Research On Crop Row Recognition And Path Planning Based On Multi-information Fusion

With the development of automation and intelligence of agricultural machinery,manual spraying operations are gradually replaced by intelligent agricultural machinery.The intelligent sprayer has high operating efficiency and can be widely used in plant protection and prevention in the agricultural field.It is one of the key research directions in the field of digital agriculture.In the agricultural field,most of the intelligent sprayers are used in greenhouse orchard scenes,but the outdoor farmland environment is complicated and the operation requirements are high.Based on farmland scenes,there is little research on autonomous navigation of unmanned agricultural machinery.Based on the self-developed unmanned sprayer platform,this paper builds a multi-sensor system and feed sprayer control system to conduct research on crop row identification and path planning in farmland scenes.The main research contents are as follows:(1)The sprayer experiment platform and control system were built.Starting from the overall structure of the sprayer,the chassis structure,walking mechanism and steering mechanism were introduced respectively.Aiming at the change of plant height and working width and the complex working environment of farmland,the boom structure and high-strength walking mechanism with automatic lifting and folding functions were studied;relying on unmanned agricultural machinery as a carrier,a multi-sensor system module was built,mainly including Kinect Sensors,RTK-GPS sensors and laser sensors,combined with multi-information sensors for data collection,to obtain dual-source information of plant crops and the location of obstacles,etc.,to provide data support for subsequent crop row identification and path planning,and the host computer through the main control module The decision information is converted into the sprayer's control signal to realize the sprayer's driving operation according to the specified strategy.(2)Improved the method of plant crop row recognition based on dual information fusion.First use Kinect to obtain the color and depth information of the target plant,convert the image from the RGB color model to the HSI color model,extract the S component(saturation)threshold,filter the depth data,and use the sampling consistency(SAC-IA)The registration method finds the transformation relationship for rough registration,and then uses the ICP algorithm to accurately register with the S component map,and respectively performs the classification center and particle fitness dimension determination,and obtains the clustering center through the particle swarm optimization(PSO)cycle optimization Use least squares to fit the path.Experiments show that the dual-information fusion crop line recognition method is optimized in real-time and accuracy compared with traditional methods,especially in real-time.(3)The improved A * algorithm farmland path planning method based on the idea of jumping points is improved.Use Kinect sensors,laser sensors,and RTK-GPS positioning devices to obtain plant crop information,farmland boundary information,and obstacle location information.Combined with the sprayer's own parameters such as turning radius,maximum operating width and other data,the improved A * algorithm of this paper is used to plan the path of the unmanned sprayer,and plan and analyze the straight path and the turning path.Finally,the applicability of the method in this paper is verified by experimental simulation.The improved A * algorithm can reduce a large number of invalid searches,reduce search time,and improve planning efficiency.The simulation results of the global planning method show that the vehicle can travel and avoid obstacles according to the planned route,and the deviation is small.Meet the autonomous operation requirements of unmanned sprayer.