The visual detection of bullet holes in a target usually can be formulated as a problem of how to exactly locate in an grayscale image the isolated and overlapping spots

which are treated as image structures without inner detail and texture. In this paper a model based matching algorithm is used to deal with this problem. A non linear model is used to describe gray and position information of those spots in a grayscale background and overlapped spots can be seen as single spots with intersection. Then the model matching process is applied to search and locate isolated and overlapping spots with various sizes

gradients and backgrounds. When we use this algorithm for holes detection and recognition in a military target

according to the characteristic of the time axis distribution of bullet holes

a feed back method is used for better performance. Experiments show it works well and achieves high reliability in real environments.