Study on image correction of misaligned line-scan imaging for tunnel disease detection

Efficient detection of tunnel surface defects is critical for reducing maintenance costs and preventing accidents. Line-scan cameras have emerged a pivotal tool for tunnel image acquisition, owing to their high resolution and continuous scanning capability. However, spatial constraints often necessitate misaligned installation of the cameras, which may introduce compound image distortions and affect defect recognition accuracy. Existing correction methods typically rely on calibration targets and accurate camera poses or address only a single type of distortion, limiting their applicability in complex tunnel environments. To address these limitations, a misaligned line-scan imaging model based on tunnel geometric priors was established to characterize the compound distortions caused by camera pose deviations. A progressive correction method was then developed, in which orientation correction, tilt correction, and orthorectification were performed in sequence to rectify the distortions stepwise. Tunnel structural features were used for automatic correction without requiring accurate camera poses. Engineering tests showed that the mean relative error of calibration-grid measurements decreased from 43.5% to 8.6%, demonstrating that the compound distortions caused by misaligned installation were effectively corrected.