KR20130031308A - Recognition of number plate system using compensation of image distort - Google Patents

Abstract

PURPOSE: A vehicle number recognition device for correcting image distortion of a camera is provided to correct an image distortion phenomenon of an HD class of a CMOS camera by using an image correction unit. CONSTITUTION: A high quality still image is obtained by correcting image distortion. A system includes a photographing unit, an image obtainment unit, an image processing unit, and an image correction unit. The image correction unit corrects image distortion corresponding to a rolling shutter. An image distortion phenomenon, which is generated by movement of a subject, is removed. [Reference numerals] (AA,DD) Image input unit; (BB,EE) Camera input unit; (CC,GG) Processing unit; (FF) Distorted image correction processing

Description

Recognition of number plate system using Compensation of Image Distort}

The present invention relates to a technique for correcting distorted image information from a distorted image, in particular, a rotary shutter type camera, rather than a general camera image, and a vehicle number recognition device using the same.

Typical license plate recognition equipment is as follows. As shown in Figure 1, the number recognition equipment is divided into a camera unit and an image input unit processing unit. In general, the image input unit is a processor unit that acquires an image by controlling a camera and a lens and recognizes the processor, that is, a vehicle number.

Complex image processing techniques, such as vehicle identification, require very sharp images, so the cameras you use are more sophisticated. In addition, at night, special lighting near-infrared, infrared or floodlights are used to obtain a visual effect close to the daytime. The acquired image is able to recognize the character by identifying the position of the character in the image by a technique such as its own neural network or pattern matching, and classifying the character of the identified position to determine whether it is a previously learned character.

However, such equipment uses expensive cameras, which greatly affects the overall system price and cannot be generalized due to such a price burden.

These special cameras use the method of frame input rather than the usual interlaced scanning method and use a fast response CCD device for each line to react simultaneously.

Currently, low-cost HD-class CMOS digital cameras with 8 times the resolution and 1/10 the resolution of these expensive CCD devices (Prograssive scan cameras) have various problems due to the characteristics of CMOS devices. However, as the processor of the processor speeds up, image distortion is corrected to obtain expensive CCD-class images, which greatly improves the performance of the entire vehicle number recognizer and makes it possible to spread the vehicle number recognizer.

2002 compound vehicle crackdown equipment

Image distortion of the rotary shutter

The present invention provides a separate processor that is responsible for image distortion correction in the configuration of a general system by correcting image distortion (Fig. 2) generated during the shooting (Rolling Shutter) of a CMOS camera, and corrects image distortion correction with a continuous image, thereby making it expensive. To construct a video analysis system. The present invention relates to the use of the continuous image in the correction of the distortion of the image and to a system using the same.

The physical configuration of the present invention is divided into a lens unit, a camera unit, and a control unit as shown in FIG. 1, and a separate image distortion correction processing unit as shown in FIG. 1B to operate an image correction processor required for the present invention. Therefore, the image analysis system and the camera unit uses an HD-class CMOS camera.

As a method of image correction, we use continuous image feature points and moving vector reference frames, which use software configuration and input stage filters to prevent minimal image distortion at the input stage.

The present invention is to replace the existing high-priced low-performance CCD camera with a low-cost high-performance HD camera by correcting the image distortion of the HD-class CMOS camera to be used in complex image processing equipment such as vehicle number recognition equipment.

1 is a configuration diagram of the existing system and improved equipment
Figure 2 Types of image distortion

In the case of using a CMOS HD camera instead of the global shutter image acquisition of a general ccd device, the configuration of the system is shown in FIG. Unlike the existing system, the image processing part is attached so that image distortion can be corrected to obtain a high quality still image. The configuration of the system includes a photographing unit, an image capturing unit processor, and an image correcting unit, and the image correcting unit corrects the following image distortion due to the rolling shutter. Image distortion is as follows

Image distortion of CMOS HD rolling shutter can be classified as follows.

Jello phenomenon

Jello is a geometric image distortion that occurs when a subject moves at high speed. As shown in FIG. 2, the horizontality is changed to an oblique line when moved laterally. This happens in reverse due to the asynchronous scan of the line when the subject moves.

In particular, such image distortion that may occur during high-speed vehicle shooting can be corrected as follows. First, the image synchronization is physically synchronized to the CMOS. In this case, since the device generates heat, the following can be corrected in the image processing. The moving subject has a mobility vector in each background image. When the vector image is corrected in the photographed consecutive frames, the image distortion phenomenon can be greatly reduced.

-Aliasing phenomenon

This is a phenomenon where synchronization occurs in each row of the sensor. It is a phenomenon that the straight line is not straight but is refracted. As shown in FIG. 2, the image is blurred and can be edited with the reference image.

-Wobbling

Even if the camera is fixed, it can be shaken under the influence of vibration.

The cause of distortion of the images listed above is a rolling shutter that does not synchronize the camera's image line at the correct time. Therefore, the Kalman technique using the time series background image and the background image designation technique correct the image distortion listed above. It can be applied directly to systems that require accurate images such as character recognition. The image correction technique is as follows.

The frame of the reference image is a reference image without deviation of the moving object. Rolling shutter is a problem caused by the inconsistency of the line scan while the subject is moving. It is possible to detect distortion caused by the intensity as a variable. Pixel blocks with low connection strength are acquired from the reference frame, and pixels with high connection strength are regarded as movement of the real object, and image noise generated during movement can be removed.

wF (x, y) = Fn-1 F (x-1, y) * a + Fn-1 F (x-1, y-1) * a + Fn-1 F (x, y) * a + Fn -1 F (x + 1, y) * a + Fn-1 F (x, y + 1) * a + Fn-1 F (x + 1, y + 1) * a

SF (x, y) = wF (x + 1, y) + wF (x + 1, y + 1) + wF (x, y) + wF (x-1, y) + wF (x, y-1 ) + wF (x-1, y-1)

The connection strength between pixels is obtained by the above equation and changes as SF value. Therefore, the noise removal value is as follows.

T( x,y) - SF(x,y)

＞ weight(F(x,y)) 과 같이 임계치를 정할수 있다.Thershold abs

T (x, y)-SF (x, y)

> Threshold (F (x, y)).

Claims (3)

Vehicle number recognition system that corrects image distortion of HD SDI Camera interface. Vehicle Number Recognition System Using HB SDI Camera Interface. Kalman series time-series filters to correct image distortion in HD SDI camera interface

Images