【作者】 王德宇；

【导师】 唐达；

【作者基本信息】 大连理工大学 ， 计算机技术（专业学位）， 2018， 硕士

【摘要】 海洋面积占地球总表面积的71%,其中蕴藏着丰富的生物资源,被称为“蓝色粮仓”。但是近些年来由于过度捕捞和养殖规划不合理等问题造成了环境的破坏以及海洋生物资源的逐渐枯竭。使用科学手段去修复海洋资源,实现鱼量增长是海洋牧场的重要意义。本文对于海洋牧场建设中的环境监测和智能捕捞探测进行研究,提出了一套基于物联网的海洋牧场养殖系统。环境监测主要是对海洋牧场水质进行长期多点监测,感知层选择溶氧、盐度、温度、PH值四种传感器进行数据的获取;网络层采用系统内部Zigbee自组网,系统外部WiFi接入互联网的拓扑结构;应用层设计一套上位机软件,作为数据的显示和存储中心。目前海洋牧场的捕捞还是靠人工来完成的,不仅具有一定的危险,而且在冬天水温较低不再适合捕捞员下海捕捞。智能捕捞方面分为机械臂的简化控制和对象识别的研究,机械臂作为感知层的主要执行器,用来进行水下的探测和水下生物的捕捞,并加入双目摄像头进行观察海下环境和自动捕捞。本文的创新之处在于对机械臂的控制算法进行简化,传统的控制算法较为复杂,控制机械臂运动到指定位置需要较长的时间。根据实际情况,对机械臂添加合适的约束,提出了一种根据末端位置坐标进行快速控制的逆运动学算法。由于海下环境复杂且光线较暗,传统的方法不能取得很好的效果,将计算机视觉和深度学习这些智能的方法应用到海洋牧场的对象识别中。通过实际应用可以看出,该系统可以进行海洋牧场的实时多点监测,机械臂逆解算算法的到位精度和到位时间满足抓取要求。对象识别可以实现对草莓和橙子的区分,识别时间短、精度高,测距精度满足机械臂自动抓取的要求。采用2017首届水下机器人目标抓取大赛(URPC)所公开的数据集进行试验,所训练的模型可以很好的识别海洋牧场中的海参、海胆和扇贝等海洋生物。更多还原

【Abstract】 The area of ocean accounts for 71% of the total surface area of the earth.It contains abundant biological resources so it is called "Blue Granary".But in recent years,due to problems such as overfishing and unreasonable aquaculture planning,environmental destruction and the depletion of marine living resources have been caused.Using scientific methods to achieve marine resources restoration and realize the increase of fish volume is an important significance of marine ranching.This paper researches the environmental monitoring and intelligent fishing detection in the construction of marine grazing land and proposes a set of marine pasture breeding system based on Internet of Things.Environmental monitoring is mainly about conducting long-term multi-point monitoring of the quality of marine ranching.Perceptual layer select dissolved oxygen,salinity,temperature,and p H values for data acquisition.Network layer adopts Zigbee ad hoc network.Topology of WiFi access to the Internet outside the system.Application layer design a host computer software as the data display and storage center for simple data analysis.At present,the fishing of the marine ranching is still done by human.It is not only dangerous,but also unsuitable for the fishermen to go fishing in the lower water temperature in the winter.Intelligent fishing is divided into the study of simplified control and object recognition of robotic arm.The robotic arm acts as the main actuator of the sensing layer and is used for underwater detection and underwater biological fishing.It also incorporates a binocular camera for observation of the submarine environment and automatic fishing.The innovation of this article is the simplification of the control algorithm of the robotic arm.The traditional control algorithm is more complex,and it takes a long time to control the robot arm to move to a specified position.According to actual condition,adding appropriate constraints to the robotic arm,an inverse kinematics algorithm based on end-position coordinates for rapid control is proposed.Due to the complexity of the undersea environment and dim lighting,it is suitable to apply computer vision and deep learning these intelligent methods to object recognition in marine grazing.Through practical application,it can be seen that the system can perform real-time multi-point monitoring of marine ranches.And in-position accuracy and arrival time of the inverse algorithm of the robot arm satisfy the crawling requirement.Object recognition can distinguish between strawberry and orange,its Short recognition time,high precision,ranging accuracy features can be applied to the robot arm’s automatic grab.Adoption of the data set published in the first underwater robot picking contest in 2017.The trained model can well identify marine organisms such as sea cucumbers,sea urchins and scallops in marine grazing lands.更多还原