This is the authors' abstract. We don't add key points for this paper.
The detection of temperature change areas on water surfaces is a challenging task owing to low contrast, coarse noise, and objects with diverse morphology in infrared videos. In this article, from the perspective of motion object detection, a spatio-temporal fusion-based object segmentation method is proposed for monitoring temperature variations on water surfaces. The proposed method is composed of three stages, i.e., density peak-based object localization, spatio-temporal fusion-based object segmentation, and enhancement. Specifically, the Gaussian mixture model (GMM) and density analysis are initially applied to periodic video frames to achieve object center localization. The foreground extracted by the GMM is then refined through the fusion of temporal and spatial information. Here, temporal information refers to the historical foreground information; spatial information refers to the adopted distance, cluster, and edge-preserving priors. In the experiments, an indoor experimental set-up for generating and monitoring water surface temperature variations is introduced, and an infrared video dataset for detecting temperature variations (IDTVs) is built. Experimental results confirm the effectiveness and utility of the proposed method on the IDTV dataset, which even surpasses the supervised methods and the large models-based methods.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Kang et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: