研究目的
To propose a ship detection method based on deep convolutional neural networks for PolSAR images that can effectively detect ships of different sizes and improve detection accuracy by comparing with the modified CFAR detector.
研究成果
The proposed ship detection algorithm based on deep convolutional neural networks demonstrates superior performance in detecting ships of different sizes in PolSAR images compared to the modified CFAR detector, achieving high accuracy and efficiency.
研究不足
The study acknowledges the limited availability of PolSAR images for training, which may cause overfitting, and the challenges posed by speckle noise and sea clutter on detection performance.
1：Experimental Design and Method Selection:
The study employs a modified Faster-RCNN method for ship detection in PolSAR images, incorporating preprocessing, modified Faster-RCNN based ship detector, and target fusion.
2：Sample Selection and Data Sources:
Real measured NASA/JPL AIRSAR datasets are used for validation.
3：List of Experimental Equipment and Materials:
NASA/JPL AIRSAR instrument for data acquisition.
4：Experimental Procedures and Operational Workflow:
PolSAR images are segmented into sub-samples using a sliding window, enriched training data through multiscale rotation, and processed through modified Faster-RCNN for detection.
5：Data Analysis Methods:
Detection probability and figure of merit (FoM) are used for performance evaluation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
