| 摘要: |
| 【目的】 虫害是影响荔枝产量与品质的重要制约因素，基于深度学习的荔枝虫害识别可以为荔枝种植过程中的虫害防治工作提供技术支持，对提高荔枝产量及品质，提高果园生态安全具有重要作用。【方法】 文章针对目前荔枝虫害识别领域存在的问题，为提高虫害目标识别精度和效率，以荔枝蝽象为目标虫害，提出一种基于YOLO v4的目标检测方法，首先使用专业摄像头、大型数据库、智能虫情测报灯3种方式采集荔枝虫害图像，配合数据增强方法，用LableImg平台进行数据标注，制作一个特征丰富的数据集，在CSP Dark net框架下进行网络模型训练，得到荔枝虫害识别模型。【结果】 基于深度学习的荔枝虫害识别技术在广州从化荔枝现代农业产业园进行应用，取得了较好的应用效果，证明该技术可以实现真实复杂环境中荔枝虫害的有效识别。【结论】 基于深度学习的荔枝虫害识别模型，能够实现虫害的科学监测，降低农户对于虫害的投入成本，减少化学农药的使用，改善荔枝生长的环境，进一步实现荔枝生产绿色化要求，增加作物的经济价值。 |
| 关键词: 虫害 目标检测 深度学习 YOLO v4网络 |
| DOI：10.12105/j.issn.1672-0423.20220404 |
| 分类号: |
| 基金项目:广东省2021年度普通高校重点科研平台和项目“真实复杂场景下的荔枝种植人工智能关键技术研究”（2021ZDZX4095）；广东轻工职业技术学院2020年校级重点项目“真实复杂场景下的荔枝病虫害人工智能关键技术研究”（KJ2020-002）；广东轻工职业技术学院2022年校级科研项目“基于区块链共识算法改进的果蔬产品溯源研究与应用”（KJ2022-08） |
|
| Application and realization of litchi pest identification technology based on deep learning |
|
Ye Yun1, Zhao Xiaojuan1, Jiang Sheng2
|
|
1.School of Chemical Engineering and Technology，Guangdong Industry Polytechnic，Guangzhou，Guangdong 510300，China;2.College of Electronic Engineering，South China Agricultural University，Guangzhou，Guangdong 510630，China
|
| Abstract: |
| [Purpose] Insect pest is an important restrictive factor affecting the yield and quality of litchi. The identification of lychee pest based on deep learning can provide technical support for pest control in the process of litchi planting，and play an important role in improving the yield and quality of litchi and improving the ecological security of orchards.[Method] Aiming at the problems existing in the field of litchi pest identification at present，in order to improve the accuracy and efficiency of pest target recognition，the target insect pest is Litchi pentatomid，this paper proposes a target detection method based on YOLO v4. Firstly，litchi pest images are collected by using professional camera，large database and Intelligent insect monitoring system monitoring lamp. Combined with the data enhancement method，use the LableImg platform for data labeling to create a feature-rich data set，train the network model under the CSP Dark net framework，and obtain the litchi pest identification model.[Result] The litchi pest identification technology based on deep learning has been applied in Guangzhou Conghua litchi modern agricultural industrial park and achieved good application results. The actual measurement shows that the technology can effectively identify litchi pests in real complex environment.[Conclusion] The development of intelligent pest monitoring lamp based on the deep learning litchi pest identification model can help farmers realize the scientific monitoring of pests，reduce the input cost of farmers for pests，reduce the use of chemical pesticides，improve the environment for litchi growth，further realize the green requirements of litchi production，and increase the economic value of crops. |
| Key words: insect pest target detection deep learning YOLO v4 network |
