With the rapid development of lidar and other sensing techniques, autonomous vehicles and mobile robotics are in the phase of real applications. But due to the poor ranging accuracy and detection range in foggy situation, the all-weather application of lidar has been limited. In this paper, the model of echo laser signals of targets in the fog is established according to the transmission and backscattering models. A combined attention mechanism network (CAMN) based on convolutional neural network (CNN) is proposed to identify the echo signal in the fog. The results of simulation and experiments show that CAMN can effectively remove the interference of fog on the detection of pulsed laser signal. The mean of absolute errors of the detection achieves 3.13 cm at the range of 10 m at the scattering rate of 30%. The detection range reaches 42 m, doubling or tripling the numbers of other approaches. The approach can effectively improve the ranging accuracy and detection range of lidar in foggy weather. It provides the basis for real applications of lidar.