基于Mask R-CNN的机场地面保护区航空器识别研究

Research on aircraft identification of airport ground protection zone based on Mask R-CNN

摘要:

为解决民航中小机场缺乏低成本运行的防跑道侵入技术的问题,提出了一种通过摄像头获取地面保护区图像信息以防止跑道侵入的模型。该模型首先利用迁移学习Mask R-CNN对航空器进行识别和跟踪,通过结合摄像头在地面保护区的布局方案,提出了一种基于摄像头的检测结果判断是否发生跑道侵入的逻辑电路。该模型通过等比例缩小的机场模型进行实验验证。结果表明,该模型的跟踪准确率为95%,可以有效地跟踪机场场面目标,进而判断跑道侵入的发生。

Abstract:

To solve the problem of lack of low-cost preventing runway incursion technology in small civil aviation airports, a model for preventing runway incursion based on the image information of ground protection zone from camera is proposed. We first establish the tracking algorithm of field aircraft by using the transfer learning Mask R-CNN. Then, a logic circuit, which combines the camera signals with the cameras layout of the ground protection zone, has been proposed for determining runway incursion. The proposed method is verified by a scaled down airport model. Simulation results demonstrate that the proposed method of its accuracy 95% is effective in the treatment of field target tracking, leading to runway incursion determination.