Aiming at the three-dimensional bin loading problem of cargo, a three-dimensional cargo loading model is established. In the model, in order to avoid the phenomenon of rolling over due to the eccentricity during the turn of the goods in the process of transportation, the gravity center constraint during the turn was added to obtain the projection of the gravity center area as an isosceles triangle or isosceles trapezoid. The cargo placement rules expand the remaining space area and increase the diversity of understanding. In order to improve the speed of iterative convergence and enhance its global optimization ability, an improved crow search algorithm is adopted to solve and optimize the model. In the improved algorithm, a multi-probability random walk strategy and a reconciliation strategy are proposed and introduced. The solution repair strategy makes the algorithm suitable for model solving and increases the diversity of solutions as much as possible. The multi-probability random walk strategy is to continue to walk randomly with different probabilities after population iteration, which makes the global optimization ability of the algorithm stronger. Simulation examples and benchmark function test results show that the improved algorithm has obvious optimization effect.