Font Size: a A A

Low-quality License Plate Character Classification Based On CNN

Posted on:2016-07-06Degree:MasterType:Thesis
Country:ChinaCandidate:P LiuFull Text:PDF
GTID:2348330536467404Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
In the background of the criminal investigation,and other special applications,we often encounter the problem how to deal with those lower quality images or surveillance videos which are blurred seriously.And these video information in images or videos are critical,especially the information of license plate.Normal license plate recognition systems can only be applied at a specified occasion,such as parking,toll booths and a variety of transportation bayonet,etc.In some special cases,the license plate in those surveillance videos are blurred seriously,even the human eye can not recognize its plate characters.So we need some way to help identify these license plate characters.But the existed image restoration techniques are applied to simulation data,they are no useful for the actual data.And degradation factors in surveillance videos are various,it is hard to estimate the degradation parameters and restore the images.This paper points out the difficulties in character recognition of such low-quality license plate.Analyzed the processes and research status of license plate character recognition,and discusses the problems of existing methods to solve such problem,and proposed solutions according to the problems.Finally,experimented the solution and analyzed the results.Proposed a solution to complete the character recognition by classify character images based on convolutional neural network.Analyzed the importance and the difficulties to recognize car plate characters from low-quality monitoring videos.Proposed a solution to complete the character recognition by classify character images based on convolutional neural network.Analyzed the degradation of car plate character in practical monitor,and then modeled and generated simulated dataset.By training the convolutional neural network on a simulated dataset,the experimental results in the actual monitoring show that this algorithm works well to classify license plate characters.And it can help identify some of the license plate characters that human eye can't identify.
Keywords/Search Tags:license character classification, convolutional neural network, low-quality license plate, degradation model
PDF Full Text Request
Related items