Condition assessment of sewers - is digital image editing and processing becoming standard in CCTV inspection of sewers?

Within the scope of maintenance of drainage systems, condition assessment plays a major role. Its task is to provide data and information on the system's actual condition and to detect defects and their causes at an early stage in order to maintain a drainage system that functions in all its components in the planned service life with the lowest effort for maintenance and rehabilitation [1].

Regarding the minimum requirements by the legislator optical inspection represents the standard when assessing the constructional condition of drains and sewers.

During an inspection the condition of the system as well as the type and extent of defects must be thoroughly observed and must be documented according to ATV-M 143-2 [2], DIN EN 13508-2 [3], for example, or other coding or abbreviation systems. Regulations on coding / abbreviation systems that are not compliant with DIN EN 13508-2 "Visual inspection coding system" [3] should have been withdrawn by May 2006 [4].

For the internal optical inspection of non-accessible sewers nowadays nearly exclusively two different versions of CCTV are used:

In the following a summary of the results of a study that tested the performance range and the fields of application of this new camera system will be given. The study was carried out by the engineering consultant Prof. Dr.-Ing. Stein & Partner (Bochum, Germany) by order of the manufacturer RICO GmbH (Kempten, Germany).

Description of the hybrid camera

Unlike conventional CCTV inspection cameras, the RPP^®DuoVision Hybrid Camera, in the following referred to as hybrid camera, possesses a camera head with two object lenses. Besides the usual zoom lens, it also has a 180° fish-eye (Fig. 2).

On the journey through the sewer - starting from the starting manhole - the condition is assessed in a conventional procedure, i.e. according to the stop-and-go principle possibly available inlets/branches, pipe joints/sockets, defects, etc. are recorded by the operator and saved in the inspection vehicle. By means of the camera pan shot or rotation of the camera head and the 10x zoom function also details are lighted and recorded if required.

The network operator and/or the client usually receives these pictures as a file on a CD-ROM and they are visualized by the "KEP viewer". Additionally, it contains other information such as basic and master data of the inspected sewer section as well operating functions.

Figure 3 shows the viewer with the selected menu point "videos" as an example of the exclusively perspective view of the sewer. It offers an enlarged view (complete picture) of the perspective internal view of the sewer in the menu option "fold-out" (cf. Fig. 4).
Activating the button "fold-out" provides the most detailed view (Fig. 4). The monitor image shows the perspective internal view of the sewer (centre) and the corresponding, synchronized fold-out (bottom). As an overview the fold-out of the complete section is located in the top right corner on the monitor. Measuring of defects and objects is only possible on the digitally enlarged lower fold-out (max. 4x digital zoom) exclusively.

The system is completed by the automatic swivelling of pipe joints by the 10x zoom lens that rotates around the own axis by 360°. Besides a fold-out in the BMP-format, on which distance and area measurements are possible by mouse click, the automatic image recognition ("socket mosaicing"), which is connected to the so-called "socket radial scan", creates so-called socket statistics, which provide important information on the width of the socket gap and possibly its development ("socket gap history") (Fig. 5).

Optionally, the operator can define whether the radial scan is saved as a digital, fold-out image only (Fig. 5), as a mere film in perspective view or in both versions. For the purpose of an holistic inspection with maximum informative contents the latter variant with both presentation modes is recommended.

According to information by the manufacturer [10], the hybrid camera is used in nominal size ranges of DN/ID 200 to 1400 mm for circular cross-sections and corresponding size ranges of other cross-section shapes such as egg-shaped, arch or rectangular profiles.

Inspections

On 8 September 2005 in-situ inspections were performed in a residential/industrial street situated in the city of Kempten/Allgaeu. Here, by means of a hybrid camera that was installed inside an inspection vehicle for demonstrative purpose (Fig. 6) a vitrified clay sewer DN/ID 500 mm with a section length of 58.77 m as well as a concrete sewer DN/ID 600 mm with a section length of 59.70 m were inspected. A repeated accessing for testing purposes of a concrete sewer DN/ID 500 mm with a section length of 57.01 m was carried out in Münster, Germany by the contractor Koch + Geist GmbH on 4 May 2006. This had additionally become necessary to particularly test this function once again after RICO had eliminated an error in the automatic socket gap measurement "socket radial scan", which had been discovered by S &P.

The study focussed on the assessment of condition patterns (patterns of defects, system parts and phenomena) of the inspected sewers – recorded by the hybrid camera – regarding their image quality and sharpness and their informative content. Here, particular emphasis was put on the measuring possibilities in the fold-out view and their accuracy or tolerance.
Feasibility studies were not subject of the study.

Results of the study

In the following paragraphs the recording of a “pipe joint” by using “socket radial scan” as well as analyses and measuring functions by means of the condition image “crack” are presented as an example.

