An adaptive dynamic programming (ADP) technique is investigated to figure out the problem of image feature tracking control of constrained visual servoing (VS) manipulators system. The complete camera-manipulator dynamic model can be established through the mapping relationship between image feature and manipulator position. The control inputs in the VS manipulators system are taken as a constrained case, and the performance index function is considered to cooperate with sliding mode function. On the basis of ADP technique, a critic neural network (NN) is constructed to derive the Hamilton-Jacobi-Bellman (HJB) equation. Furthermore, the near-optimal control policy is obtained. The Lyapunov theory illustrates that the feature tracking of input-constrained VS manipulators system is ultimately uniformly bounded (UUB). Numerical simulation examples are given to certify the validity of the developed approach.