Online Continuous Sorting for Plumule‐Removed Lotus Seeds Using Micro‐Pressure‐Airflow Excitation and 1D ‐ CNN Acoustic Recognition

ABSTRACT

Lotus seeds are high‐value food and medicinal materials. The plumule removal process of lotus seeds cannot avoid the drill deviation. The resulting defective products are difficult to discharge the inner lotus plumule and form residues, reducing the quality of the drilled lotus seeds. Based on the positional differences in drilled holes between qualified and defective products, an online continuous sorting method for drilled lotus seeds using micro‐pressure‐airflow excitation and acoustic recognition is proposed in this study. The method is to make the lotus seed pass through a micro‐pressure air curtain continuously. The airflow readily passes through the drilled hole on qualified products but has difficulty penetrating the hole on defective products, thereby generating differential acoustic signals. The signals are classified using an acoustic classification model, and then sorting is achieved. An online continuous sorting system for drilled lotus seeds was designed. The effect of the air pressure on the posture stability of lotus seeds when they pass through the air curtain was studied. A 1D‐CNN (one‐dimensional convolutional neural network) online acoustic classification model was designed for classifying drilled lotus seeds during continuous sorting. The model's classification accuracy, inference time, and parameter count were 96.38%, 18.8 ms, and 0.39 M (Million), respectively. The sorting test results show that the optimal conveying chain speed and the achieved sorting accuracy of the sorting system were 120 links/min and 93.52%, respectively. The research findings laid a theoretical basis for developing efficient and high‐quality engineering sorting/grading equipment for drilled lotus seeds.