Bag-of-word model based vehicle type identification method for unlicensed vehicle at gate

Abstract

The invention provides a bag-of-word model based vehicle type identification method for an unlicensed vehicle at a gate. The method comprises: firstly, finishing vehicle detection based on a histogram; finishing vehicle face positioning based on gradient density; performing vehicle type classification based on a bag-of-word model, which comprises two parts of training and testing; and for the problem of poor classification effect caused by correlation of throwaway vocabularies of the bag-of-word model, constructing spatial information. When the spatial information is constructed, the spatial information is divided into two parts of global spatial information and local spatial information to perform calculation, wherein the global spatial information is obtained from a spatial angle relationship between image blocks with the same marks; the image blocks, with the same marks, of an image are sorted according to distances between the image blocks and an origin first, included angles between two adjacent image blocks and a horizontal axis are calculated in sequence, and global angle histogram information about each mark is subjected to statistics; and for the local information, a local frequency histogram of occurrence of the blocks with the same marks is calculated with a specific method. The histograms of the global information and the local information are combined to finish the construction of the bag-of-word spatial information. The method is relatively high in accuracy, has relatively low vehicle face positioning requirements in positioning accuracy, is short in running time, and can meet general application requirements.

Description

technical field

[0001] The invention relates to a bayonet unlicensed vehicle type recognition method based on a bag-of-words model, which belongs to the application of image processing and analysis in intelligent video monitoring. Background technique

[0002] Due to the metal characteristics of the vehicle itself, the vehicle type recognition technology is easily affected by light, has few feature points, and is highly dependent on the license plate location. There are many types of vehicles of the same type, so the research is quite difficult. At present, the vehicle type recognition of the public security bayonet system generally detects the license plate information, thereby accurately locating the front face area, and judges the vehicle model through corresponding analysis of the front face of the car. However, the current method of this system has higher requirements on the positioning accuracy of the front face in terms of positioning accuracy, and when the front face...

Images