Aiming at the problem of vehicle video detection in blind areas at night, a method of vehicle detection in blind areas at night based on CAdaBoost algorithm is proposed. Since night vehicles and daytime vehicles have different illumination, visibility, vehicle edge characteristics, vehicle contours, etc., the test image is first grayed out and ROI is determined. Then, because the features of the car head and wheels at night are more obvious than other features, a parallel method is used to train multiple weak classifiers offline at the same time, mainly to train the Haar-like features of the car head and wheels of the test image. In the meantime, in order to improve the accuracy of detection, the weighted parameter is introduced to determine the role of the weak classifier in the final strong classifier. Then, the offline trained model is used to match the Haar-like features of the test vehicle’s car head and wheels respectively, and finally realize the real-time detection of vehicles in the blind area. The results show that: for vehicle detection in blind spots at night, the algorithm has a high detection accuracy and low detection time.