Research And Application Of Visual Real-time Recognition Of Floating Aquatic Plants In River Basins

As the core of hydrological work,hydrological data collection plays an important role in the coordinated development of people’s livelihood economy and ecological environment.With the integration and development of multidisciplinary technology,hydrological work is developing towards the direction of fully automatic unmanned flow measurement.At present,semi-automatic flow measurement has been achieved in many stations.In order to avoid floating water plants entangled on the sounding rods and threatening the safety of flow measurement equipment in hydrological operations,manual monitoring and intervention are generally used in hydrological stations.This method is susceptible to visual fatigue,has low efficiency,and interrupts the automatic connection of the flow measurement process.For the floating aquatic plants in the river basin,this paper adopts the image processing method to identify,and applies the identification results to the adaptive adjustment of the current measurement system.The main research contents include:(1)Construction of the monitoring system for the river basin.By analyzing the identification requirements,a remote video monitoring system is built based on the existing flow measurement equipment and functions of the hydrological station.It mainly includes the selection,positioning and installation of the camera,and the determination of the power supply scheme,which realizes the image capture of the upstream area of the platform.Considering the inconvenience of cable channel current measurement and wiring,the wireless bridge component LAN is used to realize real-time transmission of video images from monitoring equipment to the central control center.Finally,the water images are obtained through video framing processing.(2)Research on the identification method of floating aquatic plants in the river basin.Firstly,the preprocessing methods are compared and analyzed,and the optimal preprocessing scheme is selected from the subjective and objective perspectives.The preprocessed grayscale image of the water area not only retains local features,realizes preliminary noise reduction,but also reduces the amount of image data.Then,the optimized local threshold segmentation algorithm is used to reduce the influence of uniform brightness distribution,and the segmentation of aquatic plants and water bodies is realized.Aiming at the lack of features in the aquatic plants segmentation results,this paper proposes a mathematical morphology method to enhance the features.Finally,based on the geometric area features of aquatic plants,the identification of aquatic plants is realized,and the position of aquatic plants is calibrated in the global image through the minimum circumscribed rectangle.(3)Experimental study on aquatic weed identification algorithm.Combined with Open CV image algorithm library and C# language,software development of video surveillance system is completed,and software functional module design is completed,which provides a foundation for experimental collection and later practical application.Experimental results show that: under the sample size,the accuracy of the identification algorithm reaches 98%,which can meet the accuracy requirements of monitoring;And the average time of 0.64 s can meet the real-time demand of adaptive adjustment of suspension rod flow measurement system.(4)Application research of aquatic plants identification results.Firstly,through the fixed position of the camera and the shooting model of the camera,the transformation relationship between the image coordinate system and the world coordinate system is deduced.Then combined with the flow measurement process,the regional range and relative distance of aquatic plants are automatically acquired.Finally,a preliminary plan is made for the adaptive adjustment of the flow measurement system.