A Study Of Dynamic Weighing For Tobacco Leaves Based On IWPA-RELM Algorithm

The weight of tobacco leaves is closely related to their length and identity.When using the weight of tobacco leaves as one of the factors for intelligent grouping and grading,the static weighing efficiency is low,while dynamic weighing is more convenient.Many factors affect the dynamic weighing,not only environmental factors but also the non-linear relationship between input and output of weighing sensors.The solution to the non-linearity problem in dynamic weighing mainly focuses on the construction of mathematical models and the combination of sensors.There is less research on the use of neural networks for dynamic weighing of tobacco leaves,which cannot solve the problem of low weighing accuracy caused by non-linear factors in dynamic weighing well.The research in this thesis is based on tobacco leaves,focusing on the non-linear problem in dynamic weighing.The regularized extreme learning machine(RELM)model was optimized using multi-strategy improved wolf pack algorithm(IWPA).The IWPA-RELM tobacco leaf dynamic weighing system was implemented for nonlinear compensation.Firstly,the experimental data was preprocessed using a dynamic weighing sensor and electronic balance to obtain the dynamic and static weight data of tobacco leaves.The length and width of the leaves were obtained by grayscale image processing and Minimum Bounding Rectangle methods.The color of tobacco leaves was identified using the HSV color space,and the data was normalized.Secondly,the IWPA-RELM model was established.To enhance the optimization ability of the Wolf Pack Algorithm(WPA),Tent chaotic mapping was used to initialize the population to improve diversity.To improve the global search capability of the wolf pack during roaming behavior,individual random movement strategy from the Pelican Optimization Algorithm was incorporated.To balance the global optimization and local exploration of the pack hunting behavior,an inertia weighting strategy was proposed to improve the hunting step.The IWPA was compared with WPA and COOT algorithms on five test functions.Finally,the improved algorithm was used to optimize the initial weights and thresholds of the RELM network to achieve tobacco leaf dynamic weighing prediction.The IWPA-RELM,COOT-RELM,and RELM dynamic weighing models were compared using four evaluation indicators.The results showed that for tobacco leaves with an absolute error of 2g,the average absolute error between the static weight and IWPA-RELM predictive weight was 0.35 g,indicating the effectiveness of the improved model.The GBTD algorithm was also used to classify tobacco leaf parts,and when combining the tobacco weight factor,accuracy increased by 8.1%.For practical application,a prototype system for the dynamic weighing of tobacco leaves was developed based on the IWPA-RELM model,providing a certain reference for the practical application of dynamic weighing of tobacco leaves.