The invention discloses a cloud parking lot parking space optimization method based on a Dijkstra optimization algorithm, which includes the following steps: a video parking space detection terminal in a cloud parking lot uploads detected parking space data to a cloud platform; a weighted directed diagram of the road network of the parking lot is drawn according to the parking space data; an optimal parking space model is established according to the weighted directed diagram of the road network; the shortest driving distance and the shortest walking distance for each empty parking space are calculated using a Dijkstra optimization algorithm according to the optimal parking space model; and the value of relative driving distance Theta for each empty parking space is calculated, wherein the parking space corresponding to the minimum value in the set of values of relative driving distance Theta is an optimal parking space, and the path corresponding to the optimal parking space is an optimal path. The problem that the utilization rate of parking lot areas is unbalanced is solved under the condition of ensuring the vehicle tracking distance, and the Dijkstra optimization algorithm makes the optimization process simpler and more efficient.