MODERN MOBILE MAPPING

download MODERN MOBILE MAPPING

of 12

Transcript of MODERN MOBILE MAPPING

  • 8/14/2019 MODERN MOBILE MAPPING

    1/12

    GPS - IMAGE PROCESSING IN MOBILE

    ABSTRACT

    Mobile Mapping Systems have been widely used to map transportation

    infrastructure features for several years. First generation Mobile Mapping Systems,

    however, could not provide positioning accuracy better than one meter, especially in

    built-up urban areas. Thus, these data sets have been predominantly used to feed various

    GIS systems, primarily concerned with infrastructure inventory and facility management.

    With recent technological developments such as improving imaging sensors and,

    more importantly, the introduction of ring-laser gyro inertial systems. The prototype-

    positioning component of the system is based on a tightly integrated GPS/INS system,

    and the imaging component comprises a single downlooking high-resolution, 1K by 1K,

    digital camera.

    The process of automatically identifying centerlines, extracting image features

    and matching them is demonstrated on a variety of data sets, indicating clearly that the

    algorithmic performance has reached a sufficient threshold such that human interaction is

    no longer required, and consequently, the only limiting condition of the real-timeimplementation is the available computer processing power.

    1

  • 8/14/2019 MODERN MOBILE MAPPING

    2/12

    ABSTRACT

    INTRODUCTION

    SYSTEM CONCEPT

    PERFORMANCE OF THE AUTOMATED IMAGE SEQUENCE PROCESSING

    HARDWARE IMPLEMENTATION

    COLOR SPACE TRANSFORMATION

    CENTERLINE EXTRACTION

    FEATURE POINT EXTRACTION

    STRIP FORMATION

    POSITIONING PERFORMANCE

    SUMMARY AND CONCLUSION

    2

  • 8/14/2019 MODERN MOBILE MAPPING

    3/12

    INTRODUCTION

    Direct georeferencing of imaging sensors by means of integrated GPS/INS has

    been in the spotlight in the surveying/mapping and remote sensing communities since the

    mid-nineties (He et al., 1994; Bossler and Toth, 1995; El-Sheimy and Schwarz, 1999;

    Schwarz, 1995).

    One reason is that the primary driving force behind this process is a need to

    accommodate the new spatial data sensors, such as LIDAR or SAR (airborne systems).

    The second reason is that a substantial cost decrease, a possibility of data

    reduction automation, and a short turn-around time are the most attractive features

    offered by this technology

    The main features of the system are the high image capture rate, the online use of

    navigation estimates, and the on-the-fly image and stereo data processing. From a

    navigation standpoint, the post processing of GPS/INS data provides more accurate

    orientation as a benefit of forward and backward trajectory processing and precisely

    synchronized timing information.

    The high-accuracy GPS/INS/CCD system designed for monitoring linear highway

    features is based on the concept of tight sensor integration, combining post-processing

    with real-time image processing. The two primary components of the mobile mapping

    system currently being implemented are precise navigation and digital imaging; both

    allow for flexible and optimal system design, leading potentially to near-real time overall

    data processing.

    3

  • 8/14/2019 MODERN MOBILE MAPPING

    4/12

    SYSTEM CONCEPT

    Mobile Mapping Systems are built on the concept of combining high-performance

    georeferencing with electronic imaging on a moving platform. MMS systems have been

    using image sequences for a long time since it is an essential part of the concept.

    However, progress toward the automation of the image sequence processing has been

    slow for two reasons.

    First, economy; the actual feature extraction represents approximately less than

    20% of the overall cost, and therefore, the financial motivation is weak.

    Second is the varying image scale, which makes the object recognition task quite

    difficult in the feature-rich object space.

    Traditional MMS systems work with forward- or side-looking cameras, while our

    system uses a down-looking camera. This way the image scale changes very slightly and

    there is an almost constant scale along the vehicle trajectory. The object contents of the

    images are rather simple and predictable, such as the line marks, the primary interest to

    us, surface texture variations, cracks, potholes, skid marks, etc.

    Figure 1 shows the generic model of the dedicated centerline mapping system.

    HARDWARE IMPLEMENTATION

    The prototype of the integrated GPS/INS/CCD system designed for precision

    monitoring of the highway edge- and centerlines comprises two dual-frequency Trimble

    4000SSI GPS receivers and a medium-accuracy and high-reliability strapdown Litton

    4

  • 8/14/2019 MODERN MOBILE MAPPING

    5/12

    LN-100 inertial navigation system, based on Zero-lockTM Laser Gyro (ZLGTM) and A-

    4 accelerometer triad (0.8 nmi/h CEP, gyro bias 0.003/h, accelerometer bias 25 g) .

    Estimation of errors in position, velocity, and attitude, as well as errors in inertial

    and GPS measurements, is accomplished by a 21-state centralized Kalman filter that

    processes GPS L1/L2 phase observable in double-differenced mode together with the

    INS strapdown navigation solution.

    The estimated standard deviations are at the level of 2-3 cm for position

    coordinates, and 5-7 arcsec and ~10 arcsec for attitude and heading components,

    respectively. The imaging component is built around the Basler A201 camera, Kodak 1K

    by 1K colour CCD with 9.07 mm by 9.16 mm imaging area (9-micron pixel size) and 15

    images per second acquisition rate (15 Hz), which allows for 60% image overlap at

    normal highway speed.

