基于多特征融合DNN的重力导航适配性定量评估OA

To address the suitability assessment requirements in long-endurance underwater gravity matching navigation,this paper proposes a quantitative analysis method based on multi-feature fusion and a deep neural network(DNN).Multiple representative gravity field feature parameters are integrated,and a DNN prediction model is constructed using matching probability and mean matching error as regression targets.This approach enables the transformation of suitability region analysis from qualitative evaluation to quantitative prediction.Experimental results demonstrate that,compared with traditional suitability assessment methods based on single features,the proposed multi-feature fusion approach reduces mean matching error and improves matching probability prediction accuracy by approximately 20%.Moreover,the model maintains favorable stability and reliability under limited sample conditions.The results demonstrate the effectiveness of the multi-feature fusion DNN framework in quantitatively assessing gravity matching navigation suitability,and provide informational support for underwater gravity navigation area selection and trajectory planning.