基于STM32的农业物联网病虫害图像识别算法研究

doi:10.11959/j.issn.2096-3750.2023.00365

摘要/Abstract

摘要：

在现代农业物联网系统中，边缘计算是不可或缺的组成部分。在此背景下，可将轻量级病虫害图像识别任务置于边缘设备上，然而受限于设备计算和存储能力，该任务面临着不小的挑战。为了解决这些问题，提出了一种以经济实用的STM32为边缘设备进行病虫害图像识别的方法。该方法针对STM32的特点，基于MobileNetv2结构做出改进，并应用量化感知训练技术对神经网络模型进行压缩，提高了模型的可移植性。同时，模型使用X-CUBE-AI部署并进行了性能评估。实验结果表明，改进模型不仅保证了图像分类准确率，而且相较于其他轻量级神经网络，该模型对STM32的Flash与RAM资源的占用有所减小。

关键词: 农业物联网, 边缘计算, 病虫害识别, STM32

Abstract:

In modern agriculture IoT systems, edge computing is an indispensable component.In this context, it is feasible to deploy lightweight pest and disease image recognition tasks on edge devices.However, due to the constraints of device computation and storage capabilities, this task faces significant challenges.To address these challenges, an economically practical method was proposed for pest and disease image recognition on STM32 edge devices.Specifically, the MobileNetv2 structure was improved to better suit the characteristics of STM32, quantization-aware training technique was used to compresses the network, model portability was enhanced.Meanwhile, the X-CUBE-AI was used to arrange the model and evaluate the performance.Experimental results demonstrate that the proposed model not only ensures image classification accuracy but also reduces the Flash and RAM resource consumption on STM32 compared to other lightweight networks.

Key words: agricultural IoT, edge computing, pest and disease recognition, STM32

中图分类号:

TP389.1

许柏涛, 陈翔. 基于STM32的农业物联网病虫害图像识别算法研究[J]. 物联网学报, 2023, 7(4): 132-141.

Botao XU, Xiang CHEN. Research on agricultural IoT pest and disease image recognition algorithm based on STM32[J]. Chinese Journal on Internet of Things, 2023, 7(4): 132-141.

图/表 12

表1

倒置残差块结构"

输入	运行	输出
h×w×k	1×1 Conv2d, ReLU6	h×w×tk
h×w×tk	3×3 dwise, ReLU6	$\frac{h}{s} \times \frac{w}{s} \times t k$
$\frac{h}{s} \times \frac{w}{s} \times t k$	1×1 Conv2d	$\frac{h}{s} \times \frac{w}{s} \times k^{'}$

