白背飞虱智能识别系统的设计与实现

发布时间：2018-08-20 11:51

【摘要】：水稻是我国最主要的粮食作物,在农业生产和粮食安全中具有关键性的作用。白背飞虱[Sogatellafurcifera(Horvath)]是目前影响水稻高产、稳产的主要害虫之一,因此,必须对白背飞虱种群数量进行准确的监测和预测。根据植保专家和昆虫学家的意见,通过量化白背飞虱背部几何形态、颜色和纹理特征的方法,本文设计了一套基于图像处理技术的白背飞虱智能识别系统。为了获取清晰的白背飞虱昆虫图像,设计并搭建了野外昆虫图像采集装置。采集装置主要由采集装置机械平台、自动拍摄控制系统组成。采集装置机械平台由底座、采集工作台和传动系统组成。底座为1OOcm×100cm田字形框架,支撑整个实验平台;采集工作台大小为120cm×90cm,放置白色幕布,为白背飞虱等昆虫提供趴伏处;传动系统负责带动相机二维平面运动。自动拍摄控制系统主要由拍摄系统和运动控制系统组成。拍摄系统由相机、远心镜头、图像采集卡和RI环形冷光源组成,负责拍摄清晰的白背飞虱图像。运动动控制系统主要由伺服电机、伺服驱动器和PLC组成。通过PLC输出脉冲到伺服驱动器驱动伺服电机,并带动相机的X向和Y向的二维平面运动,且触发相机有序的扫描拍摄整个幕布,最终获取自然状态下的白背飞虱昆虫图片。用自制的野外昆虫图像采集装置,在野外环境下,采集131张白背飞虱昆虫图像。通过颜色(蓝色分量B=130)阈值分割、滤波处理后,获取昆虫图像的二值化图,然后提取出单个昆虫背部区域二值化图和背部区域灰度图。对白背飞虱的大小和颜色成分进行统计分析,剔去明显非白背飞虱的单个昆虫图像,再运用不变矩和二维傅里叶频谱数据描述昆虫几何形态、颜色和纹理共88个特征,将7个不变矩和l×l(l=1,2,...,9)个二维傅里叶频谱特征进行组合并作为输入变量,建立基于支持向量机的白背飞虱识别模型。实验结果显示白背飞虱样本的正确识别率达到95%,表明该方法可以实现田间白背飞虱的自动识别。
[Abstract]:Rice is the most important food crop in China and plays a key role in agricultural production and food security. Sogatella furcifera (Horvath) is one of the main pests affecting the high and stable yield of rice. Therefore, it is necessary to accurately monitor and predict the population of the white-backed planthopper. In order to obtain a clear image of the white-backed planthopper, a field insect image acquisition device is designed and built. The acquisition device is mainly composed of acquisition device machinery. The mechanical platform of the acquisition device is composed of a base, a collection table and a transmission system. The base is an OOcm *100cm field-shaped frame, supporting the entire experimental platform. The size of the acquisition table is 120 cm *90cm, and a white screen is placed to provide a dormant place for insects such as white-backed planthopper. The transmission system is responsible for driving the camera 2. The system consists of a camera, telecentric lens, image acquisition card and RI ring cold light source. It is responsible for taking clear pictures of the white-backed planthopper. The motion control system is mainly composed of a servo motor, a servo driver and a PLC. The pulse is output by PLC. The servo driver drives the servo motor and drives the camera to move in X and Y directions in two-dimensional plane, and triggers the camera to scan and photograph the whole curtain orderly, and finally obtains pictures of the white-backed planthopper insects in the natural state. The color (blue component B = 130) threshold segmentation is used to obtain the binary image of the insect image after filtering. Then the binary image of the back region and the gray image of the back region of the single insect are extracted. The size and color components of the white-backed planthopper are statistically analyzed, and the single insect image with obvious non-white-backed planthopper is removed. Then the invariant moments and two-dimensional image are used. Fourier spectral data describe the geometry of insects. The color and texture of insects are 88 features. Seven moment invariants and l *l (l=1,2,..., 9) two-dimensional Fourier spectral features are combined as input variables to establish a white-backed planthopper identification model based on support vector machine. This method can automatically identify white backed planthopper in the field.
【学位授予单位】：南京农业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;S435.112.3

【参考文献】