Introduction of Image-Processing in Construction Quality Management & Safety

Introduction

Automation in its numerous shapes is a technique that is likely to overtake construction sites worldwide in the near future. While for decades these sites have been observed, controlled, and monitored by human inspectors, this way of doing has often been described as time-consuming and labour-intensive, thus a possible cause for delays (Soltani et al. 2015). The presence of various inspectors throughout construction processes requires precious time as well as a considerable amount of precision from every one of them in order to ensure a continuous excellence in terms of safety and quality.

While video surveillance has been used in public spaces such as banks for decades, it is only until the beginning of the actual millennium that recorded activity shifted from being human-monitored to computer-monitored (Gelbord and Roelofsen 2002). Almost every public square inch is video-recorded for safety motives. That being said, image-processing becomes even more attractive for construction managers because of that particular reason. As the number of projects grows worldwide every single day, quality is a factor that is and will increasingly be mandatory to manage. In that extent, the positive news is that technological advances make it possible to overcome certain quality issues, such as the previously discussed time control and even the general accuracy of various observations, to mention only two.

The introduction of image-processing using deep learning in construction quality management is part of these new technological possible advances. Although this technique is still very young and therefore inclined to future updates, the actual results look promising. This paper explores the implementation of image-processing as a tool to maintain certain quality management and safety standards directly on construction sites. Since this research topic covers both computer and construction sciences altogether, an initial part discusses image-processing and deep learning while a second one focuses on their different methods of implementation regarding respectively safety, quality management, and synthetic images on construction sites.

Social Concerns

An image-processing aspect to be aware of from the start is the ethical one among workers. In a publication from 2002, privacy and anonymity were concerns among citizens regarding recorded surveillance. More than a decade and a half later, the same questions remain sometimes still difficult to answer. Who has access to the images and where are they processed exactly? This issue is a complex one about how much privacy can be surrendered in the name of public interest, public safety (Gelbord and Roelofsen).

Construction workers are powerful, in the sense that they are the ones that build what is written on plans. Inspectors were a thing, but what if workers find image-processing too aggressive and intrusive? The very last thing that is wanted is a strike caused by ethical disagreements. Labour relation managers must be aware of the implications of such an implementation on construction sites. The same 2002 publication goes on: “At this point, it seems that technology is moving faster than public awareness and is outpacing the public debate” (Gelbord and Roelofsen).

Obviously, image-processing is a management technique that goes far into concerned people’s privacy by continuously framing pictures of them. One of the main worries of this technique then is perhaps the ethical one; meticulous planning is required in order to not violate cultural manners when implementing image-processing using deep learning. Lack of awareness from workers can easily lead to privacy losses, which is very difficult situation to change (Gelbord and Roelofsen 2002). Extreme measures should be taken in order to gain privacy back, and this inevitably implies unstable working conditions. To say the least, ethical apprehensions should be given attention and understood by all parties since it is the subject of public debate and has been for basically two decades.

Conclusion

Image-processing is part of a larger technological spectrum that explores the processing of information in general. If actual technological means are accurately able to process images, chances are that near and far upcoming means will also be able to process images in an unprecedented way, therefore continuously raising the precision standards and accuracy rates. Right now, most processed images are made of movement-detected pixels, which automatically welcomes the risks of error related to random and undesired movements entering the process. Despite interesting updates regarding recognition in real images, the main difficulty lies in the interpretation of the relationships between objects. Action happens among pixels as time advances but then the identification of the precise activity is the main challenge (Bal et al. 2016).

As a matter of thought, an article published in the Advances in Military Technology Journal discusses the combination of template matching and background subtraction techniques to detect objects in infrared video sequences. Astonishing positive results from both stabilized and unstable cameras in this study suggest that combining already known techniques together may be a path that leads to success in terms of image-processing accuracy (Pham et al. 2016). While military and construction processing technologies point towards different goals, it does not prevent the former to guide the latter to the right process so accuracy demands are attained.

Anywhere advances lead the field, there is still room for explorations and invention. Further research should be conducted concerning construction site surveillance. Keeping in mind the previous findings regarding in-motion infrared camera image-processing, what if drones could survey construction sites following a certain path at various moments? This is just an instance of combination that could be worth investigation.

References:

1. Bal L., Li K., Pei J., Jiang S. (2016). “Main objects interaction activity recognition in real images.” Neural Computing & Applications, 27(2), 335-348.
2. Cromey D. W. (2010). “Avoiding Twisted Pixels: Ethical Guidelines for the Appropriate Use and Manipulation of Scientific Digital Images.” Sci Eng Ethics 16, 639–667.
3. Gelbord B., and Roelofsen G. (2002). “New Surveillance Techniques Raise Privacy Concerns.” Communications of the ACM, 45(11), 23-24.
4. Molen H. F. V. D., and Frings-Dresen M. H. W. (2014). “Strategies to reduce safety violations for working from heights in construction companies: study protocol for a randomized controlled trial.” BMC Public Health, 14(1), 1608-1619.
5. Pham I. Q., Jalovecký R., Polášek M. (2016). “Combining Template Matching and Background Subtraction Techniques to Detect Objects in Infrared Video Sequences.” Advances in Military Technology, 11(2), 151-158.