Pipe joint

Figure 7: Figures 7a to c show the view of the condition pattern "pipe joint" in a concrete sewer DN/ID 500 mm (radial scan of the socket as a digitally enlarged view shown as extracts from the viewer, cf. also Fig. 5)

Figure 7 gives a detailed view of a pipe joint in a concrete sewer DN/ID 500, which was recorded by the special hybrid camera function "socket radial scan" and is presented in the viewer in the menu option "radial scan" (cf. also Fig. 5) as a fold-out. By means of the measuring function on the fold-out a manual measuring was carried out via reticules. In this way the socket statistics, which are created during the "socket radial scan" and which are shown in Figure 5, should be analyzed in order to gain information on the width of the pipe joint and/or the joint gap as well as the condition of the pipe joint as regards defects, leakage, etc.

The automatic measurement of the socket gap delivered the following values for the socket statistics (see box "Radial Scan Info" in Fig. 5): mean (5 mm), standard deviation (5 mm), minimum value (0 mm), maximum value (11 m) and detection (32 % with reference to the circumference). The widths of the socket gaps that were measured on the digitally enlarged parts of the fold-outs of the pipe joints (Fig. 7a and b) did not deliver a value of 8.57 mm for the left springing and/or 3.06 mm for the crown and thus lie within the tolerance range given by the socket statistics (see box "Radial Scan Info" in Fig. 5). In the given case these tolerances could be put down to the structure of the concrete pipes in the pipe joint (burrs in the pipe faces), tolerances due to installation and manufacturing (deflection, straightness of the pipe faces). This is why in the "socket mosaicing" only 32% of the pipe joint were recognized as a socket gap by means of automatic image recognition; from these values a mean value for the width of 5 mm was calculated (see box "Radial Scan Info" in Fig. 5). A value that - considering the described boundary conditions - can be regarded as realistic.

However, it should be noted that in this case the expressiveness of the socket statistics is rather limited. Thus, the area on the invert, which is possibly subject to deposits or covered with sewage, cannot be scanned successfully due to lacking colour contrasts (automatic "light-dark fader") in the socket gap. In addition, cross-fading of the image, which particularly occurs in the wet region of the invert (Fig. 7c), leads to the above mentioned error consequences and the low detection level in the digital image.

Furthermore, the use of the function "socket radial scan" requires  previous training of the operator, since especially the manually adapted lighting conditions in the sewer are decisive for a good quality of the images taken of the pipe joints. If, for instance, the lightness is too high due to cross-fading and/or reflecting surfaces on the pipe wall, errors occur in the digital image as a consequence of this cross-fading as these areas are recognized and scanned as "white". To prevent them, there is an additional function in the sewer assessment software available to the operator. This function defines the scan area before activating the socket pan shot.

Longitudinal crack

Figure 8 shows an example of the condition pattern “crack formation” and/or “crack” in a concrete sewer DN/ID 600 mm (extracts from the viewer). The fold-out (Fig. 8 b) immediately reveals the longitudinal crack in the crown and the lime deposits and calc-sinters entering through it due to the colour contrast. In the right springing and/or the lower third of the fold-out in Fig. 8 b further crack formations can be detected, too. They are, however, more difficult to recognize because of the lacking white colouring and as a consequence could wrongly be regarded as scratches. The cracks in the left region of the sewer’s springing will not be dealt with in detail in the following.

A closer look with a pan shot (Fig. 8) allows to inspect the crack in the crown more closely. It is long and of only a small width and appears much larger than it actually is due to the partially strong white calc-sinters along the edges of the crack. The length of the crack can be measured relatively simple on the fold-out (Fig. 8 b). In the given case a crack length of 1505.93 mm was measured.

As it is too small the width of the longitudinal crack in the crown area of the sewer, on the other hand, not be gripped and determined  with the reticules (only the reticules’ line width is far greater in this case). Fig. 8 c chosen view of the fold-out after activating the 4x digital zoom, which already shows a high level of graining with this zoom factor. Correspondingly, with both reticules the actual crack width cannot be defined exactly. In this picture, which allows a positioning of the reticules in this enlarged version, the crack width was calculated with 3.68 mm; a value, which in the face of this crack characteristic appears too high for the experienced observer in the perspective view.

 

Measurement accuracy on the fold-out

The measurement accuracy on the fold-out (see Fig. 4 bottom) predominantly depends on the resolving capacity of the camera and the subsequent image digitalization by the analogue-digital converter (quantization error) [10]. For creating the fold-out an annulus is gripped from the digitalized image; its resolution can be adjusted and on the hybrid camera is set to 512 pixel as a standard over the entire circumference [10].
The achievable resolution (size of the displayable pixels) over the circumference of the inner pipe wall and thus on the fold-out, too, for this reason depends on the given nominal pipe size DN/ID (clear inner diameter). The resolution of the fold-out can be calculated according to: (DN/ID x p) / 512 [mm] and correspondingly decreases with growing nominal size. For DN/ID 500 mm the maximally possible resolution (pixel size), for instance, is 3.07 mm (cf. Table 1, columns 1 and 2) [10].

