<p>The combination of prefabricated type and low energy consumption has become the core development trend of the construction industry. However, existing studies often overlook non-uniform heat transfer testing of prefabricated walls based on infrared images. This paper aims to proposes a comprehensive evaluation index, incorporating cumulative temperature difference ratio and equivalent heat transfer coefficient, to analyze the non-uniform heat transfer characteristics of prefabricated exterior walls. The impact of non-uniform heat transfer on prefabricated exterior walls is assessed through non-uniform influence area and exterior wall heat transfer performance. To enhance efficiency, we employ image segmentation technology utilizing infrared thermal imaging to analyze and detect the non-uniform heat transfer patterns of prefabricated walls. The EM algorithm of an improved Gaussian mixture model is utilized for multi-classification on thermal imaging images, identifying non-uniform heat transfer affected areas and calculating the effective heat transfer coefficient. The results demonstrate the efficacy of a three-class Gaussian mixture model for visualizing non-uniform heat transfer influence areas. The image segmentation algorithm, improved based on the K-means clustering algorithm, exhibits superior performance in terms of PSNR, SSIM, and FSIM indicators. Furthermore, the proposed method enables automatic calculation of cumulative temperature difference ratio and equivalent heat transfer coefficient, showing within 5% and 10% error, respectively, compared to measured values. This study establishes a preliminary automatic recognition and calculation mode for thermal performance parameters of prefabricated walls. The findings can provide a convenient and rapid analysis of prefabricated walls thermal performance.</p>