A license plate recognition image enhancement method

Abstract

The present invention provides a license plate recognition image enhancement method, which specifically includes: S1, target display brightness adjustment; S2, sharpening processing; S3, image denoising; S4, AE processing; S5, license plate target detection, detecting, analyzing, capturing, and processing vehicles passing through a specified area. When the license plate information cannot be accurately analyzed, a license plate or a distant view of the vehicle passing by is pushed to the display screen, and then steps S1-S4 are executed to perform image enhancement processing on the license plate or distant view to accurately obtain the license plate information. The license plate recognition image enhancement method is simple to operate and does not require complex training. It can effectively improve the success rate of license plate recognition under factors such as low illumination, excessive speed, blurred license plates, unclear captured images, etc., and has a wide range of application scenarios.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a license plate recognition image enhancement method. Background Art

[0002] License plate recognition is a computer video image recognition technology used to identify and track vehicle license plates. The license plate recognition system can extract and identify moving car license plates from complex backgrounds. Through license plate extraction, image preprocessing, feature extraction, license plate character recognition and other technologies, the vehicle license plate number, color and other information can be identified. The existing license plate recognition system has a good effect on capturing slow-moving vehicles or gate detection vehicles, and can effectively identify license plate information, but the recognition effect for medium- and high-speed or street-mode vehicles is not ideal, especially for low illumination and day-night switching points. The recognition rate is not high, and the detection scene and detection mode are relatively simple.

[0003] In the prior art, there are also reports on the technology of license plate recognition image enhancement. For example, the Chinese patent application with publication number CN111325283A discloses a method for low-light license plate image enhancement recognition based on convolutional neural network, which includes the following steps: S10, constructing an image enhancement network and a vehicle license plate recognition network based on convolutional neural network; S20, training the image enhancement network using training samples; S30, training the vehicle license plate recognition network using training samples; S40, inputting the low-light vehicle number image into the trained image enhancement network to obtain an enhanced image; S50, inputting the enhanced image into the trained vehicle license plate recognition network to obtain the vehicle license plate number. This method can maintain the spatial smoothness of the illumination image and ensure that the edge structure of the illumination image is retained. The image after enhancement processing has no halo phenomenon, and the details are more prominent, achieving a good visual effect. Combined with the vehicle license plate recognition neural network system, the vehicle license plate information can be accurately obtained. However, the above method requires a lot of network training, is cumbersome to operate, and is inconvenient to use. Although the enhancement effect of low-light license plate images is good, the recognition effect of medium-speed or street mode vehicles is not well enhanced, and the application scenarios are limited. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present invention proposes a license plate recognition image enhancement method, which is simple to operate and does not require complicated training. It can effectively improve the success rate of license plate recognition due to low illumination, excessive speed, blurred license plates, unclear captured images, etc., and has a wide range of application scenarios.

[0005] To implement the above technical solution, the present invention provides a license plate recognition image enhancement method, which specifically includes the following steps:

[0006] S1. Target display brightness adjustment: according to the changes in the light brightness of the current scene, the target brightness standard specified based on human visual comfort, display performance and energy efficiency factors is preset. According to the data provided by the ambient brightness sensor and the preset target brightness standard, the display brightness suitable for the current environment is calculated to improve the visual effect of the image;

[0007] S2, sharpening processing, according to the changes in ambient brightness, adjust the sharpness of large edges, small edges, sharpness strength of the Yuv domain, white edge white points, black edge black points, to purposefully emphasize the overall or local features of the image, enhance image clarity, and expand the differences between the features of different objects;

[0008] S3, image denoising, adjust the 3D denoising parameters according to the changes in the environment, reduce the control parameters of the pure time domain filter in the darkroom environment, reduce object smear, increase the 2D denoising parameters, reduce the jump of noise points, eliminate the noise in the image, and provide clear data for license plate edge detection;

[0009] S4, AE processing, adjust exposure time, improve the image quality of moving objects, and improve intelligent recognition rate;

[0010] S5, license plate target detection, detection, analysis, capture, and processing of vehicles passing through the designated area. When the license plate information cannot be accurately analyzed, a license plate or a distant view of the vehicle passing by is pushed to the display screen, and then steps S1-S4 are executed to perform image enhancement processing on the license plate or distant view to accurately obtain the license plate information.

[0011] Preferably, in step S1, the display brightness suitable for the current environment is calculated in the following manner:

[0012] S11, divided into 16 levels from dark to bright according to different scenes, the display brightness BV changes with the target brightness AETarget;

[0013] S12, calculate the exposure value EV. For the exposure value EV of a specified size, there is a functional relationship between it and the corresponding shutter speed and aperture size: EV=TV+AV. TV is the shutter speed value, and its relationship with the exposure time T is TV=log2(1 / T); AV is the lens aperture size value, and its relationship with the lens aperture A is AV=log2A 2 ; EV is the exposure value. Any given exposure value can be obtained by adjusting the exposure time and aperture size;

[0014] S13. Calculate the display brightness BV. The exposure value EV is also a function of the image display brightness BV and the sensitivity SV. There is a functional relationship EV=f(BV, SV). SV represents the sensitivity. For an image sensor, the sensitivity SV is single and can be obtained by setting the target brightness AE Target. The target brightness AE Target is the brightness that the ISP wants the sensor to output. For the first time entering the sensor, if the ISP detects that the brightness cur_luma of the current frame is lower than the AE Target, it will increase the exposure time for re-exposure until cur_luma is equivalent to the AE Target. The f in the expression can be obtained through the metering setting, and then the display brightness BV can be calculated in reverse.

