The classification and recycling of waste are significant for environmental protection and resource conservation. In China, the fine sorting of low-value recyclable waste (LVRW) primarily uses manual sorting, which is inefficient and harmful to health. Automated sorting uses sensors to identify the type of waste on a conveyor belt, which is then automatically separated by a sorting organisation to improve the recycling process. A low-cost, high-performance sorting system was built in this paper. The system used a low-resolution industrial camera to capture colour images of waste and a pneumatic separator to sort the waste. The recognition accuracy of waste was improved by optimising the recognition model and using the copy–paste data augmentation method. Mark pneumatic valve control algorithm is proposed that achieves more than 99% waste recovery even under dense working conditions. The sorting system is suitable for sorting plants and has been applied to a case study of LVRW in Xiamen city. In the sorting experiments under complex working conditions, the purity and recycling rate of plastic and Tetra Pak wastes achieved over 95%. The experiment demonstrates the great potential of the system for waste recycling and will help solve the problem of municipal waste in developing countries.