Research On Target Pollination Method Of Kiwifruit And Design Of Control System

Kiwifruit is a cross-pollinating crop.The quality of pollination operation determines the yield and quality of fruit.At present,pollination in China mainly depends on manual work,which is inefficient,pollen consumption is large and pollination effect is difficult to guarantee.Flower recognition and target spraying are key technologies for efficient pollination.The development of kiwifruit pollination control system can promote the intelligence of kiwifruit pollination equipment.It has broad application prospects and important economic value in the development of Kiwifruit industry.In this paper,the design and research of target pollination control system of kiwifruit were carried out with the flower of "Hayward" kiwifruit as the object.The main research contents are as follows:(1)Define the best focusing range of Pixy camera and improve the accuracy of flower recognition.By analyzing the relative error between the image area of Pixy camera and the recognition mark area,it is found that the relative error between the mark area and the recognition mark area is less than 5% when the focusing distance is 15-40 cm.According to the spatial distribution of the vertical height of kiwifruit flower canopy from 160 cm to 180 cm above the ground,the optimum installation height of Pixy camera for identification and location of kiwifruit flower on scaffolding is 140-145 cm above the ground.(2)Determine the basis of Pixy camera image brightness adjustment,and build an image brightness adjustment system to improve Pixy camera’s natural illumination adaptability.By analyzing the accuracy of flower image marking under different illumination conditions,the adjustment formula between ambient illumination and camera imaging brightness is defined: y=-0.0015x+143.95.According to the relationship,the control system of image brightness adjustment is designed and constructed,which realizes the real-time automatic adjustment of image brightness of Pixy camera,effectively improves the recognition effect of Pixy camera under natural illumination conditions,and makes the accuracy of flower image marking more than 95%.(3)According to the location data of camera pixels,a region segmentation control algorithm is designed to realize the target control of pollination nozzle.The space layout of nozzle and camera ensures the accurate matching between nozzle injection area and camera field of view.Based on the mapping relationship between spatial physical size and imagepixels,by analyzing the possible distribution of flowers in the pixel plane,it is determined that the vertical coordinates of flower center points in each frame are 0~20 pixel or the vertical coordinate value is 21~40 pixel,and the pixel area is larger than 1256 pixel as the spray target.According to the corresponding conversion between image pixels and spatial areas,the horizontal division of image pixel plane is divided into three regions.For the target whose abscissa coordinates are 0~104 pixel,105~213 pixel and 214~320 pixel,combined with the delay control,the pollination work of nozzles 1,2 and 3 is delayed,which can prevent the target from repeating pollination and missed pollination.(4)Design the target pollination control system,complete the hardware circuit construction and control program.On the basis of defining the general requirements of target pollination control system,the hardware selection of the control system is completed,and the hardware circuit is designed by Altium Designer 9.According to the region segmentation and target control algorithm,the sub-programs of image brightness adjustment and flower data acquisition module,chassis remote control module,delay acquisition module and target pollination execution module are developed through Arduino IDE,and the general program of target pollination control system is designed and implemented.(5)Target identification and target pollination verification test.Through setting up a target pollination machine and pollination environment,270 Kiwifruit flowers were tested under high,medium and low illumination respectively.The experimental results show that the target recognition rate can reach 93.7% and the hit rate is 83.3%.The control accuracy of the region segmentation control algorithm and the overall performance of the target pollination control system are verified.