Chaff cloud is widely used in the field of electromagnetic interference. Chaff clouds are cheap to made and have a good interference performance. However, there are many difficulties in electromagnetic scattering evaluation of chaff cloud. One of the core problems is to establish a proper diffusion model of the chaff cloud. In fact, there are many unsolved problems in chaff cloud diffusion simulation with aerodynamic method. Therefore, the diffusion process of chaff cloud can be studied by means of high-speed photography. How to obtain the length of chaff elements from the picture is a problem that must be solved. In this paper, support vector machine and HOG algorithm are used to extract the length of single chaff element from the image. The simulation results show that the accuracy of the proposed chaff length recognition algorithm can reach 99%, and the recognition time of a single image is less than 1 s. It improves processing efficiency of chaff cloud diffusion modeling.