Research And Implementation Of Multi-source Aviation Surveillance Information Fusion Technology

With the rapid development of computer science and information technology,multi-sensor information fusion technology is also developing at a high speed,especially in military and aviation.The radar,which is the main sensor monitoring aircraft in aviation domain,is affected by its own precision,location and surroundings.As a result,there is some difference in the surveillance information when multiple radars monitoring the same aircraft.In conclusion,it is very important to fuse the surveillance information of multiple radars monitoring the same aircraft in order to get more accurate result.This paper has concerned radar surveillance information as the main research of aero surveillance information,and the research work can be divided into several parts below:firstly,the paper has digged into many types of format as well as parsing the information.And then the paper has completed the time calibration and space calibration work of parsed results,and has unified multiple radar surveillance info into same coordinates system and time scale.On the the basis of BP algorithm of neural network,the paper has put forward the multiple radar surveillance information fusion method based on BP neural network,and has designed related network structure and component modules.Finally,the paper has realized the fusion of multiple radar surveillance information of single aircraft.On the basis of methods mentioned in the second paragraph,the paper has put forward a new fusion method based on the research on GPU architechture.This method fuses multiple radar surveillance information of multiple aircraft simultaneously using BP neural network and GPU accelerating.The implementation takes the advantage of GPU parallel computing,and the accuracy of fusion results do not fall,so that the training time of multiple aircraft network has been shortened,and effiency has been improved.In the tests,the paper has tested the training time of multiple aircraft network on CPU and GPU separately when aircraft's quantity is different,and test results has shown that the training time on GPU has reduced fifty percent.Finally,the multi-radar surveillance information fusion of multi-plane is realized on the BP neural network trained above.The method which fuses multiple radar surveillance information of single and multiple aircraft based on BP neuron network and GPU,has reduced the the average distance deviation and average azimuth deviation observed by multiple radars.