Abstract: Auxiliary transportation vehicles play an important role in the transportation of personnel, materials, gangue, and equipment underground in coal mines. The narrow space and complex environment of underground tunnels in coal mines, with alternating light intensity, can easily lead to driving safety accidents caused by fatigue, blurred vision, and operational errors of auxiliary transportation vehicle drivers. With the continuous promotion of the construction of smart mines in China, improving the safety and intelligence of auxiliary transportation vehicles is of great practical significance for achieving less manned and unmanned coal mines. Aiming at the assessment of the passability of auxiliary transportation vehicles underground in coal mines, a method for evaluating the trafficability of underground coal mine auxiliary transportation vehicles based on multi-feature fusion was proposed. Firstly, an enhanced binary image of the railway track is obtained using edge detection, morphological top hat transformation, and eight connected regions. The railway track is divided into two models: straight track and curved track by setting a threshold value using vanishing point coordinates. Progressive probability Hough transform and Catmull Rom spline curve with improved feature point selection are used to fit the straight track and curved track separately, forming a railway track detection method based on a switchable model; Secondly, based on the color attributes of coal mine underground tunnel wall pipelines, a pipeline feature detection strategy is constructed through HSV color space transformation, edge detection, and least squares fitting algorithms; Once again, a weighted fusion method based on pipeline and rail feature points is proposed to extract the centerline trajectory of auxiliary transportation vehicles, and to construct rules for generating safe passage areas for auxiliary transportation vehicles; Then, for the over-segmentation problem in the traditional watershed algorithm, two-dimensional edge and area feature extraction of static obstacles is carried out by using improved watershed segmentation based on markers and GraphCut algorithm, and a small-sample classification support vector mechanism is selected to build a pedestrian detection model based on the fusion of HOG (Histograms of Oriented Gradients)features and LBP(Local Binary Pattern) features for underground coal mine, so as to realize the comprehensive detection of dynamic and static obstacles in front of auxiliary transportation vehicles; Finally, the positional relationship between obstacles and safe passage area is judged by Euclidean distance to determine the obstacle invasion, and the results of auxiliary transportation vehicle passability assessment are obtained. The experimental results show that the proposed method can realize the detection of railroad track features, pipeline features, and safe passage area in underground coal mines, and effectively evaluate the passability of auxiliary transportation vehicles in underground coal mines.