基于高密度数据和聚类分析的独立车辙识别和评价

Isolated rutting identification and evaluation based on high-density data and clustering analysis

现行规范《公路技术状况评定标准》(JTG H20—2007)以一定长度内车辙检测的平均值作为车辙评价值,而均值会对实际的车辙深度产生平滑作用。为量化现行规范车辙评价方法对车辙评价的误差,充分挖掘高密度检测数据,更加准确地对车辙进行评价,定义独立车辙,提出了基于高密度数据和聚类分析的独立车辙识别和评价方法。研究了不同数据密度识别独立车辙的结果,并利用实际1km和20km的自动化车辙检测数据,说明所提方法识别和评价独立车辙的严重程度及分布位置的有效性和准确性,将结果与现行规范车辙评价方法所得结果进行对比,对2种结果的误差进行了量化。研究结果表明:所提方法适用于所有等间隔的高密度车辙检测数据,而现行规范采用1km车辙深度平均的评价方法所得结果已不能准确反映车辙的严重程度及位置。1km路段中,利用所提方法能够找到3条独立车辙,并确定其位置和严重程度,对独立车辙的评价结果较现行规范车辙评价结果更加准确;20km路段中,利用现行规范车辙评价方法只有25.1%的车辙被识别出,其中仅18.52%的车辙能够正确判断严重程度。而利用所提方法可识别全部车辙,且车辙严重程度判断正确率达到82.3%。结果显示现行规范采用1km车辙深度平均的评价方法不适用于分布不均匀的车辙评价,且车辙严重程度越高,分布不均匀程度越大,评价误差越大。

The average detected rutting value in a certain length was regarded as the rutting evaluation value according to the current specification of ＂Highway Performance Assessment Standards＂ （JTG H20-2007）. However, average data can smooth the actual rutting depth. To quantify the error in current rutting evaluation, mine the high-density data sufficiently, evaluate rutting more accurately, isolated rutting was defined in this paper, and a rutting identification and evaluation method was proposed based on high-density data and clustering analysis. Isolated rutting identification results obtained using rutting measurement data with different density in driving direction were studied. Case studies, using 13-point laser bar data collected on 1 km and 20 km roadways, respectively, demonstrated the effectiveness and accuracy of the proposed method for identifying and evaluating the severity level and location of isolated rutting. The results were compared with those obtained using the current 1 km average method and the error of these two methods for evaluating rutting were quantified. The results show that the proposed method can be used for all equally-spaced and high-density data, but the current specification using the average rutting depth of 1 km is not able to identify the rutting severity level and location accurately. In a 1 km section, using the proposed method, three isolated ruts are identified, the locations and the severity levels are identified correctly, and the evaluation results are more accurate compared with those using the current method. In a 20 km section, using the current method, 25.1% rutting are identified, among which 18.52 % can be accurately evaluated its severity level. However, using the proposed method, all rutting are identified and the accuracy of severity level identification reaches to 82.3%. Current method using the average rutting depth of 1 km is not suitable for evaluating non-uniform rutting. The higher the rutting severity level, the more non-uniform of rutting distribution is, and the higher the rutting identification error is. 3 tabs, 7 figs, 23 refs.