Research on the Intelligent Detection Algorithm of Apple Quality Based on Deep Learning

目的：针对传统苹果品质检测方法效率低、主观性强的问题，研究基于深度学习的苹果品质智能检测算法，实现苹果外观缺陷、成熟度和品质等级的自动化精准识别。方法：构建包含15,000张高分辨率苹果图像的大规模数据集，涵盖红富士、嘎啦、黄元帅等6个主要品种；基于YOLOv8网络架构，引入多尺度特征融合模块(MSFM)、卷积块注意力机制(CBAM)和改进的Focal Loss损失函数，设计面向苹果品质检测的深度学习模型；采用迁移学习和数据增强策略优化模型性能。结果：改进算法在苹果品质检测任务上达到96.8%的准确率，精确率95.9%，召回率96.2%，F1分数96.0%，mAP值93.7%，处理速度45.1帧/秒；相比基线YOLOv8算法，各项指标分别提升4.5%、4.3%、5.4%、4.8%、5.8%和6.9帧/秒；在缺陷检测方面，对表面斑点、破损、变色等缺陷的检测准确率均超过94%。结论：所提出的改进深度学习算法能够高效准确地实现苹果品质自动化检测，为现代果品加工业提供了有效的技术解决方案。Objective: Aiming at the low efficiency and strong subjectivity of traditional apple quality detection methods, an intelligent apple quality detection algorithm based on deep learning was studied to achieve automatic and accurate identification of apple appearance defects, maturity and quality grade. Method: A large-scale dataset of 15,000 high-resolution apple images was constructed, covering six major varieties such as Red Fuji, Gala, and Huang Yuanshuai. Based on the YOLOv8 network architecture, a multi-scale feature fusion module (MSFM), convolutional block attention mechanism (CBAM) and improved focal loss function were introduced to design a deep learning model for apple quality detection. Transfer learning and data enhancement strategies were used to optimize the model performance. Results: The improved algorithm achieved an accuracy of 96.8%, precision of 95.9%, recall of 96.2%, F1 score of 96.0%, mAP value of 93.7%, and processing speed of 45.1 frames/second in the apple quality detection task; compared with the baseline YOLOv8 algorithm, each indicator was improved by 4.5%, 4.3%, 5.4%, 4.8%, 5.8% and 6.9 frames/second respectively; in terms of defect detection, the detection accuracy of surface spots, damage, discoloration and other defects exceeded 94%. Conclusion: The proposed improved deep learning algorithm can efficiently and accurately realize the automatic detection of apple quality, and provide an effective technical solution for the modern fruit processing industry.