The anti-collision system is one of units for effectual operation and safety in crane operation. Most of the system have used ultrasonic to measure position and to prevent collision of crane. But, the anti-collision system using ultrasonic is fallen the reliability of performance by the change of temperature and humidity. Hence, this study proposes neural network applied to anti-collision system to improve accuracy and stability of the measured distance data, and we evaluate performance of the system. In results of experiments, the proposed method was seen that stability and accuracy of data are improved than that of the temperature compensation method.