论文部分内容阅读
在无人机自主着陆过程中,传统地标识别方法的相似阈值确定需大量实验估计。为解决此问题,采用一种基于仿射不变矩和支持向量机的识别方法,首先设计了六圆组合的图标作为无人机自主着陆地标;由于无人机会拍摄到发生扭曲的地标图像,因此提取地标的仿射不变矩作为输入特征;最后将其输入支持向量机分类模型,完成地标的识别。与传统的几何不变矩和BP神经网络相比较,该方法提高了地标的识别精度并降低了识别测试时间,因此对地标识别具有一定的实用性。
In the process of autonomous landing of UAVs, the similarity threshold determination of traditional landmark recognition methods requires a large amount of experimental estimation. In order to solve this problem, an identification method based on affine invariant moments and support vector machines is adopted. Firstly, a six-circle combination icon is designed as an autonomous landing land-mark of a UAV. Because no-man opportunity takes a landmark image distorted, Therefore, the landmark affine moment is extracted as the input feature. Finally, it is input into the support vector machine classification model to complete landmark recognition. Compared with traditional geometric moment invariants and BP neural network, this method improves the recognition accuracy of landmark and reduces the recognition test time, so it has some practicality for landmark recognition.