Anti-Interference Recognition for Aerial Infrared Object Based on Convolutional Feature Inference Network

The infrared decoy countermeasure technology develops along with the infrared imaging guidance technology. The release of infrared decoys can block the object and destroy the integrity of the object features. The traditional statistical approach based on feature fusion matching for pattern recognition is not effective in handling this problem. To improve the characterization ability of the invariant detail features in the infrared object domain and improve the classification accuracy of the probability identification model, this article proposes an inference recognition model that combines the convolutional features and the Bayesian network. The proposed model exploits the convolutional neural network to extract the detailed feature information of the object in the infrared image. Then, it uses the 2DPCANet screening mechanism to reduce the dimensionality and select more obvious detailed information in the candidate region. Based on the combination of human prior knowledge and posterior data, an infrared object recognition probability model is established through the application of the Bayesian network. The performance of the proposed algorithm is tested on the complex air combat simulation image dataset. The experimental results indicate that the recognition rate of the proposed anti-interference recognition algorithm reaches 94.17%.