Narrow passage problem is a problematic issue facing the sampling-based motion planner. In this paper, a new approach for narrow areas identification is proposed. The quad-tree cell-decomposition approximation is used to divide the free workspace into smaller cells, and build a graph of adjacency for these. The proposed method follows the graph edges and finds a sequence of cells, which have the same size, preceded and followed by a bigger cell size. The sequence, which has the pattern bigger-smaller-bigger cells size, is more likely to be located in a narrow area. The minimum spanning tree algorithm is used, to linearize adjacency graph. Many methods have been proposed to manipulate the edges cost in the graph, in order to make the generated spanning tree traverse through narrow passages in detectable ways. Five methods have been proposed, some of them give bad results, and the others give better on in simulations