基于颜色不变量和轮廓特征的地面多圆形目标检测

研究了一种室外环境旋翼无人机对地面多圆形目标检测的方法。考虑室外环境光照变化、阴影等自然因素以及无人机飞行高度、姿态变化等因素对目标检测带来的不利影响,首先引入颜色不变量特征;并采用改进K-means算法进行图像分割;其次根据目标轮廓特征,设置面积与圆形度阈值滤除干扰区域;最后,采用基于对称性的最小二乘法与残差度确定目标位置。实际实验及无人飞行器大赛验证了所研究方法的实时性和准确性。

Multiple Ground Circular Targets Detection Based on Color Invariant and Contour Feature

The method of detecting multiple ground circular targets by UAV in outdoor environment is studied in this paper. In order to reduce the impact of natural factors such as outdoor environment illumination change, shadow, and UAV altitude, angle change as well as other factors on target detection, the method firstly introduces the color invariant feature, and uses the modified K-means algorithm for image segmentation. Secondly, the area and circularity threshold are set to eliminate interference area depend on the target contour feature. The symmetry of the least squares and the residual are used to determine the target location. The physical experiment and the contest of UAV demonstrate the proposed method is real-time and accurate.