Photometric Stereo Surface Imaging

High resolution crack detection with millimeter precision

Our patented Photometric stereo image system produces a full ultra-high-resolution digital map of the road surface.

Using a line-scan camera and four high powered lights the system captures four images of the road with illumination from the left, right front and back.

This system has the ability to detect the gradient contour for each square millimeter of the ground producing forward angle, cross angle and luminance images resulting in a wealth of information useful for both network assessment as well as detailed project based defect assessments.

Photometric Stereo Surface Imaging

The Photometric stereo image system incorporates a novel and patented technique for producing a full ultra-high-resolution digital map of the road surface. The system incorporates a line-scan camera with four bars of high power artificial lights. These lights are 20 times brighter than sunlight for their spectrum of operation, thus allowing 4.0 meter scanning without needing a sunlight shield. These lights are flashed in sequence for every millimeter of travel (60,000 times per second at 100km/h). The result is a set of four images of the road with illumination from the front, right, back and left allowing the surface structure can be separated from surface colour. Using these four images the system has the ability to detect the gradient contour for each square millimeter of the ground: producing forward angle, cross angle and luminance images.

The purpose of this detailed imaging of the surface, is to separate the "colour" or light and dark shades of the road surface, from the structure of the road surface. Cracking or other surface defects are primarily structure elements, while marks on the road surface oil drips, water from air conditions and the varying colours of the rocks that comprise the roads surface have no consequence on bearing capacity or road pavement quality. The analysis can be considered to be similar to painting the road surface an even grey, and then taking an image of the surface. If this is done, the cracks show up very clearly. This separation of luminance and structure also allows further measures of the road surface. For example the pumping of fines (indicating base distress) can still be seen in the luminance image, and linked to the crack detection and crack width measures in the structure domains. The result is that for each crack detected, not only is the width determined but also a number of other metrics including percentage pumping. The result is a wealth of information useful for both network assessment as well as detailed project based defect assessments.

The other advantage of the Photometric imaging is the vast reduction of false positives when performing crack detection. Again this is due to the detection of cracking in the structure domains. It is even possible for the system to detect cracking when covered in dirt or debris, as well as the precursors to cracking such as the formation of waves and smaller plates in the roads surface, which is especially true for thinner asphalt and chip seal surfaces.

Example #1

Normal detailed photographic image of the road surface, apart from some pumping of files, there is no clear cracking in the image as a normal surface image tends to flatten any structure, making the cracks disappear.

The same scene in the "forward angle" structure domain. Cracking becomes much clearer, making these features more obvious for both manual crack mapping and automated crack mapping algorithms.

Cross angle (transverse angle) allows the increased detection of cracking in the direction of travel.

Example #2

Asset camera image of a road pavement: this is a typical example, with shadows from trees making crack detection difficult. There is however a clear change of construction.

The next asset camera frame.

Photometric "luminance" image of the road surface. While the image is more detailed, allowing some cracking to be seen, the true cracking in the surface is still not obvious.

Photometric "angle" image of the pavement surface. Red is forward angle while blue is cross angle. In this domain, the cracking becomes pronounced, with all other features being eliminated. The result is a clearer detection of cracking.

Finally, the cracks can be detected and classified. Crack width is also automatically determined.

A word on Data storage

Due to incredibly high resolution of the system, storing the Photometric images for a complete network requires approximately 1.2 GBytes/km (depending on compression ratio). Also when combined with level of detail control, this can further increase to 1.6 GBytes/km. While initially this appears to be quite difficult to store the price for storage is rapidly reducing, to a point where this type of storage is quite acceptable for most road networks.

Some surface imaging systems choose to perform real-time processing of the imagery data into crack maps, which allows raw data to be discarded. While this produces instantaneous crack statistics for the network, in the long run it imposes a number of limits on the technology. At RPS we believe in raw image storage, which is then linked to systems that allow easy data management and rapid data turnaround. The result is the best of both worlds: data that can be delivered to customers rapidly, while having the flexibility to produce accurate, qualified and custom focused results.

Return to top