Research On Surface Floating Object Monitoring Based On Binocular Vision

The water environment problem is closely related to all human beings.The floating objects on the water surface will directly affect the water quality.These floating objects not only affect the environment,but also threaten the safety of navigation,and bring great harm to the ecosystem,thus threatening human security.In order to remove floating debris from the surface under the premise of reducing manpower,material resources and cost,the key is how to identify floating objects in a simple way.Because machine vision has many successful applications in target detection,this paper also uses machine vision.Surface floating object recognition algorithm.The traditional floating object recognition system has the problem of monitoring only,which is far from the automatic monitoring system.Therefore,this paper further studies the distance measurement and tracking.The main contents of this paper are as follows:(1)Firstly,based on the demand for unmanned boats to automatically salvage surface floats,the existing technology is used to propose a floating body salvage system framework suitable for unmanned boats.(2)The binocular camera is used to construct the image acquisition system,and the internal and external parameters of the camera are obtained by Zhang Zhengyou’s calibration method for calibration.The three-dimensional distance measurement principle is used to realize the ranging function of the whole system.(3)Based on Tensorflow framework,a target recognition module is constructed by using style migration algorithm.Style is used instead of texture.Gram matrix of symbolic style is calculated by VGG network.Floating objects are distinguished according to the small Gram difference between water image and water area image,and the large difference between water image and floating object area.Compared with the classical pattern recognition method,this method is simpler to identify the surface garbage floating matter,and provides strong data support for the follow-up target tracking and salvage.(4)When the target is recognized,the distance between the floating object and the unmanned aerial vehicle is calculated in real time by using binocular ranging module,which provides data support for real-time adjusting the position of the unmanned aerial vehicle and judging whether the current surface garbage floating object can be salvaged or not.(5)Since the movement of unmanned boats and the location of floating debris on the surface of the water will continue to change,it is necessary to use the target tracking module to continuously track the surface garbage floating objects in real time.In this paper,a target tracking module is constructed by using KCF algorithm,which has the best dynamic tracking effect at present.The target is the identified floating object,and a large number of samples are learned to obtain a template,that is,the impact response in digital signal processing.Then,the next frame of image is used as input to find the most similar point in the output,that is,the location of the new target.The final simulation results show that the calibration parameters of binocular camera are more accurate and can meet the needs of subsequent modules.Style loss function can describe the different objects,so as to identify the floating objects.Binocular ranging achieves the desired accuracy.KCF tracking can effectively describe the position relationship between the objects in the front and back frame images.In a word,the design of the whole system satisfies the requirement of fully automatic unmanned salvage of surface floats,and provides a feasible solution for the subsequent real application in the actual salvage work.