井下低照度环境下的钻孔机器人目标跟踪器

Target tracker of drilling robot under low-illumination environment in coal mine

煤矿井下是典型的低照度场景,为了解决一般摄像头在此环境下因特征提取不足导致跟踪失败的问题,提出了一种融合图像增强算法与多通道背景感知相关滤波器的低照度单目标跟踪器。采用增广拉格朗日方法来求解图像增强与目标跟踪的优化问题;提出采用目标区域光照均值判定方法来解决局部环境光源对目标亮度判别干扰的问题;设定亮度判据,实现单帧图像亮度判别与增强,保证了跟踪器在光照变化时提取目标特征的完好性。仿真实验验证了在暗光条件下,该跟踪器的跟踪准确度与重合度远高于一般的主流跟踪器。现场试验结果表明,该跟踪器实时帧率可达40 Hz,平均跟踪重叠精度为81.6%,有效地解决了机器人采用普通摄像头在煤矿低照度环境下的目标跟踪问题。

Underground coal mine is a typical low-illumination environment.In order to solve the problem of tracking failure due to insufficient feature extraction by general cameras in this environment,a low-illumination single-target tracker was proposed,which incorporated image enhancement algorithm and multi-channel background-aware correlation filter.The Augmented Lagrangian method was used to solve the optimization problem of image enhancement and target tracking.A method for determining the mean value of light in the target area was proposed to solve the problem of the interference of local ambient light source on the target brightness discrimination.The luminance detection criterion was set to realize the brightness discrimination and enhancement of a single frame image,which ensured the integrity of the target feature extraction when the illumination changes.Simulation experiments verified that the tracking accuracy and overlap of the present tracker were much higher than the general mainstream trackers under dark light conditions.The field test results show that the real-time frame rate of this tracker can reach 40 Hz and the average tracking overlap accuracy is 81.6%,which effectively solves the target tracking problem of robots using ordinary cameras in the low-illumination environment of coal mines.