Research And Implementation Of Obiect Recognition And Feature Acquisition System Based On Deep Learning In Specific Scene

In recent years,with the development of artificial intelligence and computer vision technology breakthrough,enterprises gradually promote intelligent and refined management.One of the most important tasks is to detect objects and objects characteristics in the industrial environment.The influencing factors of detection in the construction environment are more complex,such as light intensity,characteristic wear,mud on the surface of the object,shielding of people or materials and other factors will affect the detection of objects and object features.There are some limitations of existing methods for license plate detection in harsh environment.In the detection phase,there exists missed and false detection due to light,dirt and similar license plate features of interference.When the license plate line segmentation before double line recognition,the corner cutting appears and affects identification accuracy.When estimating the number of materials,existing methods can only estimate the number of visible materials in the picture,but cannot estimate the number of cubic stacked materials.In order to solve the above problems,this paper studies and realizes the feature detection system in construction site.In the license plate detection of construction site,the integrated method is used to enhance the detection effect.In the recognition stage,the angle cutting is solved by multi-point regression.In the material count estimation of construction site,two points perspective and single-view object counting are fused to estimate the number of three-dimensional stacked materials.First of all,this paper introduces the research background and the related technologies involved in object feature recognition system.Secondly,the license plate recognition algorithm is designed according to the object detection and text recognition,and the material number estimation algorithm is designed according to the object count algorithm.In the system design and implementation phase,the requirements are analyzed from the user’s perspective and the system logical model,including architecture and modules,is designed.Then detailed design and system development of each module are executed.Finally we conduct the system test.Experimental results show that the system has satisfactory accuracy and reliability,and the system can be used for object detection and feature recognition in site management of construction enterprises.