The gradient method and standard deviation method are only the inversion of the atmospheric boundary layer height of the single moment. For this issue, an image edge detection method is presented using retrieve the height of atmospheric boundary layer over a period of time. Through the introduction and analysis of the conventional method, the necessity of the research on image edge detection is illustrated. The atmospheric boundary layer height of a day and night and four kinds of different weather backgrounds are analyzed by using the independent research and development lidar to measure real-time data. The minimum root mean square errors of the image edge detection method with the gradient method and the standard deviation method for the inversion of the atmospheric boundary layer height are 9.4 m and 11.4 m, respectively. Experimental results show that the proposed algorithm is simple, reliable and high accuracy, and there is no need to choose the initial value, which has lower sensitivity and stronger adaptability than the traditional methods.