Abstract: The drum-screen winnowing machine for residual film and impurities is a key equipment for the resource utilization of residual films. Currently, manual cleaning is required after screen hole clogging, and issues such as high labor costs and low efficiency have hindered its further development. This study proposes a multi-region dual-column unclogging method and has established a multi-region dual-column unclogging system, achieving automated unclogging operations. First, through comparative experiments on YOLO series models, the YOLOv8s model was selected as the identification model. Pruning experiments were then conducted on YOLOv8s, determining a pruning rate of 40% followed by 200 epochs of fine-tuning. This process yielded the YOLOv8s-prune model. Finally, the identification performance of YOLOv8s-prune was compared with various other object detection algorithms. Results from the YOLO series model comparison experiments showed that the YOLOv8s model achieved optimal comprehensive performance on the validation set: precision (P) of 99.4%, recall (R) of 99.2%, mAP@0.5 of 99.5%, mAP@0.5-0.95 of 94.7%, and a model size of 21.4 MB. After fine-tuning, the YOLOv8s-prune model achieved a precision (P) of 99.3%, recall (R) of 99%, mAP@0.5 of 99.5%, mAP@0.5-0.95 of 92.1%, and a model size of 10.9 MB. Compared to YOLOv8s, these metrics decreased by 0.1, 0.2, 0, and 2.6 percentage points respectively, while the model size was reduced by 50%. Comparison results between YOLOv8s-prune and other object detection algorithms indicated that, relative to other algorithms, the YOLOv8s-prune model maintained high accuracy while significantly reducing the number of parameters, computational load, and model size. On the test set, it achieved a precision (P) of 98.9%, recall (R) of 99%, and mAP@0.5 of 99.4%. Finally, leveraging the coordinate information identified by YOLOv8s-prune, an identification control system was built using a microcontroller (STM32F103C8T6) as the controller. The identified coordinate information was divided into four intervals corresponding to four nozzles. The motion module traveled 360 mm per cycle to achieve queue-based positioning between the nozzles and screen holes. Positional accuracy compensation for the motion module was implemented based on the difference between the identified X-axis coordinate value of the leftmost screen hole and its actual coordinate value. Relevant performance tests showed that during sieving operations: The average recognition rate of the model was 98.3% within 0~10 min, 95.4% within 10~20 min, and 92.7% within 20~30 min. The average recognition rate for screen holes exceeded 94%. Queue positioning test results indicated that the absolute positioning deviation of the motion module ranged from a maximum of 25.6 mm to a minimum of 7 mm, with a qualification rate exceeding 94%, meeting the requirements for cyclic operations. Orthogonal experiments on the whole machine revealed that the order of influencing factors on the unblocking rate was: unblocking wind speed, operating distance, and unblocking angle. The theoretically optimal parameter combination was: unblocking wind speed of 4.959 m/s, unblocking angle of 11.14°, and operating distance of 125.373 mm, yielding a maximum unblocking rate of 93.535%. The rounded optimal parameter combination was: unblocking wind speed of 5 m/s, unblocking angle of 11°, and operating distance of 120 mm. Under these parameters, the average unblocking rate was 90.56%, with an error of 2.975 percentage points compared to the theoretical prediction, satisfying the operational requirements of the film residue winnower. The research findings provide a reference for the development of automatic unblocking systems for film residue winnowers and the setting of operational parameters.