Ship Targets Classification and Recognition of Random Forest Based on AIS Data Feature Optimization
Accurately and efficiently classifying and recognizing ship targets is of great significance to promote mari-time intelligent transportation management and enhance maritime situational awareness.In the existing methods of ship targets classification and recognition,the multi-dimensional feature space is generally constructed,which is easy to cause information redundancy,resulting in the reduction of classification accuracy and efficiency,etc.A method of ship targets classification and recognition of random forest based on AIS data feature optimization is pro-posed in this paper.Firstly,18-dimensional features related to ship speed,acceleration,course and distance are extracted from AIS data.Secondly,the average impurity reduction method is used to evaluate the importance of fea-tures and optimize the best feature combination.Finally,the ship targets are classified and recognized by using the feature optimized random forest,and the classification results are compared and evaluated with those of the original random forest.The experimental results show that the speed features and distance feature extracted from AIS data play an important role in the classification of ship targets.The optimal 14 features ordered by importance are select-ed,which can efficiently use the abundant information contained in AIS data,reduce the complexity of the model and better distinguish different types of ships.The overall classification accuracy reaches 86.2%,and the classifi-cation efficiency is better than that of the original random forest,which can meet the needs of accurate and efficient classification and recognition of ship targets.
ship targetsclassification and recognitionAIS datafeature optimizationrandom forest
万方数据
维普
