Urban structure plays a key role in providing available paths for pedestrian flow through urban areas. Land-use planning influences the accessibility behaviours of pedestrian movement controlled by urban structures, activities, and street networks with the unique attributes of each urban area. To improve urban spatial planning in terms of adopting effective land-use options and enhance a better public transportation accessibility, we consider combining the following two techniques; detection of pedestrians using computer vision, and trajectories of crowd movement using land-use agglomeration pattern analysis. Applying the proposed method to a high-density area composed of multi-directional crossings at a T-way junction in front of Hakata station, Fukuoka, Japan, it is shown that the derived correlation coefficient between the closeness value and the volume of commercial building space indicates a strong relationship between these two variables, resulting in the conclusion that the proposed method is useful for application in the design of urban spatial plans.