Due to the fact that in the given case a pixel represents the smallest visible and thus also measurable unit on the digitalized fold-out (Fig. 9), the image resolution (and thus the local gripping rate) as well as the size of the information saved on the pixel (such as the colour depth) in digitalized images is limited, whereas one pixel is only an approximation of reality. This limitation of resolution (cf. Table 1, column 2) leads to image information being lost.

It can be taken from Fig. 9 that objects and/or defects (especially crack widths) whose dimensions are smaller than a pixel, can only be displayed “blurry”, i.e. with blurred outlines and/or lacking contrast in the areas of pixel transition, on the digitally enlarged fold-out.
As consequence a measurement of very small distances does not make sense in this range as the value that was measured and/or gripped with the reticules is likely not to reflect the actual conditions and can thus cause high measurement tolerances and/or errors.

This effect will be exemplified on the crack shown in Fig. 8 and 9:

Here it is started from the assumption that the actual width of the crack is 2.00 mm at a specific point that is looked at in detail on the concrete pipe DN/ID 600 mm. The dimensioning of a square pixel was calculated mathematically by 3.68 x 3.68 mm (see Table 1). Now the case is assumed that at this point the thin crack in the digitalized image lies in the range of a pixel's width. Due to this also assumed black-and-white contrast (line/crack is to be featured white, background/pipe wall is assumed to be black) in the following the crack is displayed with exactly one pixel-width with bright contrast and is measured correspondingly with 3.68 mm. In a second case it is hypothesized that the crack is situated directly in the transition area of two pixels, i.e. both pixels each would correspondingly display the crack over the width and/or height with a darker white and/or bright gray. Now the crack would be measured 2x 3.68 mm = 7.36 mm. With reference to the width of a pixel the absolute measuring error therefore is +3.68 mm.

If now the crack width was assumed 5.00 mm, i.e. larger than a pixel, for example, the absolute measuring error would correspondingly be 3 pixel – 2 pixel = 1 pixel, i.e. 3.68 mm.   
With reference to the actual value of 2.00 mm and/or 5.00 mm this means a relative measuring error of 3.6 8mm/2.00 mm = 1.84 or 184 % and/or for the 2^nd value 3.68 mm/5 mm = 0.74 or 74 %.       
Especially from the first, very high value it becomes clear that the precision of the measurement might not lead to any reasonable or acceptable result with very small dimensions, and definitely not if the width of the scratch or a crack is even smaller than assumed above.

In contrast the measurement of crack lengths as regards the accuracy does not represent a problem. If the crack on the fold-out was measured with a length of 1506 mm (Fig. 8 b), the relative measuring error, on the other hand, would only be 3.68 mm/1506 mm = 0.0024 or 0.24 % and would thus lie in an acceptable range.

This example shows that, when using the measuring function on the fold-out, also the uncertainty and/or the accuracy of the measuring must always be scrutinized critically. This applies, as shown above, especially for very small dimensions, especially crack widths which are smaller than the achievable resolution and/or displayable pixel size. Here in terms of nominal size the dimensioning is confronted with limits of accuracy that are no longer tolerable.         
In order to assist the user, especially with the determination of the crack width, in which case the measuring function on the digitalized fold-out of the hybrid camera starts to make sense and can be used with sufficient  accuracy, S & P developed the 3^rd column in Table 1. In the determination of the dimensioning of the minimum object/defect with reference to the nominal size it was assumed that a relative measuring error of 20% especially for measuring the crack width would lie in a tolerable range and would thus lead to an acceptable measuring result.

With an extreme zoom the above described phenomenon of “graining” can be noticed on each digitalized image with reference to the given resolution and chosen zoom factor.

Conclusion: With the given nominal pipe size of DN/ID 600 mm the value of 3.68 mm (Fig. 8 c) that was ‘measured’ on the fold-out with 4x digital zoom mathematically corresponds exactly to the width of one pixel (cf. Table 1, column 2), i.e. the maximally achievable resolution, and thus does not represent the smallest visible and/or measurable unit (precision limit). In this case the measuring function on the fold-out does not lead very far as great inaccuracies in the values must be taken into account due to the loss of information caused by the digitalization of the fold-out and the effects resulting from it (cf. Table 1, column 3).       
In the given case the crack width could only be estimated with 1 mm based on the images in perspective view that were recorded with the 10x zoom lens.

