Research Of Target Recognition Based On Unmanned Ship And Machine Vision Memory Polyurethane

With the continuous development of marine resources,the monitoring and management of marine activities in all regions is particularly important.The main purpose of this paper is to realize the monitoring of the sea surface by unmanned ships through machine vision,that is,to realize the automatic recognition of the target of the surface vessel and the automatic tracking of the vessel’s target.This article first analyzes the functional requirements of the sea target recognition and tracking system for unmanned ships,and analyzes the working environment of unmanned ships and the characteristics of sea-air images based on the environmental conditions on the sea surface.Based on this,the design is based on The overall design of the sea surface target recognition and tracking of the ship.For the problem that the background part is far more than the target part in the sea-air pictures taken by unmanned ships,this article appears near the sea antenna according to the target’s maximum probability in the sea-air image,and reduces the sea surface noise in the image after filtering by Canny.The image edge is detected by edge detection,and the Hough line detection method based on the improved sea-strait features successfully extracts the sea antenna,and the subsequent processing range is reduced,thereby greatly improving the speed of subsequent target recognition.For targets existing on the sea surface,their contours need to be segmented to obtain the position and contour of the sea surface target for follow-up target identification work.In this paper,we study the method based on gray thresholds and the region growing method,propose a seed point selection method based on gray statistics and direction gradient,and improve a region growing rule based on gray mean value and energy mean value.The successful use of the region growing method to mark the background area achieves the segmentation and extraction of sea targets.For the unknown objects that have extracted contours and locations,this paper uses the method based on fusion features to extract features from the unknown objects and classify them according to the features.The fusion feature uses the invariant moment feature,texture feature,and custom set feature of the target.The classifier uses BP neural network algorithm to train the features extracted from the sample and successfully implements the recognition of the sea surface vessel.In order to achieve target tracking of successfully identified ships,this paper uses SURF feature point extraction and matching algorithm to identify and extract the identified vessel targets,so as to achieve target tracking.In order to reduce the amount of calculation and improve the effect and scope of target tracking,this paper combines the Kalman filtering prediction target position with the stabilization image pan/tilt head tracking target to track the detected sea surface target,which greatly improves the effectiveness of unmanned ship sea surface monitoring.Based on the above research,this paper designs a complete set of sea surface target identification and sea surface activity monitoring solutions for tracking ship targets,which can significantly improve the efficiency of sea surface activity monitoring and management.