渔业现代化2025，Vol.52Issue(1)：99-109,11.DOI:10.3969/j.issn.1007-9580.2025.01.010
基于FasterYOLOv9-Slim的轻量级工厂化养殖鱼群识别
Lightweight factory-farmed fish identification based on FasterYOLOv9-Slim
摘要
Abstract
For factory farming environments with limited computing resources,enhancing identification speed and reducing model size while maintaining high accuracy is of paramount importance.Traditional object recognition algorithms such as YOLO,despite their superior performance,require substantial computational power due to their large model parameters and complex computations,thereby posing a significant bottleneck in settings unequipped for large-scale deep learning models where real-time monitoring is critical.Lightweight models provide potential solutions but often at the expense of reduced accuracy,which is unacceptable in factory farming as it directly impacts operational effectiveness.Therefore,there is an imperative need for methods that can reduce complexity without adversely affecting recognition performance.We propose FasterYOLOv9-Slim,a lightweight fish school identification model designed for efficient operation,compact size,and high precision through targeted enhancements.Specifically,the FasterYOLOv9-Slim model,based on improvements from YOLOv9 and FasterNet.First,based on the YOLOv9 model,FasterNet is introduced as a lightweight backbone network to replace the complex backbone structure used in the original model,effectively reducing the large number of parameters and computing requirements brought by the traditional convolution operation of the YOLOv9 backbone network.Secondly,HDPrune technology was used to prune two high-dimensional detector heads in the YOLOv9 model head network to reduce the network depth,reduce the calculation amount,and effectively reduce the accumulation of interference information in the network.Finally,based on partial convolution(PConv),the original feature fusion module RepNCSPELAN4 is improved to obtain a lighter and more efficient version of FasterRepNCSPELAN4.Combined with advanced downsampling modules ADown and DownSimper,the neck network structure of the model is redesigned.A more efficient feature fusion framework DFA-Neck was constructed.At the same time,it enhances the ability of the network to express features on different scales,and realizes the efficient coordination among feature extraction,fusion,information transfer and detection head output.To fully verify the effectiveness of the improvement,ablation and comparison tests are designed.In the ablation test,the results show that FasterNet effectively reduces the number of parameters in the model,proving its significant advantage in reducing the consumption of computational resources,while HDPrune plays an important role in weakening the accumulation of interfering information in the network,DFA-Neck successfully coordinates the functions of FasterNet and HDPrune in the overall network,ensuring the model's high efficiency in the process of feature extraction and information transfer.In the comparison test,the model was compared in detail with the advanced identification models of the same size in the YOLOv7,YOLOv8 and YOLOv10 series models in terms of performance,and the results showed that FasterYOLOv9-Slim achieved 34.14%,64.02%and 22.22%significant reductions,and demonstrates excellent overall performance in terms of model size,inference speed,and recognition accuracy in comparison with state-of-the-art lightweight networks such as ShuffleNet,MobileNet,and RepViT.The study verifies the effectiveness of the FasterYOLOv9-Slim model in balancing accuracy and speed in the fish identification task under factory farming conditions with limited computational resources,and also provides valuable experience and guidance for the design and optimization of the model in similar application scenarios in the future.关键词
养殖鱼群/YOLOv9/目标识别/模型剪枝/轻量化Key words
fish farming/YOLOv9/object recognition/model pruning/lightweighting分类
农业科技引用本文复制引用
张鑫,于红,吴子健,程志澳,高陈成,杨宗轶,王悦..基于FasterYOLOv9-Slim的轻量级工厂化养殖鱼群识别[J].渔业现代化,2025,52(1):99-109,11.基金项目
辽宁省重点科技项目(2023JH26/10200015) （2023JH26/10200015）
国家自然科学基金(62406052) （62406052）
辽宁省自然科学基金计划博士科研启动项目(2024-BS-214) （2024-BS-214）
辽宁省教育厅基本科研项目(LJ212410158022) （LJ212410158022）
辽宁省属本科高校基本科研业务费专项资金资助(2024JBQNZ011) （2024JBQNZ011）