表1

图1

表2

图2

图3

图4

表3

表4

图5

图6

表5

表6

参考文献 30

[1]	聂鹏程, 张慧, 耿洪良 ,等. 农业物联网技术现状与发展趋势[J]. 浙江大学学报(农业与生命科学版), 2021,47(2): 135-146.
	NIE P C , ZHANG H , GENG H L ,et al. Current situation and development trend of agricultural internet of things technology[J]. Journal of Zhejiang University (Agriculture ＆ Life Sciences), 2021,47(2): 135-146.
[2]	TZOUNIS A , KATSOULALS N , BARTZANAS T . Internet of things in agriculture,recent advances and future challenges[J]. Biosystems Engineering, 2017(164): 31-48.
[3]	陆林峰, 管孝锋, 黄海龙 ,等. 基于农业物联网的应用平台构建[J]. 浙江农业科学, 2020,61(7): 1455-1457.
	LU L F , GUAN X F , HUANG H L ,et al. Construction of application platform based on agricultural internet of things[J]. Journal of Zhejiang Agricultural Sciences, 2020,61(7): 1455-1457.
[4]	ZHOU Y Q , TIAN L , LIU L ,et al. Fog computing enabled future mobile communication networks:a convergence of communication and computing[J]. IEEE Communications Magazine, 2019,57(5): 20-27.
[5]	ZHOU Y Q , LIU L , WANG L ,et al. Service-aware 6G:an intelligent and open network based on the convergence of communication,computing and caching[J]. Digital Communications and Networks, 2020,6(3): 253-260.
[6]	ZHANG X H , CAO Z Y , DONG W B . Overview of edge computing in the agricultural internet of things:key technologies,applications,challenges[J]. IEEE Access, 2020(8): 141748-141761.
[7]	SUNYAEV A . Cloud computing[M]// Internet computing. Cham: Springer International Publishing, 2020.
[8]	DILLON T , WU C , CHANG E . Cloud computing:issues and challenges[C]// Proceedings of 2010 24th IEEE International Conference on Advanced Information Networking and Applications. Piscataway:IEEE Press, 2010: 27-33.
[9]	ALAM T . Cloud computing and its role in the information technology[J]. IAIC Transactions on Sustainable Digital Innovation (ITSDI), 2020,1(2): 108-115.
[10]	杨英茹, 吴华瑞, 张燕 ,等. 基于复杂环境的番茄叶部图像病虫害识别[J]. 中国农机化学报, 2021,42(9): 177-186.
	YANG Y R , WU H R , ZHANG Y ,et al. Tomato disease recognition using leaf image based on complex environment[J]. Journal of Chinese Agricultural Mechanization, 2021,42(9): 177-186.
[11]	赵立新, 侯发东, 吕正超 ,等. 基于迁移学习的棉花叶部病虫害图像识别[J]. 农业工程学报, 2020,36(7): 184-191.
	ZHAO L X , HOU F D , LYU Z C ,et al. Image recognition of cotton leaf diseases and pests based on transfer learning[J]. Transactions of the Chinese Society of Agricultural Engineering, 2020,36(7): 184-191.
[12]	钟林忆, 刘海峰, 董力中 ,等. 计算机视觉下的农作物病虫害图像识别研究[J]. 现代农业装备, 2021,42(1): 51-55.
	ZHONG L Y , LIU H F , DONG L Z ,et al. Image recognition of crop diseases and insect pests based on computer vision[J]. Modern Agricultural Equipments, 2021,42(1): 51-55.
[13]	季力 . 基于 STM32 芯片的电参数测量与数据传输[J]. 自动化与仪器仪表, 2010(3): 137-139.
	JI L . Power measurement and data transmission based on STM32 chip[J]. Automation ＆ Instrumentation, 2010(3): 137-139.
[14]	JACKO P , BERE? M , KOVá?OVá I ,et al. Remote IoT education laboratory for microcontrollers based on the STM32 chips[J]. Sensors (Basel,Switzerland), 2022,22(4): 1440.
[15]	SANDLER M , HOWARD A , ZHU M L ,et al. MobileNetv2:inverted residuals and linear bottlenecks[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 4510-4520.
[16]	OSMAN A , ABID U , GEMMA L ,et al. TinyML platforms benchmarking[C]// International Conference on Applications in Electronics Pervading Industry,Environment and Society. Cham:Springer, 2022: 139-148.
[17]	CHEN Y P , DAI X Y , CHEN D D ,et al. Mobile-former:bridging MobileNet and transformer[C]// Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2022: 5260-5269.
[18]	QIN Z , ZHANG Z N , CHEN X T ,et al. FD-MobileNet:improved mobilenet with a fast downsampling strategy[C]// Proceedings of 2018 25th IEEE International Conference on Image Processing (ICIP). Piscataway:IEEE Press, 2018: 1363-1367.
[19]	HE K M , ZHANG X Y , REN S Q ,et al. Deep residual learning for image recognition[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2016: 770-778.
[20]	HAN K , WANG Y H , TIAN Q ,et al. GhostNet:more features from cheap operations[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 1577-1586.
[21]	YANG J W , SHEN X , XING J ,et al. Quantization networks[C]// Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 7300-7308.
[22]	LIU Z , WANG Y , HAN K ,et al. Post-training quantization for vision transformer[J]. Advances in Neural Information Processing Systems, 2021(34): 28092-28103.
[23]	JACOB B , KLIGYS S , CHEN B ,et al. Quantization and training of neural networks for efficient integer-arithmetic-only inference[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 2704-2713.
[24]	周敏敏 . 基于迁移学习的苹果叶面病害 Android 检测系统研究[D]. 杨凌:西北农林科技大学, 2019.
	ZHOU M M . Research on android detection system of apple leaf diseases based on transfer learning[D]. Yangling:Northwest A ＆ F University, 2019.
[25]	JORDAN A A , PEGATOQUET A , CASTAGNETTI A ,et al. Deep learning for eye blink detection implemented at the edge[J]. IEEE Embedded Systems Letters, 2021,13(3): 130-133.
[26]	FALBO V , APICELLA T , AURIOSO D ,et al. Analyzing machine learning on mainstream microcontrollers[C]// International Conference on Applications in Electronics Pervading Industry,Environment and Society. Cham:Springer, 2020: 103-108.
[27]	ALONGI F , GHIELMETTI N , PAU D ,et al. Tiny neural networks for environmental predictions:an integrated approach with miosix[C]// Proceedings of 2020 IEEE International Conference on Smart Computing (SMARTCOMP). Piscataway:IEEE Press, 2020: 350-355.
[28]	SAILESH M , SELVAKUMAR K , PRASANTH N . A novel framework for deployment of CNN models using post-training quantization on microcontroller[J]. Microprocessors and Microsystems, 2022(94): 104634.
[29]	MERENDA M , PORCARO C , DELLA CORTE F G . LED junction temperature prediction using machine learning techniques[C]// Proceedings of 2020 IEEE 20th Mediterranean Electrotechnical Conference (MELECON). Piscataway:IEEE Press, 2020: 207-211.
[30]	CAPOTONDI A , RUSCI M , FARISELLI M ,et al. CMix-NN:mixed low-precision CNN library for memory-constrained edge devices[J]. IEEE Transactions on Circuits and Systems II:Express Briefs, 2020,67(5): 871-875.

