The traditional passive heat dissipation method has low heat dissipation efficiency, which is not suitable for the heat dissipation of the concentrated heat source inside the long-focal aerial camera, resulting in temperature level changes and temperature gradients in the optical system near the heat source, which seriously affect the imaging performance of the aerial camera. To solve this problem, an active heat dissipation method of liquid cooling cycle is proposed in this paper. To improve the solving efficiency and ensure simulation accuracy, a dynamic boundary information transfer method based on grid area weighting is proposed. The thermal simulation results show that the liquid cooling method reduces the heat source temperature by 70.12%, and the boundary temperature transfer error is 0.015%. The accuracy of thermal simulation is verified by thermal test, and the simulation error is less than 6.44%. In addition, the performance of the optical system is further analyzed, and the results show that the MTF of the optical system is increased from 0.077 to 0.194 under the proposed active liquid cooling cycle heat dissipation method.

To address the limitations of traditional methods in feature extraction and multi-modal information fusion, this paper proposes an infrared–visible image object detection architecture that integrates Convolutional Neural Networks (CNNs) and Deformable Transformers. This method leverages the advantages of CNN in local feature modeling and the capabilities of Transformer in capturing global contextual information, facilitating the fusion of semantic consistency and structural details across modalities. By introducing a detection-aware multi-task optimization mechanism, the model improves object detection in challenging scenarios such as low-light conditions, occlusion, and complex backgrounds. Experiments on multiple standard datasets, including M3FD and LLVIP, indicate that the proposed method achieves competitive or better performance than the compared methods in key metrics such as mAP. Specifically, our method obtains the best mAP50 among the evaluated methods with an mAP50 of 74.2% on the M3FD dataset and 98.6% on the LLVIP dataset, surpassing the second-best PIAFusion by 4.3% and 2.5% respectively. These quantitative results support the practicality and effectiveness of our approach in the evaluated complex environments.