[0015] Preferably, in step S5, the license plate target detection process includes a vehicle priority mode and a license plate or vehicle mode, wherein the vehicle priority mode only detects, analyzes, captures, and processes vehicles passing through the designated area, and then pushes the obtained vehicle license plate information to the display screen. When the license plate information cannot be accurately analyzed and an alarm is triggered due to factors such as excessive vehicle speed, blurred license plates, and unclear captured images, a distant view of the passing vehicle will be pushed for user reference; the license plate and vehicle mode is used for scenes in low illumination or night vision where the vehicle's shape is difficult to see clearly. Vehicles and license plates passing through the designated area are simultaneously detected, analyzed, captured, and processed, and then the obtained vehicle license plate information is pushed to the display screen.

[0016] Preferably, in step S5, when detecting the license plate target, the entrance and exit mode, street mode, and external trigger mode can be selected according to the actual usage scenario, wherein the entrance and exit mode means that the camera is set to monitor the entrance and exit, and the license plate detection is performed after a vehicle appears, and the detection information is compared with the intelligent library information to see if they match, and then a judgment is made whether to open the gate to release the vehicle; the street mode means that when a vehicle passes through the detection area and triggers the detection, the license plate information of the detected vehicle will be uploaded and pushed; the external trigger mode means that an I / O alarm will trigger the license plate capture and recognition action, and this mode will continuously push pictures during the alarm process.

[0017] Preferably, in step S3, image denoising is implemented using a 3DNR interface, and image edges and textures are preserved at different gains by adjusting parameters of a pure time domain filter, a pure space domain filter, and a time domain and space domain hybrid filter in an adjustment tool.

[0018] The beneficial effects of a license plate recognition image enhancement method provided by the present invention are:

[0019] (1) This license plate recognition image enhancement method can effectively enhance the useful information in the license plate image and improve the visual effect of the image. For the application of a given image, it can purposefully emphasize the overall or local characteristics of the image, make the originally unclear image clear or emphasize certain interesting features, expand the difference between the features of different objects in the image, and suppress the features of no interest, so as to improve the image quality, enrich the amount of information, and enhance the image interpretation and recognition effect, providing a good foundation for the subsequent license plate positioning, segmentation and character recognition, and greatly improving the accuracy and efficiency of vehicle recognition;

[0020] (2) This license plate recognition image enhancement method is simple to operate and does not require complex training. It can effectively improve the success rate of license plate recognition under factors such as low illumination, excessive speed, blurred license plates, and unclear captured images. It has a wide range of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the steps of the present invention.

[0022] Figure 2 It is a diagram of target brightness parameter debugging and target display brightness parameter in an embodiment of the present invention.

[0023] Figure 3 This is a diagram for debugging sharpening processing parameters in an embodiment of the present invention.

[0024] Figure 4 This is a diagram for debugging image denoising parameters in an embodiment of the present invention.

[0025] Figure 5 This is a diagram for debugging AE processing parameters in an embodiment of the present invention.

[0026] Figure 6 Schematic diagram of license plate image processing information in entrance and exit mode in an embodiment of the present invention.

[0027] Figure 7 Schematic diagram of license plate picture information push in vehicle priority mode in an embodiment of the present invention.

[0028] Figure 8 This is a vehicle perspective image pushed before being processed by the present invention in an embodiment of the present invention.

[0029] Fig. 9 This is a distant view of the vehicle after being processed by the present invention in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary persons in the art without creative work are within the scope of protection of the present invention.

[0031] Embodiment: A license plate recognition image enhancement method.

[0032] In image analysis, the quality of the image directly affects the design of the recognition algorithm and the accuracy of the effect. Therefore, during image preprocessing, it is very necessary to eliminate irrelevant information in the image, restore useful real information, and enhance the detectability of relevant information.

[0033] Based on the above principle, a license plate recognition image enhancement method provided by the present invention specifically includes the following steps:

[0034] (1) Target display brightness adjustment

[0035] According to the changes in the brightness of the current scene, a series of brightness standards are preset ( Figure 2 The brightness standard is specified based on factors such as human visual comfort, display performance, and energy efficiency. Then, based on the data provided by the ambient brightness sensor and the preset target brightness standard, the display brightness suitable for the current environment is calculated to improve the visual effect of the image and enhance the recognition of the image.

