Research On Sorting System Of Waste Plastic Beverage Bottles Based On Deep Learning

This paper focused on the color and position recognition of waste plastic beverage bottles in the whole bottle visual sorting process,as well as the pneumatic spray separation of waste plastic beverage bottles.Based on the theory of deep learning classification,the color recognition of waste plastic beverage bottles was studied.The classification data set of waste plastic beverage bottles was collected and built,and then three classic deep learning networks(Alex Net,VGGNet and Darknet-53)were trained and tested on the classification data set.In this way,a Bottle Net classification network model suitable for sorting waste plastic bottles was proposed as well as studying the adaptability of the classification network model under different working conditions(simultaneous interpreting of different belt colors and light intensity).On the basis of the deep learning objective detection method,the location recognition problem of waste plastic beverage bottles was studied.First,the detection data set of waste plastic beverage bottles was collected and made.And then,the Bottle-YOLOv3 location network model,which was suitable for the location recognition of waste plastic beverage bottles,was designed and constructed.Aiming at the problem of signal repetition in continuous multi frame images,an algorithm of avoiding omission and position information repetition was proposed in the process of sorting waste plastic beverage bottles.By using the fluid simulation software FLUENT and the orthogonal test analysis method,the effects of the type,total length,contraction angle and outlet diameter of the pneumatic nozzle on the jet performance were simulated,and the structure of the nozzle were optimized.Using the fluid simulation software FLUENT to simulate the impact load of the jetted waste plastic beverage bottle under the jet air flow,the flight trajectory model of the waste plastic beverage bottle in the external flow field was constructed.Based on the model,the flight trajectory of the waste plastic beverage bottle under different working conditions(working pressure,injection angle,nozzle layout and location)and nozzle structure were studied.The working condition parameters and nozzle structure parameters were optimized.Building the experiment platform was accomplished,and the feasibility of the color separation scheme of waste plastic beverage bottles was verified by experiments.