Aiming at the problems of current intelligent illegal lane-crossing detection system

such as the fixed location and the cumbersome steps to determine the solid dotted line

a mobile terminal illegal lane-crossing detection system based on the fusion of multiple information is designed.The system uses target detection and semantic segmentation technology to obtain vehicle and lane line information

and uses network pruning to simplify and compress the target detection network to achieve model size reduction; using a lightweight spatial convolution module to enhance the feature extraction of lane line segmentation network structure

to achieve High-precision segmentation of lane lines; the use of pixel statistics to complete the segmentation results of lane lines under occlusion

so that the segmentation of lane lines is not affected by occlusion.The vehicle and lane line information is fused to detect the indentation behavior

and the solid dotted line is judged using pixel-based statistics（horizontal projection algorithm）in multiple areas of interest.Experiments show that this system improves the accuracy and robustness of line-breaking violation detection

realizes mutual supervision of traveling vehicles

and has certain application value.