    Figure 2 shows the system design, including the sensors, dataflow, processing steps.

    PERFORMANCE OF THE AUTOMATED IMAGE SEQUENCE PROCESSING

    5

  • 8/14/2019 MODERN MOBILE MAPPING

    6/12

    To assess the feasibility of automated line extraction with 3D positioning and

    consequently its real-time realization, a rich set of the potential image processing

    functions was developed in a standard C++ programming environment.

    Figure 3 shows the overall dataflow and processing steps, which will be

    illustrated in more detail later. In short, the real-time image processing is feasible due to a

    simple sensor geometry and the limited complexity of the imagery collected.

    COLOR SPACE TRANSFORMATION

    Color images are preferred over monochromatic ones. Figure 4 illustrates various

    cases, including an extreme situation where the yellow solid lines are hardly visible in the

    B/W image. A simple histogram analysis in the RGB (Red, Green, Blue) color space

    easily reveals two peaks in the red and green channels representing the yellow color of

    the centerline.

    6

  • 8/14/2019 MODERN MOBILE MAPPING

    7/12

    Although there are many image processing algorithms working with various color

    data (multichannel gray-scale imagery), the great majority of the core functions work

    only on simple monochrome image data. Therefore, if possible, a color space conversion

    is desirable, in other words, moving from the 3D color space into one dimension, a color

    direction, which shows the best possible separation for the objects we want to distinguish.

    Since the quality of the extracted centerlines still show visible differences, a

    filtering process has been implemented to remove this dissimilarity. The output of the

    median filter is converted to a binary image. End results show no significant difference

    between the centerline segments extracted from the very different images as illustrated in

    Figure 6.

    7

  • 8/14/2019 MODERN MOBILE MAPPING

    8/12

    CENTERLINE EXTRACTION

    After the RGB to S transformation and filtering, the geometry of the centerlines is

    extracted from the binary images For a raster line, its centerline is of primary interest.

    Skeleton is more like a one-pixel-width line, while centerline can be used to express a

    one-pixel-width line in both raster and vector lines.

    The skeleton is then extracted by shrinking the raster line from its boundary in all

    directions until the one-pixel-width eight-connected line remains. In the medial axis

    transformation method (Montanvert, 1986), the discrete medial axis pixels are the local

    maximum of a transformation value.

    A robust recursive filtering technique can eliminate noise such as gaps, although

    most of the gaps and grey-scale irregularities already have been removed during the color

    space transformation, as well as provide segmentation for multiple centerlines such as

    double solid lines. Figure 7 depicts the results of this processing step.

    Once boundary points are extracted, a line-following routine can generate the boundary

    lines, which are subject of further cleaning such as removing irregularities by applying

    geometrical constraints. In the final step, the midpoints are computed and the centerline is

    extracted as shown in Figure 8.

    8

  • 8/14/2019 MODERN MOBILE MAPPING

    9/12

    FEATURE POINT EXTRACTION

    To achieve the highest accuracy possible, the 3-dimensional centerline positions must be

    obtained from stereo imagery. Knowing the camera orientation, both interior and exterior,

    and the matching (identical) entities between the 2-dimensional centerlines, the 3-

    dimensional centerline position can be easily computed.

    I denotes the smoothing operation on the grey level image I(x, y). Ix and Iy indicate the x

    and y directional derivatives respectively. Figure 9 depicts feature points extracted

    around the centerline region from overlapping images.

    9

  • 8/14/2019 MODERN MOBILE MAPPING

    10/12

  • 8/14/2019 MODERN MOBILE MAPPING

    11/12

    POSITIONING PERFORMANCE

    The multisensor system calibration is a key task to achieving the ultimate

    accuracy of the given sensors. System calibration is defined here as the determination of

    spatial and rotational offsets between the sensors as well as imaging sensor calibration.

    Continuous calibration of the INS system is provided by GPS and thus is very dependent

    on GPS anomalies such as satellite signal obstructions, multipath, interference, etc. The

    effective ground pixel size was about 2-4 mm. Figure 11 shows the calibration range, and

    the road area with control points.

    A comprehensive analysis of the system calibration and positioning performance

    are available in (Grejner-Brzezinska and Toth, 1999, and 2000). Using these boresight

    parameters, the comparison of ground coordinates obtained by the photogrammetric

    methods from the directly oriented imagery to the GPS-measured ground truth delivered

    the ultimate accuracy performance of the overall system. The control points used in this

    test were GPS-measured with an accuracy of ~1.5 cm per coordinate and were located

    about 18 m from the perspective center of the camera.

    SUMMARY AND CONCLUSION

    This paper introduced a concept of an all-digital mapping system designed for

    precise mapping of highway linear features. The test results presented here indicate that

    an integrated, land-based system supported by a medium to high quality strapdown INS

    11

  • 8/14/2019 MODERN MOBILE MAPPING

    12/12

    and dual frequency differential GPS offers the capability for automatic and direct sensor

    orientation of the imaging sensor with high accuracy.

    In addition, the concept of real-time extraction of highwaylinear features such as

    centerlines was demonstrated. The overall system performance was extensively tested by

    a prototype positioning module (for more details see Grejner-Brzezinska and Toth, 1999,

    and 2000), while the feasibility of automated feature extraction was evaluated only by

    simulations.

    12