Automobile sign accurate positioning method based on road checkpoint blurred image

Abstract

The invention provides an automobile sign accurate positioning method based on a road checkpoint blurred image. The method comprises the steps of according to the salient image characteristics of theautomobile sign area, by means of an image processing method of four stages, positioning area of the automobile sign accurately; obtaining automobile logo image information, the problem that in the prior art, clear images need to be provided by road level photographing equipment is solved. Even if the provided image is relatively fuzzy, the identification and positioning precision of the automobile logo is still very high; in an experiment, a test picture comes from materials under different illumination and weather conditions; the comprehensive recognition rate reaches 91%, the automobile logo positioning processing effect is accurately judged, the accuracy and timeliness of the algorithm are good, and the method has robustness and high efficiency, can adapt to various automobile brands,has universality and transportability, and has huge market popularization and application potentials.

Description

technical field [0001] The invention relates to a method for clearing and processing fuzzy text images, in particular to a method for precise positioning of automobile signs based on fuzzy images of road checkpoints, and belongs to the technical field of fuzzy text image processing. Background technique [0002] Vehicle logo positioning is an important part of vehicle information capture and collection in the field of intelligent transportation development. At present, people have more stringent demands on the quality of data information, and the acquisition of information must meet high integrity and authenticity. Regional traffic control systems, license plate recognition and search systems, vehicle speed detection systems, and red light detection systems for pedestrian passages have been applied in many cities and have begun to take shape. At present, the field of intelligent road traffic is mainly the application of image processing related technologies. The essence is t...

AI Technical Summary

Problems solved by technology

[0010] One is some car logo recognition and positioning methods proposed in the prior art, including car logo recognition based on principal component analysis combined with edge invariant moment features, car logo recognition method for edge features combined with image feature invariant moments, multi-feature combination automatic Adapting to sign recognition, most of them can only be used for standard car sign training and recognition. If it is an incomplete car sign, or a slightly disturbing car sign, the recognition rate will drop rapidly. However, this will be encountered in real work. A series of problems encountered, road inspectors and police officers just want to quickly locate and search the type of car signs through the car sign classification algorithm. If each car image needs to be manually intercepted or cropped, it is obviously half the effort with twice the result, and there is almost no use and promotion. value

[0011] The second is that the existing technology based on the vehicle license plate location extracts the car logo, which will seriously affect the positioning accuracy due to the large external interference; the car logo model based on the existing car logo recognition, the accuracy and timeliness of the algorithm are poor

However, this algorithm is only effective for some brands of cars, and it is impossible to only have these types of cars in practical applications. The inherent shortcomings of this method make it impossible to popularize and apply, and the method has great limitations.

[0012] The third is that if there are reflections and other phenomena in the morphologically processed image of the existing technology, there may be multiple large block areas, and the vehicle logo area that needs to be located exists in those areas, making the method completely invalid; The technology's feature recognition search directly matches and locates, and it needs a large number of samples to achieve a good recognition effect. If there are insufficient positive samples for training, it will be difficult

Images