基于YOLO改进算法的远程塔台运动目标检测

远程塔台由于其低成本高时效远程实时控制技术正越来越受到民航业界的青睐，其中运动目标自动检测和显示是远程塔台的核心技术，作为增强现实技术更好地为管制员提供服务。本文在分析了远程塔台机场场面背景复杂、场面目标多为远场景、小目标等特点基础上，提出了一种改进的YOLO算法来实现远程塔台运动目标的检测，算法核心思想以Darknet-53为基础网络，多尺度预测边界框，以运动目标图像坐标（x,y）的偏移量作为边框长宽的线性变换来实现边框的回归，改善了传统YOLO算法损失函数不同大小的边框未做区分的问题，提高了检测准确性和速度。机场真实数据实验表明，该算法能快速、准确的检测出远程塔台的运动目标，并准确的回归运动目标边框及分类。

Moving Target Detection of Remote Tower Based on Improved YOLO Algorithm

Remote tower is becoming more and more popular in the civil aviation industry because of its low cost, high efficiency and remote real-time control technology. The automatic detection and display of moving targets as an augmented reality technology is the core technology of remote tower, and it can provide better service to controllers. Based on the analysis of the characteristics of remote tower airport, such as complex scene background, remote scene and small target, this paper proposed an improved YOLO algorithm to detect the moving target of remote tower. The core idea of the algorithm was to predict the boundary frame of multi-scale based on Darknet-53 network and took the coordinates of moving target image as the basis (x, y) offset as a linear transformation of the length and width of the border to achieve the border regression, improved the traditional YOLO algorithm loss function of different sizes of the border was not distinguished, improved the detection accuracy and speed. Experiments on real airport data show that the algorithm can detect the moving targets of long-distance tower quickly and accurately, and accurately regress the moving object boundaries and classification.