Research And Implementation Of Vehicle Characteristics

In recent years, with the rapid development of economics and technology in ourcountry, the vehicle has become an essential tool in our lives, leading to the rapidlyincrease of their number. The management of vehicles becomes heavier and heavier. Forexample, the charge management in highway and parking lot, and the management ofcar violation require a lot of staff to manage, resulting in the biggest disadvantage thatthe efficiency is still very low.In recent years, the automatic recognition of vehicle features has been developed,especially in the recognition of license plates. Many companies have launched their ownlicense plate recognition system. But most of those systems can work only in somespecific conditions, while fail in certain bad condition, especially under bad illuminationor at night. At the same time, the recognition of vehicle logo is also not solved very well.So, the further research in the recognition of vehicle features has great significance.Under the above background, the thesis will deeply research the license platerecognition and vehicle logo recognition, including exploring all kinds of algorithms onthe recognition of vehicle logo and license plate, proposing a new character recognitionmethod based on priori knowledge and making great contributions to the recognition ofvehicle logo. The main contexts of this thesis are as followes:1. License plate detection: Firstly, the input image will be pre-processed. Then thecandidates of license plate will be detected using Adabosst classifier, the real licensedplates will be filtered out from those candidates by the energy of candidates’image, andfinally the precise location of license plate will be extracted by the up-bottom borderand left-right border of plates.2. Segmentation of license plates: this thesis proposes a new methodology forcharacter segmentation to extract the characters from a group of binary image which arebinarized by a group of threshold.3. Recognition of characters: Chinese characters are recognized in this thesis usingthe method of EHMM. While about the recognition of other characters, the thesis decomposites the strokes of a character firstly, mathches the strokes in dataset withthese strokes abtained from the former step, and then gets the result of recognition.4. The detection of vehicle logo: First, get the coarse location of vehicle logo bythe location of license plate, then analyze the texture in the area using sobel operator,and finally extract the precise location of the logo.5. The recognition of logo: First, detecte the corners in the logo image, then extracttheir features, and match with that in dataset to get one of the most similar results.All algorithms in this thesis are developed using C language based on Opencv,using the development tool of Microsoft Visual Studio2005.