With the increasingly prominent phenomenon of rocky desertification in various karst regions worldwide, accurate monitoring of rocky desertification is crucial for solving this ecological problem. However, previous studies have relied mostly on comprehensive indicator methods that might exaggerate the contributions of certain indicators and overlook the interactions among evaluation factors. In this study, we established nine different feature space models for rocky desertification, which explicitly presents the interaction relationships between parameters. Then the most effective method for monitoring rocky desertification was then determined through a comparative analysis. The results indicate that: (1) the proposed rocky desertification monitoring model based on a point-to-line bare land index-land surface temperature feature space has good accuracy and applicability, with an overall inversion accuracy of 87.1%; (2) from 2000 to 2020, rocky desertification showed an improvement trend, with the proportion of non-rocky desertification areas increasing by 14.5% and (3) the internal structure of potential rocky desertification is unstable, which should be regarded as a critical object in the process of rocky desertification control. These findings can provide an effective reference for remote sensing monitoring technology of karst rocky desertification and a scientific basis for comprehensive and precise control of regional rocky desertification.