| Due to its unique climate and geological conditions in the Yellow River Basin,the relationship between water and sediment in the river is unbalanced,and the sediment content in the water body is relatively large.The sediment in the river is mainly suspended sediment,which will continue to move and settle during the process of transporting with the water flow,which will have a certain impact on the water quality,soil erosion control,river channel shape evolution,reservoir sedimentation,reservoir operation scheduling,etc.Therefore,it is very important to detect and analyze the size distribution and sediment content of suspended sediment particles in rivers.In this paper,in the laboratory,the water body is oscillated by the grid turbulence device and natural sediment particles are added to it,so that the sediment particles are suspended.The image of the suspended sediment particles is collected by the machine vision recognition system,and the sediment particles are subsequently identified by algorithmic image recognition technology.,calculate the particle size and sediment content,and study its vertical distribution characteristics at different heights,which provides a new idea for the measurement of sediment particles and the study of the movement mechanism of sediment particles.The main contents are as follows:(1)Based on the theory of turbulent flow,a set of grill turbulent experimental device was designed,and the turbulent characteristics of the turbulent flow generated by it were studied in depth.Under the vibration stroke and vibration frequency,there is an approximately isotropic uniform turbulent flow area in a certain height range in the middle region of the adjacent twolayer grid.(2)On the basis of the isotropic turbulent flow generated by the grid turbulence device,the suspended sediment experiment is carried out.The 2mm slit shooting space independently designed on the side wall of the turbulent flow device is used to collect the suspended sediment particle images,and the clustering algorithm and the U-net algorithm are used.The two types of image recognition technologies identify sediment particles,and the respective recognition effects and advantages and disadvantages are compared and analyzed.The results show that the recognition effects of the two types of algorithms are relatively good.In order to reduce the amount of data and image calculations,the clustering algorithm is selected in this paper.(3)Based on the clustering algorithm,the images collected at different heights of the grid layer are recognized and analyzed,the particle size is calculated,and the vertical distribution characteristics are studied.The results show that the vertical particle size distribution at different heights of the grid layer is relatively uniform overall,and the particle size difference at different heights is small.The diameter has increased,and the particle size identification results are compared with the results measured by the laser particle size analyzer.The relative error is 5%10%.(4)On the basis of accurate particle size identification,the particle volume is calculated from the particle size,and then the sediment concentration at different heights is calculated to study its vertical distribution characteristics.The results show that the distribution of sediment concentration is closely related to the amount of sand input.When the amount of sand input is small,the sediment particles are less affected by gravity and the interaction between particles,and the vertical distribution of sediment concentration is more uniform;The sediment content in the water body increases,the turbulent diffusion effect is suppressed,and the gradient change in the vertical distribution of sediment concentration begins to appear.Continue to increase the amount of sediment input,and the vertical distribution of sediment content gradually stabilizes.The overall sediment content in the upper part is relatively small,the distribution characteristics of slightly larger sediment content at the bottom.The sand content distribution results obtained by image recognition are compared and analyzed with the results obtained by the theoretical formula of Rouse and Zhang Ruijin to calculate the sand content of each layer,and the sand content distribution characteristics obtained by the three calculation methods are consistent as a whole,and the error of the results obtained by image recognition and the results calculated by Rouss formula is smaller,and the relative error is about 1%-10%. |
PDF Full Text Request