The　problem　of　hollowing　threatens　the　integrity　of　cultural　relics.　Infrared　thermography　provides　a　basis　for　identifying　hollows　but　struggles　to　accurately　determine　their　boundaries.　Existing　methods　predominantly　rely　on　conventional　thermographic　parameters　and　lack　systematic　studies　on　the　extraction　and　analysis　of　thermally　sensitive　features　of　hollows　under　complex　field　conditions.　In　particular,　research　on　the　precise　detection　of　hollow　boundaries　in　rock　art　remains　limited.　Based　on　the　theory　of　heat　transfer,　this　paper　constructs　a　distribution　model　of　the　temperature　field　in　the　hollow　region,　quantifying　the　dynamic　characteristics　of　heat　diffusion　in　the　hollow　area.　A　new　method　for　calculating　the　hollow　radius　based　on　infrared　thermography　is　proposed,　and　a　correlation　analysis　of　the　unknown　parameter　β　in　this　method　is　conducted　using　numerical　simulation.　Based　on　the　analysis　results,　a　quantitative　relationship　between　the　parameter　β,　ω　and　R　is　established.　To　test　the　accuracy　of　the　computational　method　under　complex　surface　conditions,　calculations　were　performed　on　200　measurement　points　with　different　hollow　radii,　among　which　73　%　of　the　points　had　a　relative　error　of　less　than　14　%.　The　experiment　also　conducted　a　comprehensive　analysis　of　the　error　performance　of　different　methods　from　the　perspectives　of　time　and　hollow　size.　The　experiment　also　analyzed　the　error　performance　of　different　methods.　This　analysis　considered　both　the　time　dimension　and　the　size　of　the　hollow　areas.　The　experiment　also　conducted　a　comprehensive　analysis　of　the　error　performance　of　different　methods　from　the　time　and　hollow　size　dimensions.　The　results　show　that　the　method　in　this　paper　achieves　an　average　relative　error　rate　of　9.89　%　and　a　standard　deviation　of　0.22,　compared　to　16.69–24.82　%　and　0.32–0.42　respectively　in　existing　methods,　demonstrating　clear　advantages　in　accuracy　and　stability,　demonstrating　the　smallest　fluctuation　and　the　best　stability　compared　to　other　methods.　This　method　was　also　applied　to　on-site　hollow　detection　in　the　Damaidi　rock　paintings.　Its　performance　was　compared　with　the　on-site　hardness　testing　method.　Both　methods　gave　similar　results　in　identifying　hollow　boundaries.　This　further　confirms　the　reliability　and　applicability　of　the　proposed　method.　©　2025　Elsevier　Ltd