The linear inversion of hydro-geological prospecting relies on the initial model deeply and its accuracy is poor, so the paper applies the hybrid algorithm of ant colony algorithm (ACO) and BP neural network. Through the MATLAB programming to verify the classical model of the stone with different position date. The inversion results are in agreement with the actual models. In order to highlight the advantages of the joint inversion, the BP algorithm is compared with the ACO-BP algorithm, The results show that the ACO-BP algorithm has its advantages in the inversion occurrence of the abnormal body, it can achieve the goal of advance warn in geological engineering.