The solution of nonsmooth composite function problem with polyhedral constraints was studied. For the problem of nonsmooth composite function, a smooth function was firstly constructed to approximate the nonsmooth objective function, and the purpose of solving the original problem was then achieved by handling the smooth approximation problem. On this basis, considering the special structure of polyhedral constraints, the idea of sequential quadratic programming algorithm was adopted. A quadratic programming subproblem was solved to obtain the search direction. The active set strategy was adopted by the solution of quadratic programming to approach the optimal solution of the search direction by solving a series of quadratic programming problems with equality constraints step by step. After that, the step size was obtained by line search. The next iteration point was then obtained. Finally, the global convergence of the algorithm was proved theoretically. A preliminary numerical experiment was carried out to compare the algorithm with the smoothed sequential projection shrinkage algorithm. The results show that the algorithm has certain advantages not only in iteration times but also in computing time.