In terms of point-in-polygon tests, a grid method proposed recently exhibits high computational efficiency. This method organizes the polygon fragments within each grid cell into stripe structures, ensuring that edges in each stripe intersect with both the left and right boundaries of the stripe. In this way, localization computation is enhanced, and GPUs are used for convenient parallel computation, resulting in a detection efficiency superior to that of various previous methods. However, stripe structures constructed based on grid cells generate redundant stripes. Besides, the method has a high space requirement for stripe construction, making it inconvenient to construct stripe structures on GPUs. In response to this, this study proposes to construct stripe structures via grid rows. Thus, redundant strips can be eliminated, and the space requirement for the creation of computation is reduced, due to which stripe structures can be constructed on GPUs, and work efficiency is improved. Experimental results show that, compared with the original method, the new method significantly accelerates the construction of stripe structures, even by over 40 times. Moreover, it has a faster detection speed and can handle dynamic polygons more efficiently.