4 CURRENT APPROACHES TO IMPROVE VISUAL INSPECTIONS
4.3.6 Potential for using technologies in routine visual bridge inspections
There are similarities between some archaeological sites, and some bridges, indeed some of the heritage sites are bridges (Lubowiecka, et al., 2009). However, there are also some significant differences in the purposes of the data collection, and the limitations placed on the collection of the data. The collection of data on archaeological or cultural sites is not a safety critical task, and is not usually used to plan/programme further work. The purpose of the data is primarily to record the condition of the site or artefact of interest as accurately as possible to enable work to continue offsite or once site is destroyed/inaccessible (Allen, et al., 2004).
The consultation responses (discussed in Section 2.3) indicated that engineers would want to be able to detect minimum defect widths of 0.2mm to 1mm if the inspection was to be successful. A datapoint spacing of 1 point per mm would result in 100 datapoints in a square cm, or 1 million datapoints in a square m. The typical datasets produced for archaeological applications do not produce data at sufficiently high resolutions. For example, Haddad (2011) states that 3 million data points were collected on a series of sandstone sculptures, with a surface area of 40m2 (Haddad,
2011). This is approximately 1 datapoint every 13mm – good enough to give a visual representation of the appearance and shape of the sculpture, but not good enough to detect or display fine details or cracks. The data can be used for a variety of purposes, and can provide a lasting record of the site. There are many similarities between the needs of the archaeological surveys and those of bridge inspectors, but the data available from the systems used in archaeology does not meet the needs of the engineers.
There are a number of issues related to the collection of accident scene data which make it somewhat of a special case and have affected the development of the data collection systems. The scenes of interest can be many tens of metres long, but are
often only a few metres wide, and most of the information of interest is confined to a narrow vertical band of only a couple of metres height. The techniques used are fast and accurate, but are also expensive and difficult to use for non-specialist staff (Fraser, et al., 2005). The data does not need to be photorealistic as it is parameters such as the relative positions of vehicles and obstructions, road geometry and sightlines which are of primary interest, not the presence or appearance of fine details. Consequently, although image data is often collected along with the laser data this is not collected or presented at resolutions adequate for detecting fine details which would be looked for in a routine visual bridge inspection.
Images collected on construction sites are used to enable visualisations of progress, or to provide condition records. The data is not routinely used to identify fine details and cracks such as those that would be of interest to engineers involved in bridge maintenance, and consequently high resolution is not required.
The methods used for pavement condition monitoring are not directly applicable to those which would be needed to perform routine bridge inspections, although there are some similarities and lessons which could be learned. The main difference is that the pavement is always (unless something has gone catastrophically wrong with the survey) in the same place relative to the vehicle: the survey vehicle drives along the pavement, and collects data on the part of the pavement it is on at any time. The pavement is also, essentially a two dimensional ribbon (for the purposes of these condition monitoring surveys it is only the surface of the pavement which is considered). Bridges, on the other hand, are three-dimensional structures, and can be of many varied sizes and shapes. This makes it much harder to collect data at the desired resolution covering all parts of interest on the structure without developing a tailored approach for each structure.
The pavement condition monitoring techniques have had good success in monitoring certain surface condition parameters, but less success on visual defects such as cracking and fretting. It is precisely these types of defects which would have to be detected in routine visual bridge inspections.
Bridge engineers have indicated that image data is already used when considering inspection reports, and that the images are very useful. The collection of inspection images is currently done non-systematically, and is only as methodical as the inspector taking the images. No inspectors interviewed as part of this research routinely collect a full image set over all visible parts of a bridge. This leaves some areas of the bridge unimaged. For bridge elements where no image exists, and with no notes in the inspection report, it must be assumed that the inspector inspected the element and found nothing to report, but there is no way of knowing that the inspector did not overlook that area of the bridge, or miss a defect which was present.
The use of images in GIs suggests that there is no objection in principle to using images to determine the condition of a bridge. Additionally, the adoption of advanced technological approaches such as GPR or acoustic monitoring shows that bridge inspectors and engineers are open to the idea of using new techniques and tools. Problems with visual inspection data have been addressed elsewhere successfully by using technology, but there are sufficient differences between bridge inspection and these other applications to make their solutions not directly suitable for Image-Based Inspection (IBI) of bridges. Therefore the questions are, could inspections be performed to an acceptable standard using images; and if so, how could these images be recorded, processed and analysed; and how would such inspections fit into the existing inspection regime?
An examination of the goals and procedures of the five types of inspection on UK highway structures would suggest that an image-based system could perform
surveys which were between General and Principal Inspections in their scope and coverage. The development and adoption of Image-Based Inspections (IBI) could address many of the problems with visual inspections, produce a tool for defect progress tracking, and be a necessary first step towards the use of automated image processing and analysis techniques.
Images must be collected in such a way as to provide full, detailed coverage of all visible elements of the structure, providing no less detail than can currently be obtained when performing a GI with no artificial aids. The images would have to be accurately locatable on the structure so that the presence, severity, type and extent of any defect could be accurately recorded.
In order to produce such a system there are a number of areas which need to be investigated, such as appropriate ways of collecting, processing, analysing, interpreting and reporting the data. Successfully overcoming these problems could result in inspection data which was not as reliant as the current regime on the opinion and performance of a single on-site inspector, was easier to share and discuss, provided a complete visual record for tracking and monitoring changes over time and which could be suitable for automated image-processing, opening up the possibilities of further advances in inspection data processing, analysis and reporting.