On-Board Real-Time Ship Detection in HISEA-1 SAR Images Based on CFAR and Lightweight Deep Learning

Pan Xu,Qingyang Li,Bo Zhang,Fan Wu,Ke Zhao,Xin Du,Cankun Yang,Ruofei Zhong
DOI: https://doi.org/10.3390/rs13101995
IF: 5
2021-05-19
Remote Sensing
Abstract:Synthetic aperture radar (SAR) satellites produce large quantities of remote sensing images that are unaffected by weather conditions and, therefore, widely used in marine surveillance. However, because of the hysteresis of satellite-ground communication and the massive quantity of remote sensing images, rapid analysis is not possible and real-time information for emergency situations is restricted. To solve this problem, this paper proposes an on-board ship detection scheme that is based on the traditional constant false alarm rate (CFAR) method and lightweight deep learning. This scheme can be used by the SAR satellite on-board computing platform to achieve near real-time image processing and data transmission. First, we use CFAR to conduct the initial ship detection and then apply the You Only Look Once version 4 (YOLOv4) method to obtain more accurate final results. We built a ground verification system to assess the feasibility of our scheme. With the help of the embedded Graphic Processing Unit (GPU) with high integration, our method achieved 85.9% precision for the experimental data, and the experimental results showed that the processing time was nearly half that required by traditional methods.
environmental sciences,imaging science & photographic technology,remote sensing,geosciences, multidisciplinary
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: **How to achieve efficient real - time ship detection on the HISEA - 1 SAR satellite**. Specifically, due to the lag in satellite - to - ground communication and the large number of remote - sensing images, the traditional ground - centered processing method cannot meet the requirements for rapid analysis and real - time information in emergency situations. To solve this problem, the author proposes an on - board ship detection scheme based on the Constant False Alarm Rate (CFAR) algorithm and lightweight deep learning. ### Problem Background 1. **Advantages and Challenges of SAR Satellites** - Synthetic Aperture Radar (SAR) satellites can operate under various weather conditions and are widely used in ocean monitoring and large - scale offshore ship target detection. - However, due to the lag in satellite - to - ground communication and the large number of remote - sensing images, it is difficult for traditional methods to achieve rapid analysis and real - time information acquisition. 2. **Limitations of Existing Methods** - Traditional methods rely on ground - processing platforms, resulting in a large delay in information extraction time and reducing the timeliness of data. - Although deep - learning methods have high precision, they have high computational complexity and are difficult to implement for real - time processing on satellite platforms with limited resources. ### Solutions To overcome the above problems, the author proposes the following solutions: 1. **Combining CFAR and Lightweight Deep Learning** - Use the CFAR algorithm for preliminary rapid and rough detection to reduce the computational burden of the subsequent deep - learning model. - Input the CFAR detection results into the YOLOv4 - tiny network for further precise positioning and classification. 2. **Hardware Acceleration Strategies** - Utilize the powerful computing power and low - power consumption characteristics of the embedded GPU (such as NVIDIA Jetson TX1) to achieve real - time processing. - Allocate logic - operation - intensive tasks to the CPU and floating - point - operation - intensive tasks such as matrix convolution to the GPU to maximize computational efficiency. ### Main Contributions 1. **Land - Sea Segmentation** - Remove land interference through geographical prior information and the region - growing method to improve detection efficiency and accuracy. 2. **CFAR Preliminary Detection** - Use K - distribution to model sea clutter and perform preliminary ship detection based on the CFAR algorithm. 3. **Lightweight Deep - Learning Refined Detection** - Input the CFAR detection results into the YOLOv4 - tiny network and use its ability to efficiently extract global context information to obtain more accurate final detection results. Through the above methods, the author has achieved efficient real - time ship detection on the HISEA - 1 SAR satellite, significantly improving the processing speed and detection accuracy.