用于边缘计算设备的果树挂果量轻量化估测模型  

A Lightweight Fruit Load Estimation Model for Edge Computing Equipment

作　　者：夏雪[1] 柴秀娟 张凝 周硕 孙琦鑫 孙坦 XIA Xue;CHAI Xiujuan;ZHANG Ning;ZHOU Shuo;SUN Qixin;SUN Tan(Agricultural Information Institute,Chinese Academy of Agricultural Sciences/Key Laboratory of Agricultural Big Data,Ministry of Agriculture and Rural Affairs,Beijing 100081,China;Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]中国农业科学院农业信息研究所/农业农村部农业大数据重点实验室,北京100081 [2]中国农业科学院,北京100081

出　　处：《智慧农业（中英文）》2023年第2期1-12,共12页Smart Agriculture

基　　金：国家重点研发计划项目(2022YFD2002205);中央级公益性科研院所基本科研业务费专项(Y2022QC17,JBYW-AII-2022-05,JBYW-AII-2022-26);中国农业科学院科技创新工程(CAAS-ASTIP-2021-AII-08,CAAS-ASTIP-2023-AII)。

摘　　要：[目的/意义]挂果量是果树栽培管理的重要指标。传统人力抽样估测果树挂果量的方法不仅耗时费力,而且容易产生较大误差。本研究提出一种用于边缘计算设备的轻量化模型,实现视频中树上柑橘挂果量的自动估测。[方法]该模型采用CSPDarkNet53+PAFPN结构作为特征提取网络,实现更快的推理速度和更低的模型复杂度,在果实跟踪过程中引入Byte算法改进FairMOT的数据关联策略,对视频中的柑橘进行预测跟踪,以提升挂果量估测准确性。[结果和讨论]在边缘计算设备NVIDIA Jetson AGX上进行模型性能测试结果表明,本研究所建模型对柑橘挂果量的平均估测精度(Average Estimating Precision,AEP)和处理速度(Frames Per Second,FPS)分别达到91.61%和14.76,模型估测值与人工测得真实值的决定系数R^(2)为0.9858,均方根误差(Root Mean Square Error,RMSE)为4.1713,模型参数量、计算量(Floating Point Operations,FLOPs)和模型大小分别为5.01 M、36.44 G和70.20 MB,展现出较对比模型更优的挂果量估测性能和更低的模型复杂度。[结论]试验结果证明了本研究所建模型在边缘计算设备上对柑橘挂果量估测的有效性,基于算法模型研发的果园挂果量远程监测系统可满足用于果园移动平台行进状态下的果树挂果量估测需求。本研究可为果园生产力自动监测分析提供技术支持。[Objective] The fruit load estimation of fruit tree is essential for horticulture management.Traditional estimation method by manual sampling is not only labor-intensive and time-consuming but also prone to errors.Most existing models can not apply to edge computing equipment with limited computing resources because of their high model complexity.This study aims to develop a lightweight model for edge computing equipment to estimate fruit load automatically in the orchard.[Methods] The experimental data were captured using the smartphone in the citrus orchard in Jiangnan district,Nanning city,Guangxi province.In the dataset,30 videos were randomly selected for model training and other 10 for testing.The general idea of the proposed algorithm was divided into two parts:Detecting fruits and extracting ReID features of fruits in each image from the video,then tracking fruit and estimating the fruit load.Specifically,the CSPDarknet53 network was used as the backbone of the model to achieve feature extraction as it consumes less hardware computing resources,which was suitable for edge computing equipment.The path aggregation feature pyramid network PAFPN was introduced as the neck part for the feature fusion via the jump connection between the low-level and high-level features.The fused features from the PAFPN were fed into two parallel branches.One was the fruit detection branch and another was the identity embedding branch.The fruit detection branch consisted of three prediction heads,each of which performed 3×3 convolution and 1×1 convolution on the feature map output by the PAFPN to predict the fruit's keypoint heat map,local offset and bounding box size,respectively.The identity embedding branch distinguished between different fruit identity features.In the fruit tracking stage,the byte mechanism from the ByteTrack algorithm was introduced to improve the data association of the FairMOT method,enhancing the performance of fruit load estimation in the video.The Byte algorithm considered both high-score and low-sc

关 键 词：精准园艺 挂果量估测 边缘计算 深度学习 多目标跟踪 轻量化模型 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] S225.93[自动化与计算机技术—控制科学与工程]