Aiming at challenge small target recognition in complex environment, this paper proposes an airborne radar target recognition method based on the spatiotemporal distribution characteristics of power towers. The proposed method comprehensively considers the actual working conditions of helicopter low altitude flight, utilizes the agile and adaptable characteristics of artificial intelligence technology, and adopts the CNN-LSTM improved model to achieve high-precision recognition of power tower line targets by airborne collision avoidance radar. This article introduces strategies such as weighted entropy and sample expansion in sample construction, effectively avoiding model capability defects caused by sample imbalance. The experimental results demonstrate that the proposed method has higher accuracy and robustness compared to traditional radar target detection methods and typical CNN/LSTM networks in complex backgrounds.