The increase in vehicle ownership every year causes a lack of information monitoring on each vehicle. As one of the methods used to find vehicle information, recognizing each number plate is a solution for recognizing vehicles. Utilizing object detection techniques using computer vision in recognizing vehicle number plates can simplify the plate recognition process. The process of identifying and classifying the characters on the plates is conducted simultaneously with a simple implementation which is a benefit of using computer vision in recognizing vehicle plates. The use of the Haar cascade classifier algorithm in this research overcomes the problem of plate detection combined with the Convolutional Neural Networks (CNN) to conduct Optical Character Recognition (OCR) on vehicle plates. The results of vehicle plate recognition in-situ experiments in four real-time tests obtained an average accuracy value of 42.67%.