Abstract: A well-illuminated environment contributes to the reduction of traffic accident rates, and artificial intelligence-based streetlight brightness detection has rapidly become a research hotspot in the fields of intelligent lighting and smart transportation. However, existing research on brightness detection focuses on simply stacking the depth of neural networks to increase the sensory field of features, often neglecting their spatial features, which leads to issues such as low detection accuracy and inability to meet real-time requirements. To address these limitations, a streetlight brightness detection algorithm that integrates spatial features is proposed. This algorithm innovatively introduces a spatial attention mechanism into the residual network for brightness detection to extract spatial features. The extracted spatial features are fed into a multi-scale detector constructed by the proposed feature pyramid network to achieve feature-level fusion and obtain the final streetlight brightness detection results. Experimental results show that the proposed algorithm achieves a mean average precision of 96.8% on the streetlight dataset, representing a 1.4% improvement compared with the YOLOv3 algorithm. This performance meets the demands of practical engineering applications and enables intelligent detection of streetlight brightness.