data acquisition way, 3D laser scanning technology has the advantages of high precision, good real-time performance, strong anti-interference ability and high efficiency [1] , which is widely used in road point cloud segmentation [2] , calculation of forest canopy area [3] , landslide analysis [4] , and analysis of tunnel deformation monitoring [5] , etc. However, there is little research on the application of 3D laser scanning technology in traffic infrastructure automation and refinement management. In reality, many tests rely on manpower. For the reason above, application framework of 3D laser scanning technology in traffic infrastructure management is designed, which focuses on using advanced 64-line 3D lidar sensor equipment to collect point cloud and processing the point cloud data. Data processing is divided into two aspects, one is fitting the point cloud data of transportation infrastructure, getting its state equation in space, and recording it in digital form. The other is calculating the geographical location of transportation infrastructure, recording and managing the location information by combining GPS and QGIS.
There are some advantages of this framework as follows: (1) Compared with two-dimensional camera, the data collected by the 3D lidar sensor contains depth information and strength information, which is more comprehensive and rich, and data quality and accuracy are also higher. Moreover, the 3D lidar sensor is not affected by light and has better environmental adaptability; and (2) Compared with modeling, data fitting can extract more information of traffic infrastructure; compared with RFID (radio frequency identification) technology [6] , the real-time performance of this technology is better, and the status information of infrastructure can be updated in real time; and (3) Combined with GIS and GPS technology, not only the storage and management of traffic infrastructure data is more convenient, but also the spatial information can be utilized, which provides great convenience for managers.

Methods
The structure design of framework proposed in this paper is shown in Figure 1. The whole framework is composed of three layers: perception layer, network layer and application layer. The perception layer belongs to the device side, which is the front end of the overall framework. 3D lidar sensor collects the point cloud, which will be transmitted to the network layer. The network layer belongs to the algorithm side, which is the data processing part. One part of data processing is data fitting. The total least squares method is taken as the fitting criterion, the point cloud data is fitted and the central coordinate information will be extracted. The other one part is calculating the geographical location. Combined with GPS, the spatial position information can be obtained and transmitted to the application layer. The application layer belongs to the platform side, storing the traffic infrastructure information and facing to the managers.

Visualization tool of point cloud: Veloview
Veloview is a kind of point cloud data reconstruction software, which can visualize and process 3D data captured by HDL-64E in real time. The measurement results of lidar sensor are returned in 3D position information and attribute data, such as laser ID, point ID, azimuth, timestamp, etc. The real-time point cloud image displayed by lidar sensor in veloview is shown in Figure 3. As can be seen from Figure 3, in addition to the real-time point cloud image, the point ID, 3D coordinates (X, Y, Z), azimuth, distance, intensity, laser ID, vertical angle and other information can also be checked.

Criterion of data fitting
As for data fitting, traditional least squares (LS) method only considers the measurement error of the observation value. For the data measured by the lidar sensor, there are measurement errors existing in both coefficient matrix and observation value, so the traditional least squares method can not be used as the fitting criterion, but the total least squares(TLS) method is appropriate, which considers both measurement errors [7] . The difference is shown in Figure 4.  Figure 4 Difference between LS and TLS method

Total least square model and its solution
As for linear equation system , the error of coefficient matrix B is set E B , and the error of observation L is set E L . In this case, the linear equations can be written as: (1) Equivalent to: . m is the number of observation value and n is the number of unknown parameters.

, ∆
,formula (2) can be written as: Singular value decomposition of C can get: The rank of augmented matrix C is n+1 because 0 . According to Eckart-Young-Mirsky matrix approximation theorem [8] [9] , The best approximation matrix B L of matrix [B L] must be decomposed into: The constraint criterion of the total least squares method is: Among formula (6), ‖ ‖ is Frobenius norm. Its correction is: ⃗ are the n+1 column of matrix U and V respectively. So it is easy to get that the rank of the residual error is 1. So: Then the total least squares solution of unknown parameters can be obtained from the n+ 1 column of V, which is: So the total least squares solution is: According to the steps above, the collected data can be processed programmatically.