Summary

As a summary it can be said that for the first time the hybrid camera combines the strengths of a conventional pan and tilt camera – with 10x optical zoom for the perspective detailed view in a high image quality and resolution – with the advantages of the independent, digitally editable and processable images of a fish-eye in the form of a fold-out. Generally, the advantages of editing and processing of digital images are that in the case of the fold-out, for instance, a complete 2D-view of the sewer is created out of an originally 3D-image. This view additionally allows measurements over the single pixels.

This technology however also bears limits where the given results should be critically questioned especially in terms of their measuring precision. Furthermore, by signal conversion and data compression artefacts can develop.

As far as the reproducibility of data, more precise quantification of defects (especially crack widths), archiving, plausibility checks after multiple inspections, substance predictions, rehabilitation planning, etc. are concerned, digital imaging has advantages that conventional camera systems cannot offer alone.
It turns out that digital image editing and processing in CCTV inspections of drains and sewers can no longer be held up and can already considered standard technology. It reflects the general trend of time and the manufacturers’ eagerness for technical progress. On the long run the ‘old’ systems will either be combined with this new technology as in the case of the hybrid camera, or will be completely replaced by fully digital inspection systems. Only the users and their interest and/or demands on high-quality and expressive inspection data will decide at which intensity and speed this development process will enter condition assessment of sewers.

In spite of the growing ‘digitalization’ of condition assessment of drains and sewers, the operator (inspector) is still an important influencing factor to achieve the best possible image quality and expressiveness of the inspection data through manual fine adjustments (e.g. brightness) with reference to the conditions although there are often a lot of automatisms.

Bibliography

[1] Stein, D.: Rehabilitation and Maintenance of Drains and Sewers. 3rd edition, Ernst & Sohn, Berlin 1998.
[2] ATV-M 143: Part 1: Inspektion, Instandsetzung, Sanierung und Erneuerung von Entwässerungskanälen und -leitungen - Grundlagen (12.1989). Part 2: Inspektion, Instandsetzung, Sanierung und Erneuerung von Abwasserkanälen und -leitungen - Optische Inspektion (04.1999).
[3] DIN EN 13508: Condition of drain and sewer systems outside buildings. Part 1: General requirements (02.2004). Part 2: Visual inspection coding system (09.2003).
[4] Keding, M.: Zustandsbeurteilung von Entwässerungssystemen nach DWA-M 149. KA - Abwasser, Abfall 52 (2005), No 6, pp 719–724.
[5] Stein, D., Körkemeyer, K.: Entwicklungen bei der TV-Inspektion von Abwasserkanälen. Contribution to 36. Essener Tagung of 26 to 28 March 2003 in Aachen.
[6] Stein, R.: Trends und Entwicklungen der Zustandserfassung. bi umweltbau (2004), Issue 3, pp 64–72.
[7] Hunger, W.: Optische Kanalinspektion mit Panoramo. Contribution 12th European Symposium on Water, Wastewater and Refuse "TV-Inspektion für Betrieb und Instandhaltung (EWA)" on 14 May 2002 in connection with IFAT 2002 in Munich.
[8] Stein, D., Körkemeyer, K., Brauer, A.: Vergleichende Analyse des neuartigen PANORAMO-Inspektionssystems mit den Standardverfahren zur Inspektion von Abwasserleitungen und -kanälen am Beispiel des ARGUS 4-Kamerasystems. Expertise by Prof. Dr.-Ing. Stein & Partner GmbH, Bochum by order of IBAK Helmut Hunger GmbH & Co. KG, Kiel. Bochum, May 2004 (published on www.unitracc.de, for viewing free registration necessary).
[9] Stein, D., Brauer, A., Broziewski, A.: Optische Zustandserfassung von Kanalisationen - volldigital oder analog? KA - Abwasser, Abfall 2005 (52), Issue 3, pp 259–268.
[10] Company information RICO Gesellschaft für Mikroelektronik mbH, Kempten.
[11] Hinn, A. K.: Die industrielle Bildauswertung und -verarbeitung im Kanal. In: Rohrleitungen - Für eine sich wandelnde Gesellschaft. Schriftenreihe aus dem Institut für Rohrleitungsbau Oldenburg, Vol 30, pp 702-706. Vulkan-Verlag, Essen 2006.
[12] Oxford Advanced Learner's Dictionary of Current English. 5^th edition. Oxford UP, 1995.
[13] Stein, D., Stein, R., Brauer, A.: Untersuchung und Beurteilung des Leistungsspektrums sowie der Einsatzbereiche der RPP^®DuoVision-Hybridkamera. Expertise by Prof. Dr.-Ing. Stein & Partner GmbH, Bochum by order of RICO Gesellschaft für Mikroelektronik mbH, Kempten. Bochum, May 2006.

More News and Articles