To enhance intelligence and accuracy of the spatial rotating project’s recognition

a new strategy based on aim body’s moment invariants and neural BP network is developed.Target project’s 7 hu moment invariants（HMIs） and 3 affine moment invariants（AMIs） are given weights by Relief algorithm to evaluate their recognition capability.The moment invariant with bigger weight is chocen as BP neural network input features to train the net and recognize project.Simulation test shows that this method in which choose rotary object’s moment invariants by relief algorithm to recognize aim body is efficient.Choose BP neutral net input feature with Relief can not only deduce features extraction time and decrease its compute time but also help to net’s convergence and increase its recognize accuracy.