论文部分内容阅读
农田环境中农作物大多呈近似直线的行垄分布特点,农用车辆自主视觉导航时通常利用这些景物特征作为跟踪目标。提出了一种计算车辆相对于跟踪目标位姿的新型方法,首先分析了传统算法中存在的计算量大、忽视图像平面中各像素权重不同等缺陷,而后依据跟踪路径局部线性模型假设,详细地推导了算法过程。基于视觉导航原型车辆的试验结果表明,与人工测量值相比,横向距离和航向角的误差均值都等于零,标准差分别为3cm和0.62deg。
Most of the crops in the farmland environment have the characteristics of straight line and ridge distribution, and the characteristics of these vehicles are usually used as tracking targets in the autonomous visual navigation of agricultural vehicles. A new method to calculate vehicle pose relative to tracking target is proposed. Firstly, the traditional algorithms are analyzed, such as large amount of calculation, ignoring the defects such as the different weight of each pixel in the image plane. Then, based on the local linear model assumption of tracking path, The algorithm process is deduced. The test results based on visual navigation prototype vehicles show that the mean values of lateral distance and heading angle error are equal to zero with standard deviations of 3cm and 0.62deg, respectively, as compared with manual measurements.