类人足球机器人场地线检测及定位

Detection and location of humanoid robot soccer field line

为了解决传统类人足球机器人场地线检测算法在光照不稳定时准确率低的问题,提出一种基于全卷积网络(FCN)改进的场地线检测模型,通过增加卷积层数、构建残差块、在上采样过程中融合更多特征图,提高了检测准确度。实验结果表明,改进模型能够克服光照因素干扰,对场地线进行准确检测,平均像素精度与平均交并比均提高了约5%。为提高机器人自定位精度,提出一种基于后方交会法的机器人自定位方法,通过3个特征点的像面坐标与世界坐标,实现机器人自定位,实验结果表明该方法具有可行性,最大定位距离误差为17.5 mm。

In order to solve the problem of low accuracy of traditional humanoid soccer robot field line detection algorithm under unstable illumination,an improved field line detection model based on the full convolutional network is proposed.By increasing the number of convolution layers,building residual blocks,and fusing more feature maps during up-sampling,the detection accuracy is improved.Experimental results show that the improved model can overcome the interference of illumination factors and detect the field lines accurately,and the average pixel accuracy and crossover ratio are improved by about 5%.In order to improve the accuracy of robot self-positioning,a robot self-positioning method based on the rear intersection method is proposed,which realizes the robot self-positioning through the image plane coordinates and world coordinates of three feature points.The experimental results show that the method is feasible,and the maximum positioning distance error is 17.5 mm.