A model design method for efficient recognition and classification of radar targets of moving UAV swarms
By using evolutionary algorithms to optimize the neural network architecture in the inter-marching drone group radar target recognition and classification tasks, the problem of high computational complexity of existing models is solved, and high precision and real-time target recognition is achieved.
CN115421121BActive Publication Date: 2025-04-01NORTHWESTERN POLYTECHNICAL UNIV
2 Cites 0 Cited by
Patent Information
- Application Number
- CN202211033451.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-26
- Publication Date
- 2025-04-01
- Estimated Expiration
- 2042-08-26
AI Technical Summary
Technical Problem
Although the existing neural network model can ensure high accuracy in the radar target recognition and classification tasks of the inter-marching drone swarm, it is difficult to meet the real-time requirements.
Method used
The neural network architecture search technology based on evolutionary algorithms is used to optimize the coding of graph structures to generate a simpler neural network model, thereby reducing the computational complexity and improving real-time.
Benefits of technology
On the basis of retaining the recognition accuracy, the complexity of the classification model is significantly reduced, the calculation amount of target recognition is reduced, and real-time is improved.
✦ Generated by Eureka AI based on patent content.
Abstract
The present invention discloses a model design method for efficient recognition and classification of radar targets of a group of moving unmanned aerial vehicles, and constructs a first classification model based on a neural network; the first classification model is used to perform target recognition on echo data of radar targets; the first classification model is converted into a graph structure code; an evolutionary algorithm is used to optimize the graph structure code to obtain the optimized graph structure code; a second classification model is generated according to the optimized graph structure code, and the second classification model is used to perform target recognition on the echo data of radar targets; the present invention converts the constructed first classification model into a graph structure code, and then optimizes the graph structure code using a neural network architecture search technology based on an evolutionary algorithm, so that the complexity of the classification model can be greatly reduced on the basis of retaining the accuracy of the first classification model, thereby reducing the amount of calculation for target recognition and improving the real-time performance of target recognition.
Need to check novelty before this filing date? Find Prior Art