For the PDF version of this article, click here.
Many utilities now use helicopters to conduct transmission line inspections to identify conductor and component damage and corrosion. This condition-based monitoring forms part of asset management strategy. The inspection task is a two-stage operation: video recording followed by video inspection to identify the damage. The inspection is particularly difficult for staff trying to spot conductor damage resulting from a lightning strike (for example, arc marks or cut strands). Research at the Central Research Institute of Electric Power Industry (CRIEPI; Tokyo, Japan) has resulted in the development of a system that automatically detects conductor damage and improves the overall effectiveness of this condition-monitoring procedure.
The Automatic Detection System
The video system is set up and tracked as follows:
- Storage of videos in DV format using an HDD
Figure 1 shows the hardware configuration of the developed system. The system requires taking digital video (DV) from a helicopter, which is then transferred to a hard disk drive (HDD). Following the video storage in the HDD, the process of detecting conductor damage can begin.
It is preferable to input the videos directly into the computer, simultaneously detecting the damage. However, because the developed system is not a real-time process system, damage detection demands all frames be processed. A missing frame is not allowed because this is equivalent to overlooking damage. When video data are stored in the HDD, it uses an IEEE 1394 interface and stores it in a DV format that includes a time code. Using the time code, a viewer can correlate the recorded video with that stored in DV format on the HDD and make a direct comparison.
- Reference point for an initial position on an inspected conductor
During video recording, the position on the cable in the images changes. To check the cable, it is necessary to select it correctly from the image. Figure 2 shows the initial position of the conductor shown as “input” — a single point being specified. The developed system calculates the position of the conductor using three points, the other two being determined automatically.
- Conductor tracking in each image
The position of the conductor is calculated automatically from the initial position based on the similarities among the brightness pattern of the images, which can be determined mathematically.
To extract the contour of the cable, the image area to be used is narrowed (Fig. 3). Generally, when a contour is selected, you find the difference in brightness between two adjacent pixels and define the point where the difference is greater than or equal to the threshold as a part of the contour. By narrowing the area, the calculation time is reduced to a sixth of that required if the whole image was used.
Next, to check whether the conductor contour has an ideal shape, an ideal conductor shape is determined and used when a cut strand or wire is detected. The ideal conductor shape defined by a line of points is calculated by a line-fitting process using least-squares approximation. This fitting process eliminates all points not on the contour.
Detection of a Damaged Conductor
The detection process comprises two parts: The brightness check detects arc marks and the shape check detects cut strands or wires.
- Brightness check
After determining the ideal conductor shape, the average brightness of a conductor image on each vertical pixel is calculated. Figure 4 shows a sample image, the average brightness in a conductor image and the threshold of brightness of an undamaged conductor. Next, the average brightness of the conductor and the standard deviation are calculated. If arc marks are found on the conductor, then the average brightness at this the arc position is beyond the threshold.
- Shape check
When there is a cut wire/strand on the conductor (Fig. 5A), the contour appears like that shown in Fig. 5B. The system searches for this type of contour, and the result is shown in Fig. 5C. The line-fitting process is shown in Fig. 5D from which the ideal conductor shape is calculated. Figure 5E compares the actual and ideal conductor contours.
Performance Check on System Developed
To confirm the performance of the detection system developed, it was applied to 20- to 10-second line inspection videos that contained arc marks and a cut wire. In total, some 6000 images (20 videos × 10 seconds × 30 images/second) were checked.
The detection results given in Table 1 show that 97.8% of the images (5509 + 358) were correctly processed. The results also indicate that 362 images contained damaged conductors, but the system incorrectly misread 129 images as arc marks.
|Result — No Damage||5509||4|
|Result — Damage||129||358|
The developed system is installed on a personal computer, and following the transfer of the videos to the computer, it checks the images for arc marks and cut wires/strands on the transmission line conductor. The arc marks are detected by statistical analysis of the conductor brightness, while the cut wires/strands are detected using conductor shape information.
The statistical analysis uses the mean brightness of the conductor and its standard deviation. In the event the conductor brightness is beyond the threshold range, the system classifies the position as an arc mark. Shape information is obtained from a comparison between a real and ideal cable contour. If the difference between the contours exceeds 10 pixels, the system judges that the conductor has a cut wire/strand.
Because the developed system detected all conductor positions with an arc mark or cut wire/strand, it has proven to be a useful tool for processing the data collected during transmission line aerial inspections. Currently, this development for asset management is on trial at two Japanese utilities.
The authors gratefully acknowledge the contributions of Shinya Okuda and Michihiko Sugimoto of Shikoku Electric Power Co. Inc. for offering their aerial images.
Yuichi Ishino joined the Central Research Institute of Electric Power industry (CRIEPI) in 1991. He researches vision-based monitoring systems for utility equipment.
Dr. Fujio Tsutsumi received a Ph.D. in information science and electrical engineering (ISEE) from Kyushu University, Japan, in 2003, and he joined CRIEPI in 1990. He researches visualization for large volumes of data and human interface.