输入	运行	t	c	n	s
160²×3	Conv2d	-	32	1	2
80²×32	STM32-block	-	16	1	1
80²×16	bottleneck	6	24	2	2
40²×24	STM32-block	-	32	2	1
40²×32	bottleneck	6	64	2	2
20²×64	STM32-block	-	96	3	1
20²×96	bottleneck	6	160	2	2
10²×160	STM32-block	-	320	1	1
10²×320	Avgpool	-	-	1	-
1×1×320	Conv2d 1×1	-	1 280	1	1
1×1×1 280	Conv2d 1×1	-	k

类别	数量/张
Alternaria_Boltch	5 343
Brown_Spot	5 655
Grey_Spot	4 810
Mosaic	4 875
Rust	5 694

类别	数量/张
Early_Blight_Fungus	792
Healthy	1 381
Late_Blight_Water_Mold	1 513
Leaf_Mold_Fungus	755
Powdery_Mildew	1 469
Septoria_Leaf_Spot_Fungus	1 403
Spider_Mite_Damage	929

网络模型	准确率（C model）	准确率（Python model）	Flash/KB	RAM/KB	参数量	CPU推理时间/s
STM32-MobileNet（×0.35）	90.69%	91.89%	256.52	137.26	139 038	512.745
FD_MobileNet（×0.35）	86.39%	88.72%	324.49	96.12	251 429	300.733
MobileNet（×0.35）	90.71%	93.67%	425.40	234.63	319 232	978.16
MobileNetv2（×0.35）	90.67%	92.44%	498.41	165.20	420 667	661.169

网络模型	准确率（C model）	准确率（Python model）	Flash/KB	RAM/KB	参数量	CPU推理时间/s
STM32-MobileNet（×0.35）	87.28%	87.40%	257.84	137.37	139 919	158.373
FD_mobilenet（×0.35）	80.01%	84.37%	325.20	96.12	252 147	98.934
MobileNetv2（×0.35）	87.52%	89.76%	436.23	237.25	320 134	216.960