Selection of calculation tool
Matlab is selected as the calculation software, and the calculation results are presented in the form of three-dimensional scatter diagram and optimum fitting result diagram.

Application layer
QGIS is selected as the application layer of the framework. As one kind of GIS, there are many advantages in QGIS, such as user-friendly interface, supporting cross platform and operating system, and supporting dozens of GIS data file formats. In addition to supporting rich data formats, data visualization, editing and analysis are also supported by QGIS, which is convenient for managers to query and meet the needs of management. OpenStreetMap online map can be imported and 4326 GCS_GPS_WGS_1984 geographic coordinate system is selected in QGIS. The interface of software is shown in Figure 5.

Test road section
Xinmin Road of Tsinghua university, lontitude from 116.32448 to 116.32456 and latitude from 40.00000 to 40.00109, is taken as the test road section. Display on OpenStreetMap in QGIS is shown in Figure 6 and live picture is shown in Figure 7.   Figure 7 Road condition of test section

Test vehicle
Digital road integrated acquisition vehicle, mounted with lidar sensor HDL-64E is shown in Figure 8.

Point cloud data in perception layer
The point cloud data collected by HDL-64E is shown in Figure 9. When selected, the point cloud will be highlighted immediately. Then, the data information such as 3D coordinates, azimuth and distance of the point cloud to be detected can be extracted as the original data for data fitting.  Figure 9 Point cloud picture(point cloud to be calculated was marked in red dotted box)

Fitting results of road lamps
The fitting algorithm of point cloud data in Matlab is shown in Figure 10.   (116.32463,40.00030). Based on Geodesy, when the horizontal distance is less than 10km, the geoid or ellipsoid can be replaced by a horizontal plane, regardless of the influence of the earth curvature. On the horizontal plane, when the difference of X coordinate is 1m, the difference of longitude is 0.00001 degree, when the difference of Y coordinate is 1.1m, the difference of latitude is 0.00001 degree. The calculated longitude and latitude of the road lamp 1 are (116.32451°, 40.00038°). Similarly, the same processes are done on the other two road lamps in the test section of Xinmin Road as shown in Figure 12, and the fitting results are shown in Figure 13 and Figure 14.

Figure 12
The other two road lamps and one signboard (The signboard is marked with white dotted line frame and the road lamps are marked with red dotted line frame)

Fitting result of sign board
Picture of signboard's point cloud is shown in Figure 15.

Figure 15
Picture of sign board's point cloud Point cloud data of signboard is extracted and the fitting algorithm is shown in Figure 16, and the fitting result is shown in Figure 17.

Results displayed in application layer
The infomation obtained from Table 2 is displayed in the form of layer in QGIS, which is shown in Figure 18. Traffic infrastructure is displayed in the form of red dot, whose detailed information can be queried in open attribute table. Picture of the attribute table is shown in Figure 19. Picture of the attribute table

Conclusions
In this paper, road lamps and signboard in the test section are taken as examples, an application framework of traffic infrastructure management is put forward based on 3D laser scanning. The framework is introduced comprehensively from the perception layer, network layer and application layer. The advantages of 3D laser sensors, such as high precision, good real-time performance and strong antiinterference ability are brought into play well. Therefore two conclusions can be drawn as follows: (1) This paper focuses on the application of 3D lidar sensor equipment in traffic infrastructure management, gives the specific usage of 3D laser scanning technology in automatic and refined management of traffic infrastructure, and expresses the status and location information of infrastructure in digital form through data fitting.
(2) Combined with the use of GPS and QGIS, the spatial information of infrastructure can be developed and utilized, and the data is collected by integrated acquisition vehicle instead of manual detection, which can save time and effort, improve the management efficiency, and provide a certain solution for traffic infrastructure management department in the future.