Design Of Floating Objects Collection Equipment And Research On Autonomous Recognition Algorithm

With the awakening of people’s awareness of environmental protection and governance,water environment monitoring and protection has become a research hotspot in the field of environmental protection.In view of the problems of low salvage efficiency,high cost,low safety,low degree of automation,few cleaning equipment and so on in the current floating objects collection equipment,this topic designed a novel floating objects cleaning equipment,suitable for small and medium-sized waters,with simple structure,easy processing,high degree of automation,low cost and efficient characteristics.It can realize intelligent image segmentation and detection,independent path planning,automatic fishing and other functions.It provides certain technical support and reference means for environmental protection,governance and detection of floating objects on the water surface,and has a wide range of application prospects.Firstly,Equipment overall scheme and mechanical structure design.Aiming at the pollution problem of small and medium-sized open waters,the overall scheme of the equipment is designed mainly by analyzing the task requirements and functions of the floating objects collection equipment.In addition,the main structure design and construction of floating objects collection equipment,salvage module design and propulsion system design were carried out.Solidworks software was used to conduct three-dimensional modeling of the equipment and various parts,detailed processing drawings were drawn by CAD software,and the prototype of floating objects collection equipment was processed.The necessary equipment parameters were obtained through pool testing.Then,Research on multi-task learning technology based on image segmentation and object detection.This topic draws on the ideas of YOLOP algorithm,adds segmentation head to YOLOv5 s target detection algorithm to realize multi-task learning,and performs image segmentation and target detection on the collected images at the same time.Image segmentation is mainly related to the surface segmentation of water and land.The training set uses ORCA data set to test the trained model,and the surface segmentation performance reaches a high level.Most of the existing object detection methods for convolutional neural networks only have high accuracy when detecting a single type of aquatic target,but the effect is not ideal when detecting multiple types.The recognition accuracy and recognition speed need to be improved,and the robustness of the algorithm cannot be verified under bad weather.Based on the above shortcomings,this thesis adopts self-made floating objects data set and proposes an improved YOLOv5 s target detection algorithm,which effectively improves the accuracy and accuracy of the algorithm.Experiments show that the improved YOLOv5 s algorithm has strong robustness and generalization.Finally,Research on path planning of floating objects collection equipment and design of equipment control system.In this chapter,an improved sparrow search algorithm is proposed to plan it,and a stable and reliable navigation path can be obtained by combining the electronic route chart.In addition,this chapter carries out the design of the control system of floating objects collection equipment,mainly carries out the overall system hardware design,core control chip and sensor module selection,in addition to the system process design,and develops the automatic identification software of floating objects on the water,which can monitor and count the types and quantities of floating objects in real time.