2024 Volume 80 Issue 9 Article ID: 23-00158
We propose an image recognition method that automatically measures the deterioration size from photographed images by a digital camera to improve the efficiency of inspections of communication tunnels used to accommodate communication cables. The proposed method detects equipment and deterioration areas from photographed images using a deep learning, and estimates pixel resolution (mm/pixel) from the detection results of metal rack areas and size information. The actual deterioration size is measured from the pixel resolution and the detection result of the deterioration area. We evaluated the detection results of the proposed method using 1, 700 images, and confirmed that the proposed method can detect with an F-measure of 0.92 in the metal rack areas and 0.84 in the exposed rebar area. We evaluated the measured values by the proposed method and the measured values by the measuring tape using 91 exposed rebars, and found that the correlation coefficient was high at 0.9773. The measured performance of the proposed method are sufficient to determine the scale of deterioration and can be applied to the inspection of communication tunnels.
