Development Of Spaceboren Sea Region Extract Module For Large-field Optical Remote Sensing Image

The on-orbit sea targets detection based on visible light remote sensing images can greatly improve the timeliness of traditional processing methods,and has been widely concerned in military defense and civil field,It has become a hot spot in the field of remote sensing research.The on-orbit sea target detection is performed on the sea region in the remote sensing image,but the cloud and land areas contained in the image bring a lot of invalid interference information to it,which not only reduces the processing efficiency,but also as a false alarm severely restricts the accuracy of the test results.Sea region extraction can remove the land and clouds in the image and retain the sea region of interest.Therefore,sea region extraction is a necessary means to enhance the effect of on-orbit sea targets detection.In particular,the continuous expansion of the remote sensing image width,although it brings a larger field of view for target detection,has also caused the traditional sea surface extraction method to be wrongly segmented in application,thereby reducing the accuracy of extraction.In the implementation of satellite engineering,due to the limitations of the power consumption,volume and computing power of the spaceborne information processing platform,the current sea region extraction method is still difficult to achieve a good balance between the extraction effect and the computing performance,which unable to meet the real-time requirements of on-orbit target detection.The content of the research on the above problems is as follows:(1)Aiming at the problem of low accuracy of sea region extraction caused by the weak scene adaptability of the simple and efficient gray threshold segmentation method,a multi-level threshold joint discrimination is proposed.This method uses the local feature analysis of the sea surface area,cloud and land to identify the overall spatial characteristics of the image,and uses the statistically obtained thresholds to realize the type discrimination of the wide image slice sub-pictures,and provide information guidance for spatial characteristics of gray threshold method to improve the accuracy and efficiency of sea region extraction.Compared with the single gray threshold method,the experiment shows that this method can greatly improve the effect of extracting the sea surface area of wide remote sensing images when the algorithm complexity is slightly increased.(2)Aiming at the challenge of real-time calculation of the sea region extraction method,the accelerator design of the sea region extraction method is carried out,which based on the “ARM+FPGA” heterogeneous So C processor.Using the method of task division,the computationally intensive steps of image processing in the method are deployed in parallel on the FPGA side,and the steps of control and subgraph classification run on the ARM system,so that the sea region extraction method is divided into the most suitable device for performance improvement.At the same time,due to the limited computing resources of the on-orbit platform,by reducing the image and then restoring,the overhead of the cache and the calculation amount is reduced,thereby improving the calculation efficiency(3)Aiming at the problem that it is difficult for the on-board processing platform based on SRAM process heterogeneous processor to have both high performance and reliability,the hybrid system architecture design of "commercial grade heterogeneous So C device and Anti-radiation Flash FPGA" is adopted to realize the system has high performance and flexibility while providing space reliability.Based on the actual needs of remote sensing image on-orbit processing,a hardware module platform was developed,and carry on the firmware design of the corresponding function to finish the work flow of the module on the track.Finally,the sea region extraction accelerator was verified on this platform.Tests show that better hardware acceleration can be achieved.Experimental results show that the proposed sea region extraction method can improve the accuracy of processing wide-range remote sensing images.Finally,after the actual development of the module is completed,the feasibility of the spaceborne processing platform and on-orbit sea region extraction is verified,which has application value.