【摘要】 为提高视觉着陆过程中无人机的相对定位精度，选取视觉图像中的直线交点作为结构化约束特征点，设计了基于梯度一致性的边缘检测算法，并结合Shi-Tomasi角点检测算法进行结构化约束特征点的粗定位。对LSD直线检测算法进行改进并设计了亚像素角点定位精度改进算法，在结构化约束特征点粗定位的基础上，将其精度提高到亚像素级。基于实际场景中固有约束的结构化约束特征点具有鲁棒性、旋转和尺度不变性，抗干扰能力更强，其高精度定位有利于提高视觉着陆相对定位的精度与可靠性。更多还原

【Abstract】 In order to improve the relative positioning accuracy of the UAV during visual landing process, the straight line intersections in the visual image are selected as the structural constrained feature points. The structural constrained feature points detection method is designed based on gradient consistency edge detection algorithm, and the rough positioning of structural constrained feature points is realized combined with Shi-Tomasi corner detection algorithm. On the basis of the rough positioning of structural constrained feature points, the improved LSD algorithm and subpixel corner detection algorithm are also put forward to raise the accuracy of feature points to subpixel level. Based on the inherent constraints in the actual scene, the structural constrained feature points are more robust and have invariance property in rotation and scale, which also have the capability of resisting disturbance. The highly precise positioning of the structural constrained feature points is beneficial to the accuracy and reliability of visual landing.更多还原