25-01-2013, 12:46 PM
Automated Defect Recognition Method by Using Digital Image Processing
1Automated Defect Recognition.pdf (Size: 232.08 KB / Downloads: 62)
ABSTRACT
As existing infrastructure systems are aged and deteriorated rapidly, state agencies started searching for more advanced ways to maintain their valuable assets to the acceptable level. One of them is the application of digital image processing. Recently, in the civil engineering domain, digital image processing methods have been developed to the areas of pavement conditions, underground pipeline inspection, and steel bridge coating assessment. The main reasons to count on the advanced technology are due to such advantages as accuracy, objectivity, speed, and consistency. These distinct advantages have brought attention to state agencies to minimize the shortcomings of existing inspection practices. This paper deals with a digital image processing method to apply it to the evaluation of steel bridge coating conditions. Infrastructure condition assessment can be made more accurately and quickly with the aid of computerized processing system. The proposed method in this paper was designed to recognize the existence of bridge coating rust defects. It was developed by making pair-wise comparisons between a defective group and a non-defective group and generating eigenvalues to separate two groups. An automated defect recognition method can make a decision whether a given digitized image contains defects.
Introduction
As existing infrastructure systems are aged and deteriorated rapidly, state agencies started searching for more advanced ways to maintain their valuable assets to the acceptable level. One of them is the application of digital image processing. Recently, in the civil engineering domain, digital image processing methods have been developed to the areas of pavement conditions, underground pipeline inspection, and steel bridge coating assessment (H. D. Cheng et al. 1999, S. K. Sinha et al. 2003, S. Lee et al. 2005). The main reasons to count on the advanced technology are due to such advantages as accuracy, objectivity, speed, and consistency. These distinct advantages have brought attention to state agencies to minimize the shortcomings of existing inspection practices.
The conditions of steel bridge painting surfaces can be evaluated accurately and quickly by applying digital image processing. Also, machine vision-dependent inspections can provide more consistent inspection results than human visual inspections. Because conventional inspection heavily relies on individual abilities, inspection results are error-prone and may have wide variations between inspectors. The results can be different depending on personal preferences, work experiences, and the workload of the inspectors. It is pretty important to develop reliable infrastructure condition assessment for better maintenance of the assets. In case of bridge coating, bridge managers can more realistically develop long-term cost-effective maintenance programs if they have dependable coating condition data. Also, they can make decisions as to whether a bridge shall be painted again immediately or later. Efficient coating condition assessment is also essential for the successful implementation of steel bridge coating warranty contracting. Under the warranty contracting, an owner and a contractor inspect steel bridge coating conditions on a regular basis and decide whether additional maintenance actions are needed. However, it is extremely difficult to determine if a bridge contains more defects than an allowable level. If they are in conflict, they will go through a lengthy process to reach an agreement.
Deteriorated Bridge Infrastructure Conditions
The report cards published by American Society of Civil Engineers (ASCE) are important indicators to understand current conditions of major civil infrastructure systems in this country. A wide range of civil facilities are included for evaluation such as bridges, roads, dams, schools, transit, energy, and so on, and graded on an A to F grading scale (ASCE report card 2009). Lots of leading civil engineers are involved in preparing the report card and the analysis of reports, studies, and other sources are performed. Unfortunately, the overall American infrastructure received a failing grade since the beginning of the studies. The overall grades in 1998, 2001, 2005, and 2009 are D, D+, D, and D, respectively. Also, the report cards indicate the estimated investment needs for the next 5 years to recover infrastructure systems to the acceptable level. The dollar amount needed has been increased. The 2001 ASCE study indicated that the estimated cost for infrastructure renewal was $1.3 trillion dollars, $260 billion annually. But, the 2005 study addressed that the renewal cost was $1.6 trillion, and the 2009 study recorded the highest point, $2.2 trillion dollars. As existing infrastructure systems are aged rapidly, more and more investment funding becomes necessary to eliminate deficiencies. However, the available funding amount is much less than required and is typically limited. Thus, it is very important to set up an efficient management plan on how to consume limited resources each year. In case of bridges, the GPAs in 1998, 2001, 2005, and 2009 are C-, C, C, and C, respectively (see Figure 1). The points are a little bit higher than the other infrastructure facilities, but still are not satisfactory.
Data Analysis & Results
In this stage, bridge coating images are processed to generate eigenvalues. Two kinds of pair-wise comparisons were performed: two non-defective images and a non-defective image and a defective image. Total 105 data points were obtained from the comparison of two non-defective images (Group A). Also, total 225 data points were achieved from the comparison of a non-defective image and a defective image (Group B). Figure 4 shows the gray-level distribution of Group A and Table 1 presents descriptive statistics based on the figure. Five values (minimum, maximum, average, standard deviation, and variance) were calculated to a small eigenvalue and a large eigenvalue.
Conclusions and Limitations
This paper presented a novel approach to recognize the existence of bridge coating rust defects by utilizing a digital image processing to better assess a bridge coating surface. The image defect recognition method was developed by making pair-wise comparisons and calculating eigenvalues which were chosen as a key feature to distinguish defective images from non-defective images.
The rust defect recognition method was realized by taking the following three stages: image acquisition, image processing, and data analysis. In the image acquisition stage, bridge painting digital images were acquired and prepared to generate two types of data sets: defective and non-defective. In the image processing stage, a pair-wise comparison was performed to generate eigenvalues. The first comparison was performed between two different non-defective images where total 105 data points were generated. And, the next comparison was carried out between a defect image and a non-defective image where total 225 data points were generated. Large and small eigenvalues were generated and distributed on a two-dimensional distribution map. Also, five statistical values were calculated and presented in tables. The results from this experimental study were summarized in details in the above discussion section. Experimental results demonstrated that an eigenvalue-based defect recognition method is effective to distinguish defective images from non-defective images.