研究目的
Aiming at the problem of traditional low-level recognition of key targets in remote sensing images, a method for target detection and recognition based on Faster R-CNN is proposed.
研究成果
The simulation results show that the method has high recognition accuracy and speed, and can provide reference for recognition of the key targets of remote sensing images. In the future, different convolutional neural networks can be considered for integration to increase speed and accuracy.
研究不足
The paper does not explicitly mention limitations, but potential areas for optimization could include computational efficiency on different hardware or generalization to other datasets.
1：Experimental Design and Method Selection:
The method uses Faster R-CNN for target detection and recognition in remote sensing images. It involves converting the NWPU VHR-10 dataset to VOC 2007 format, refining hyper-parameters based on training set categories, training the Faster R-CNN neural network to generate a model, and using this model for detection and recognition of unknown images.
2：Sample Selection and Data Sources:
The NWPU VHR-10 dataset is used, containing 800 very-high-resolution remote sensing images cropped from Google Earth and Vaihingen dataset, manually annotated by experts.
3：List of Experimental Equipment and Materials:
Hardware includes NVIDIA GeForce GTX's TITAN X GPU with 4 GB memory. Software includes TensorFlow deep learning framework.
4：Experimental Procedures and Operational Workflow:
Steps include dataset conversion, hyper-parameter refinement, neural network training with Faster R-CNN, and model application for detection.
5：Data Analysis Methods:
Simulation results are compared with mobilenet-SSD-v1 network in terms of training time and accuracy, using metrics from the training and test sets.
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
