On-line measurement of trash contents in cotton based on BP neural network

To improve the existing trash content, in cotton statistical models, which are inaccurate and imprecise in measurement as there are many factors influencing trash content in cotton, a model based on the BP neural network was constructed. A measurement system was designed to pick up image feature parameters of trash in cotton and to dispose these parameters. As for the slow convergence rate of BP algorithm, a momentum item was introduced into BP algorithm so that the convergence rate was increased. Experimental results show that the relative residual standard deviation of the fitted value of the BP neural network is 1.76% and the accuracy of its fitted value is much higher than that of other statistical models.