基于图像后处理与人工智能的温室病虫害防治方案

病虫害综合治理(IPM)是降低在温室中有害的化学物质使用量的重点策略，IPM策略包含早期的害虫检测与后续的种群监控，IPM不仅耗时而且高度依赖人工判定，容易产生错误。提出一个基于图像处理算法与人工神经网络检测与监控温室中粉虱与蓟马的方法。首先，通过图像采集系统获得粘虫板的数字图像；然后，对于每个检测目标使用图像的目标检测、分割、形态学分析与颜色特征估算等方法进行处理；最终，通过前向多层神经网络算法将目标分类处理。本系统的粉虱识别精度高达0.96，召回率为0.95，蓟马的识别精度为0.92，召回率0.96。

Image post-process and artificial intelligence based pest control approach of the greenhouse

IPM(Integrated Pest Management) is a important strategy to reduce the usage amount of harmful chemicals of the greenhouse, IPM consists of the early pest detection and the continuous swarm monitoring, IPM is not only high time cost but also highly dependent on the manual decision-making, it performs much errors. A whitefly and thrips detecting and monitoring method based on the image process and artificial neural network is proposed. Firstly, the digital images of sticky traps are collected by image acquiring system; then, the target detection, segmentation, morphometric analysis and color feature estimation methods of the image process are used to process each pest target; lastly, forward multilayer neural network algorithm is adopted to classify the targets. The proposed system realizes a recognition accuracy of 0.96 and recall rate 0.95 for whitefly, and a recognition accuracy of 0.95 and recall rate 0.92 for thrips.