The invention relates to a vehicle environment perceiving and controlling method based on cloud machine learning. The method comprises the steps that A1, image data of dead zones of the front portion, the rear portion and behind the left side and the right side of a vehicle, and millimeter wave radar data of a barrier in the area in front of the vehicle are collected; A2, character vector sets of the image data are extracted, and the distance between the barrier in front of the vehicle and the vehicle is obtained by processing the millimeter wave radar data of the barrier; A3, machine learning is conducted on all the character vector sets, and barriers on the front side, the rear side, the left side and the right side are recognized; A4, target tracking is conducted, and environment perceiving data of a target and current state data of the vehicle are obtained; A5, the safe distance is worked out, and active vehicle safety control is executed correspondingly; and A6, the environment perceiving data of the target are uploaded to a cloud service system for learning, and finally target recognition parameter sets and algorithm control parameter sets are upgraded. By the adoption of the vehicle environment perceiving and controlling method, the accuracy of environment perceiving, the vehicle control precision and measurement distance are effectively improved, and the performance of an active vehicle safety system is improved integrally.