Raindrops size and shape from videosonde and image processing

This paper describes a new videosonde designed for providing shape and size of raindrops. It makes use of a CCD camera and image processing to acquire image sequence, to detect raindrops, to segment and measure shape and size of each detected particle. Depth is obtained by one fixed camera with 2 lights, which is a key feature of the video sonde. Raindrops are detected by their 2 shadows. As no model are accurate to represent raindrops, image processing uses a-contrario method for detection, and constraint filtering for matching. Results are evaluated on glass ball and real rain simulator.