视觉感知的水稻株间机械除草与同步液肥环施机理研究

According to the environmental characteristics of paddy field and agronomic requirements of the weeding and fertilizing operation, this study proposes a new operating method, which combines root-surrounding fertilizing with intra-row mechanical weeding. It develops a mechanism of synchronous operation integrating root-surrounding fertilizing in the intra-row mechanical weeding as well. This way, it reduces the consumption of herbicide and improves the utilization rate of fertilizer. To be concrete, visual recognition model of rice plants is established using deep learning algorithm according to the plant-type characters under Side-View image perspective. Accurate positioning of rice plants is implemented by bidirectional projection, which solves technical difficulties of identification and location caused by the overlapping of neighbor crops due to the small short distance. Coupling mechanism among weeding component, fertilizer-water jet and paddy field soil is analyzed using explicit dynamics theory by means of ANSYS platform, which can explain rules of weeds damage, soil deformation and multiphase medium coupling. Multidisciplinary coupling model of synchronous operation is simulated with Modelica on MWorks platform, which involves multi-system of machinery, electricity, hydraulic, pneumatic and controlling. It helps to ascertain the main affecting causes of state switching performance and provide theoretical and design basis for development of new type of synchronous operation mechanism. Not only can this study solve technical problem of synchronous root-surrounding fertilizing with mechanical intra-row weeding, but also provide reference for precise management of other crops.

为了减少除草剂的施用量，提高化肥利用率，本项目根据水田环境特点、除草与追肥作业的农艺要求，提出了株间机械除草与环株同步施肥的作业方式，并设计视觉感知的新型除草施肥同步作业机构。在研究方法上，根据Side-View图像视角下的先验株型特点，建立基于深度学习的稻株视觉识别模型，利用双向投影法实现稻株的准确定位，解决小株距叶冠遮叠作物识别和定位的技术难题；应用显式动力学理论在ANSYS平台上定量描述株间除草部件及肥料射流与水田土壤互馈耦合机理，揭示杂草破坏、土壤变形及多相介质耦合规律；采用Modelica语言在MWorks平台下建立除草施肥同步作业机构的机、电、液、气、控系统的多领域耦合模型，探明影响其状态切换性能的主要原因，为新型同步作业机构的研制提供理论基础和设计依据。本项目的实施不仅可以解决水稻精准机械除草和施肥的技术难题，也可为其他作物精准田间管理技术的研究提供参考和借鉴。

本项目开展了稻株识别和定位方法研究、除草部件与水田土壤耦合机理研究、株间除草机构与液肥深施装置工作机理研究、除草机和施肥机研制等。首先根据稻穴内稻株群茎基部成束的特点，提出了基于茎基部边缘拟合的稻株定位方法，定位误差小于7mm；提出了基于深度学习的稻株识别与定位方法，定位精度大于93%；2种定位方法均满足株间机械除草对稻株定位的精度要求。分别基于有限元ANSYS/LS-NYDA和离散元EDEM建立了水田土壤模型，开展了除草部件与水田土壤的互作耦合仿真分析；优化了行间螺旋刀齿式除草轮，除草率大于85%；优化了株间气动摆动式除草部件，除草率大于80%。利用计算流体力学软件Fluent建立了射流式施肥器的仿真模型，设计了一种新型施肥器。开发了3ZSC-190型水田智能除草机、3ZSQ-40型轻简式水田除草机、3GY-1920型宽幅水田除草机、射流式水稻追肥机等4种装备。3种除草装备的除草率均大于80%、伤苗率小于5%。射流式水稻追肥机各行施肥量变异系数小于1%，总排肥量稳变异系数为0.39%，施肥断条率为0%，施肥深度合格率为85.6%，各项性能指标均满足精准施肥要求。本项目完成后，形成了比较完整的水稻株间机械除草和液肥深施的理论方法，部分解决了水稻精准机械除草和施肥的关键科学技术问题。