Satellite ship target on-orbit detection device determination method and device and detection method

By quantifying, pruning and layer fusion processing of the SAR satellite target detection model, the problem that traditional SAR satellites cannot achieve real-time detection in orbit is solved, efficient and real-time detection of ship targets is achieved, and reliable maritime security guarantees are provided.

CN119580113AActive Publication Date: 2025-03-07SOUTHWEST JIAOTONG UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411514161.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-29
Publication Date
2025-03-07
Estimated Expiration
2044-10-29

AI Technical Summary

Technical Problem

Due to limited resources, traditional SAR satellites cannot achieve real-time detection of ship targets in orbit.

Method used

By obtaining the trained object detection model, determining the weight value and activation value of the convolution kernel, calculating the L1 norm, quantizing and pruning, obtaining a quantitative pruning object detection model, and using the layer fusion method to fuse the convolution and fully connected layers to achieve on-orbit real-time detection.

Benefits of technology

Real-time detection of ship targets is achieved, and can efficiently monitor marine dynamics and provide timely and reliable maritime security guarantees.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119580113A_ABST
    Figure CN119580113A_ABST
Patent Text Reader

Abstract

The invention provides a satellite ship target on-orbit detection device determination method and device and a detection method, and relates to the technical field of ship target satellite on-orbit detection. A target detection model is obtained, and a weight value and an activation value are determined; according to the L1 norm, sorting the channels from large to small according to the importance; dividing the sorted channels into m equal parts, and performing quantization and pruning to obtain a quantization pruning target detection model; fusing a convolution + BN layer and a full connection + BN layer of the quantitative pruning target detection model by using a layer fusion method to obtain a layer fusion target detection model; determining an FPGA hardware operator required in a forward reasoning model for realizing the layer fusion target detection model; and determining required FPGA board card resources according to the required FPGA hardware operator in combination with the target detection frame rate. According to the invention, soft and hard collaborative lightweight ship target detection and on-orbit ship target detection are combined, so that the on-orbit satellite can efficiently detect the marine target ship in real time.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Ship detection device and method for optical remote sensing satellite image

    CN112347857A

  • SAR image ship detection method and system based on lightweight deep learning

    CN113469073A

  • Image recognition method and device, and computer-readable storage medium

    US20230334632A1

  • Method for enhancing SAR image data for ship target detection

    WO2023050746A1