基于深度学习算法的水稻飞虱虫害图像测报系统设计

Design of deep learning algorithm based image detection andalarm system for rice planthopper pests

为了实现水稻飞虱虫害诊断,解决传统测报系统中存在的误报率高的问题,利用深度学习算法从硬件和软件两个方面对水稻飞虱虫害图像测报系统进行优化设计。测报硬件系统主要由主机、分机、传感器以及图像采集设备组成,并通过电源电路为硬件设备提供电力支持。在硬件设备安装完成的基础上,建立数据库为软件功能的实现提供基础数据,并利用深度学习算法通过采集图像预处理、识别水稻飞虱虫害和启动异常报警程序三个步骤实现图像测报功能。为了检测设计的基于深度学习算法的水稻飞虱虫害图像测报系统的性能,设计系统测试实验。经过与传统测报系统的对比可以发现,设计的虫害图像测报系统的误差率降低了0.19%,且平均时间消耗节省了0.64 s。

In order to diagnose the rice planthopper pests and reduce the high false alarm rate in the traditional detection and alarm system,an optimization design of image detection and alarm system for rice planthopper pests is carried out by means of deep learning algorithm in the aspects of hardware and software.The detection and alarm hardware system is mainly composed of host,extension set,sensor and image acquisition equipment.The power of the hardware equipment is supported by power supply circuit.On the basis of the hardware equipment,the database is established to provide basic data for the realization of software functions.The image detection and alarm function is realized with deep learning algorithm by three steps:preprocessing of the collected image,identification of rice planthopper pests and initiation of abnormal alarm program.Experiments were designed to test the performance of the deep learning algorithm based system for rice planthopper pest image detection.In comparison with the traditional detection and alarm system,it can be found that the error rate of the designed image detection and alarm system reduced by 0.19%,and the average test duration is shortened by 0.64 seconds.