Indoor target positioning is a crucial technique in the intelligent logistics system, visible light positioning technique becomes a feasible solution in the intelligent logistics center because it does not need extra communication equipments. However, the reflection of walls would lead to accuracy reduction of indoor visible light positioning, in view of this problem, a hybrid neural networks based indoor visible light positioning method is proposed. Firstly, the gated recurrent unit is used to capture the dependency of the optical power value in the optical power vectors, the one dimensional convolutional layer is used to extract the local spatial features of the optical power vectors; then, these deep features are fused to enhance the ability of feature learning for the optical power vectors, so as to improve the accuracy of indoor light positioning. Simulation results show that, compared to the other neural network based indoor visible light positioning methods, the positioning error of the proposed method is lower, and the time efficiency is within a reasonable range.