Abstract: The motion regularity of non-uniform bedload particles has been a difficulty in sediment research, and extracting the parameters of non-uniform bedload particles is a basic work to overcome the difficulty. Through flume experiment of non-uniform bedload on rough bed, a large amount of videos of particle motion were obtained, and some of the videos were selected for processing and research: firstly the video shot was converted into images by image conversion software, and then the deformity was corrected and the background noise was removed of the converted images by spatial transformation and cutting; lastly the sediment particle contours were extracted by selecting a reasonable threshold. Furthermore, parameters inclusive of particle size and particle volume were calculated to provide basis for obtaining the basic information of bed load particles motion. The research offers scientific and improved, reliable data for researching the transport mechanism of non-uniform bed load sediment.

Key words: water quality prediction, neural network model, dynamic equation, mechanism priori knowledge, Mackey-Glass chaotic system