For different colors of recyclable beverage bottles with different recycling values and the need to solve the color identification sorting problem,a lightweight model based on YOLOv5s and with DELTA parallel robot sorting equipment is proposed for intelligent identification sorting.This model reduces thenumber of C3 in the original Backbone and uses 1×1 convolution kernel instead of C3 and 3×3 convolution kernel in the Conv module part.GhostConv is used instead of traditional Conv.CIOU Loss increases the Loss of detection box scale,length and width to improve the rectangular box regression effect.CIOU Loss is chosen as the bounding box loss function,and the effectiveness of this model is verified by comparing experiments with other traditional models.The results show that the number of parameters and computation volume are reduced by 33.80%and 36.84%,respectively,compared with the original model,and the recognition time for the color of recycled beverage bottles reaches 0.008 s,which can recognize 125 images per second,and the recognition accuracy reaches 97%.Compared with the traditional model,the improved YOLOv5s model has higher recognition accuracy and faster recognition speed.