Vehicle bodywork attitude measurement has been important in automotive technology. This paper presents a non-gyro bodywork attitude measurement program. Using a three-axis acceleration sensor output real-time body posture through BP neural network. Program skillfully made physical model and the algorithm simple according to the characteristics of motor cars. Taking advantage of the nonlinear function learning and generalization ability of ANN, the system made the input three-axis inertial acceleration values of harmful components filtered so as to mapoutput Attitude. And finally, it presents the data of experiments. By taking full advantage of ANN which precessing in parallel, and good at dealing with problem uncertainty, this program provides a solution of calculating real-time bodywork attitude, while with minimal hardware overhead.