面向田园监测的低成本多光谱图像远程采集节点设计

利用多光谱图像采集设备获取农作物敏感波段的多光谱图像并对其进行分析,已经成为当前农作物生长状态评估的重要手段。但现有的光谱图像采集设备存在价格昂贵、体积较大、分辨率低、需人工搬运设备至田间作业等不足。为此,该文设计一种低成本、高分辨率的多光谱图像远程采集节点,并部署在田园对农作物光谱图像进行实时采集,并远程传输到服务器端,从而实现对农作物状态的持续监测。节点可同时采集RGB彩色图像和3种不同波段光谱图像,图像分辨率可达500万像素,且滤片可更换,因而其可应用于不同作物的图像获取。节点的硬件平台由嵌入式处理器芯片S3C6410、CMOS图像传感器OV5640和自行设计的机械式滤光片转换装置构成。为了实现图像的远程传输,节点集成了3G无线通信模块。采用嵌入式Linux搭建了节点的软件平台,并在此基础上设计了节点的软件系统,实现了滤光片转换的驱动控制、图像采集、图像压缩和3G无线传输等功能。为了验证节点的性能,将节点部署在田园进行了一系列试验测试。测试结果表明,节点能够自动采集RGB彩色和3种波段光谱图像,并通过3G网络无线传输到服务器端。采集和传输大小为133 k B的4幅图像所需总时间约为45.27 s,丢包率为0.54%。试验表明该文所设计的节点能够实时采集和远程传输多光谱图像,可满足田园现场环境多光谱图像获取的需求。

Design of remote acquisition node of low-cost multispectral image for field monitoring

Analyzing the multispectral images of sensitive bands of crops obtained through multispectral image acquisition devices, has currently become an important means to assess the growth status of crops.However, the existing multispectral image acquisition devices have some shortcomings, such as expensive in price, huge in volume, low in resolution.Further, it is laborintensive and timeconsumed to carry the device to reality environment.To this end, a low cost and high resolution multispectral image remote acquisition node is designed in this paper, and it is deployed in a field to capture multispectral images of crops in real time which are remotely transmitted to the server in order to achieve continuous monitoring for the growth status of crops.The designed node can simultaneously capture RGB images and spectral images in 3 different bands with the resolution of up to 5 mega pixels, and its filters are replaceable.Thus, the node can be deployed in different environments to capture images of different crops.The system control module of the node is constructed by integrating a lowcost highperformance S3C6410 chip produced by Samsung Electronics Co., Ltd, and a peripheral circuit designed by ourselves.The multispectral image acquisition module of the node is constructed by combining a CMOS image sensor OV5640 with a mechanical filters switching device designed by ourselves.To transmit the images remotely, a third generation (3G) module is integrated into the node, which communicates through 3G wireless networks widely used in China.S3C6410 is a low cost chip, OV5640 as a CMOS sensor is much cheaper than CCD sensors, and the mechanical filters switching method is much cheaper than others, so the cost of the node is inexpensive.To ensure the stability and reliability of the node, the powerful embedded Linux operating system is employed as the software development platform.And based on this platform, a modular designing method is adopted to program the software system of the node, realizing the drive control of the filter switching device, capturing images periodically, compressing images and transmitting images through 3G networks in C/C++ language.In order to verify the performance of the node, the node is deployed in a farmland for a series of experiments.To ensure the node works stably in the field, where the climate is changeable and the infrastructure is absent, the node is encapsulated in a waterproof spherical shield, and a solar panel plus rechargeable batteries are used to supply power for it.The experimental results indicate that the node can automatically capture RGB images and images of 3 different bands of 550,0 and 800 nm, and transmit them to the server through 3G wireless communication networks.The total time consumed to capture and transmit 4 images with a size of 133 kB is about 45.27 s.Aiming at the package loss problem arising in experiments, an image automatic retransmission mechanism is designed based on selective repeat ARQ (Automatic Repeat reQuest) protocol, which makes the package loss rate dropped to 0.54%, realizing the reliable transmission of images.The experiments validate the node designed in this work can capture multispectral images in real time and transmit them remotely, and satisfy the requirements of multispectral image acquisition in real agricultural field environments.