Embedded front vehicle detection system based on fusion of machine learning and traditional methods

For the detection of remote and fuzzy small targets, machine learning detection has a higher probability of missed detection. It is more dangerous when applied to vehicle detection. Therefore, this paper proposes and uses a detection method that combines machine learning and traditional methods to deal with the problem that the vehicle in front with a high risk factor is missed. Through the designed prediction mechanism for fusion, the video frame images will be processed by different modules. Compared with target detection using machine learning only, the missed detection rate was reduced by 29.1%. Finally, we deploy it in an embedded device.