To address the challenges in recognizing number plate images in car park environments caused by factors such as lighting and tilt, this paper proposes a method for number plate localization and correction based on morphological analysis, the Hough transform and the minimum variance of projected point coordinates. Character segmentation and recognition are achieved using vertical projection and template matching. First, coarse localization of the license plate is achieved through the top-hat transform and Canny edge detection, followed by fine-tuning using prior knowledge such as area and aspect ratio. Next, the Hough transform is employed to detect straight lines in the border for horizontal correction, whilst vertical shearing correction is performed using the criterion of minimum variance of projected point coordinates. Subsequently, the corrected binary image is vertically projected to segment the characters, and finally, the license plate number is output by matching it against a standard template. A parking management system with toll management functionality was developed in the MATLAB environment. Testing was conducted on 57 real-world license plate images, achieving an overall recognition accuracy of 91.2%, thereby validating the effectiveness of the method.