Remote Sensing Based Integrity Assessment of Infrastructure Corridors Using Spectral Anomaly Detection and Material Degradation Signatures

Abstract

This study developed and empirically evaluated a quantitative remote sensing–based framework for integrity assessment of infrastructure corridors by integrating spectral anomaly detection, material degradation signature engineering, multisensor fusion, and time-series change analytics. Corridor integrity was operationalized as a continuous, segment-level outcome derived from multispectral, thermal, and synthetic aperture radar observations constrained to corridor footprints. The analytical dataset consisted of 480 corridor segments, retained after quality screening from an initial extraction of 520 segments, and stratified across land-cover adjacency, terrain class, and exposure conditions. Spatially blocked partitions were applied to reduce spatial dependence bias, producing 336 development segments and 144 testing segments with comparable contextual and condition distributions. Descriptive analysis showed moderate right-skew in spectral anomaly intensity (mean = 0.42, SD = 0.21) and wider dispersion in degradation signatures (mean = 0.55, SD = 0.27), while fused integrity scores exhibited reduced variance (mean = 0.48, SD = 0.18), indicating stabilization through multisensor integration. Reliability analysis confirmed strong internal consistency for composite constructs, with Cronbach’s alpha values ranging from 0.79 to 0.88 across exposure strata. Multiple regression modeling demonstrated substantial explanatory power, with the final model achieving R² = 0.62 and remaining statistically significant (F = 153.4, p < 0.001). Material degradation signatures showed the strongest standardized effect on corridor integrity (β = 0.42, p < 0.001), followed by spectral anomaly intensity (β = 0.31, p < 0.001) and temporal change constructs (β = 0.19, p < 0.001), while contextual controls exhibited smaller but significant contributions. Diagnostic checks indicated low multicollinearity (maximum VIF = 2.14) and minimal residual autocorrelation after spatial blocking (Durbin–Watson = 1.96). Hypothesis testing results remained stable under conservative multiple-testing adjustment, confirming robustness of inferential conclusions. Overall, the findings demonstrated that physically interpretable, multisensor-derived indicators can be transformed into reliable and statistically defensible measures of corridor integrity, supporting quantitative monitoring across heterogeneous and spatially extensive infrastructure systems.