Abstract: The previous researches mainly focused on traffic flow state recognition in advance (namely traffic flow forecasting) and realtime recognition (namely incident detection or traffic flow breakdown detection), but recognition of quantitative change rule of traffic flow was neglected in some degree. Based on the traffic flow theory and the elements of projective transform in descriptive geometry, the key to the changepoint recognition was analyzed by the complanation of dualistic linear regression. This method is calibrated and tested on the field data of a road to verify the validity and the feasibility of the theory.

Key words: traffic flow, linear regression, three-dimensional scatter plot, projective transform

摘要： 交通流状态辨识的以往研究主要集中在交通流状态预辨识(即交通流预测)和实时辨识（即事件检测或交通流突变检测）上，在一定程度上忽略了对交通流量变规律的辨认。依据交通流理论，结合画法几何中投影变换原理，以二元线性回归平面化研究为例，对变点辨识的关键问题进行了分析。结合某路实测数据对本方法进行了标定和检验，验证了方法的有效性和可行性。

关键词: 交通流, 线性回归, 三维散点图, 投影变换