The objective of this research is to develop an automatic system for thinning out radish seedlings. In part 1, we discussed a new cultivation method and the best time for thinning out, suited to automatic system. This paper proposed an algorithm how to determine which radish seedling should be thinned out. The approach, based on the OTSU's automatic thresholding method, was firstly applied to acquire the binary images, using the information of ExG (Extra Green) and brightness of radish seedling images. Secondly, we took a projection of the binary image onto horizontal X direction, particularly calculated the summation of binary image's pixels on the vertical Y direction. And then using the summation, the radish seedlings that should thin out, are determined. Finally, the experiments performed on 439 radish seedling images taken from the field, showed the proposed algorithm is effective for about 97% of total images.