Abstract: The movement track of agricultural machinery includes not only field operation track, but also road driving track．Effectively distinguish operation tracks of field operation and road driving during driving of agricultural machinery, accurately divide effective operation plots and accurately evaluate operation efficiency of agricultural machinery, so as to realize remote intelligent management of agricultural machinery．Typical characteristic data was extracted by analyzing attributes of agricultural machinery trajectory points, and training model established by the method of BP_ AdaBoost was used to recognize track points of agricultural machinery．After re-marking error prone track points at junction of road and field, it was trained again．The correct rate of track recognition was 96.89%．It not only avoided dependence of traditional clustering algorithm on threshold and parameters, but also effectively solved problem of mistaking road driving trajectory into field operation trajectory．