[0036] Specifically, the display brightness suitable for the current environment is calculated in the following way:

[0037] First, the image display brightness BV changes with the image target brightness AE Target, and can be divided into 16 levels from dark to bright according to different scenes. Then calculate the image exposure value EV. For a specified exposure value EV, there is a functional relationship with the corresponding shutter speed and aperture size EV = TV + AV, where TV is the shutter speed value, and its relationship with the exposure time T is TV = log2 (1 / t); AV is the lens aperture size value, and its relationship with the lens aperture A is AV = log2A 2EV is the exposure value. Any given exposure value can be obtained by adjusting the exposure time and the aperture size. Finally, the display brightness BV is calculated. The exposure value EV is also a function of the image display brightness BV and the sensitivity SV. There is a functional relationship EV = f (BV, SV). SV represents the sensitivity. For an image sensor, the sensitivity SV is single and can be obtained by setting the target brightness AE Target. The target brightness AE Target is the brightness that the ISP (image signal processor) wants the sensor to output. For the first time entering the sensor, the ISP detects that the brightness cur_luma of the current frame is lower than AETarget, and the exposure time will be increased for re-exposure until cur_luma is equivalent to AE Target. And the f in the expression can be obtained through the metering setting. When the exposure value EV is determined and the expression f is determined, as long as the sensitivity SV is determined by the change of the target brightness AE Target, the display brightness BV can be calculated in reverse, and then the display brightness BV of the picture follows the floating change of the target brightness AE Target of the picture to obtain the best image visual effect, enhance the recognition of the picture, and provide a good visual basis for the subsequent picture enhancement processing.

[0038] (2) Sharpening

[0039] Manually adjust the values ​​of large edge sharpness, small edge sharpness, sharpening intensity of Yuv domain, white edge white point, and black edge black point according to the changes in ambient brightness (such as Figure 3 As shown in the figure, the overall or local features of the image can be purposefully emphasized to enhance the image clarity and to enlarge the differences between the features of different objects, thereby improving the capture quality and recognition rate.

[0040] (3) Image denoising

[0041] like Figure 4 As shown in the figure, the 3D noise reduction parameters are adjusted according to the changes in the environment. In the darkroom environment, the control parameters of the pure time domain filter are reduced to reduce the object smear, and the 2D noise reduction parameters are increased to reduce the noise point jump, eliminate the noise in the image, and provide clear data for license plate edge detection. Noise elimination is mainly achieved using the 3DNR interface. By adjusting the pure time domain filter, pure spatial domain filter, and time domain and spatial domain hybrid filter in the tool, the image edge and texture are preserved at different gains.

[0042] Especially in low-light scenes, data comparison shows that the impact of noise on captured image recognition is smaller than that of smear. Therefore, in 3DNR processing, the time domain parameters are reduced, the noise is released appropriately, the upper limit of gain is limited, the soft photosensitivity parameters are adjusted, the black and white mode is entered in advance, and the color mode is entered later to avoid the impact of high gain on license plate recognition.

[0043] (4)AE processing

[0044] In the past, the license plate recognition system has a good effect on capturing slow-moving vehicles or vehicles detected at gates, but the effect is not ideal for medium-speed or street-mode vehicles. Through a large amount of data testing and comparison, the reasonable use of exposure time (such as Figure 5 As shown), it can improve the quality of pictures captured of moving objects and thus improve the intelligent recognition rate, and to a large extent solve the problem of license plate recognition rate.

[0045] (5) License plate target detection: Detect, analyze, capture, and process vehicles passing through the specified area. When the license plate information cannot be accurately analyzed, a license plate or a distant view of the vehicle passing by is pushed to the display screen, and then steps (1)-(4) are executed to perform image enhancement processing on the license plate or distant view to accurately obtain the license plate information.

[0046] When detecting license plate targets, you can select entrance and exit mode, street mode, or external trigger mode according to the actual usage scenario:

[0047] 1. Entrance / Exit (entry and exit mode)

[0048] Indicates that the camera is set to monitor the entrance and exit. When a vehicle appears, the license plate is detected, and the detection information is compared with the information in the smart library to determine whether to open the gate to release the vehicle. Right-click on the pushed license plate image to pop up a dialog box (such as Figure 6 As shown), you can choose to add the acquired information to the smart library as needed. This mode only supports 1 lane.

[0049] 2. City Street (Street Mode)

[0050] It means that when a vehicle passes through the detection area and triggers the detection, the license plate information of the detected vehicle will be uploaded and pushed. This mode supports 2 lanes.

[0051] 3. Alarm Input (external trigger mode)

[0052] Indicates that an I / O alarm will trigger license plate capture and recognition and other actions. Please note that in this mode, images will be pushed continuously during the alarm process.

[0053] License plate detection target, currently the program supports European license plate and American license plate.

[0054] The license plate target detection process includes vehicle priority mode and license plate or vehicle mode.

[0055] Vehicle priority mode: This mode only detects, analyzes, captures, and processes vehicles passing through the specified area, and then pushes the obtained vehicle license plate information to the screen, such as Figure 7 shown.

[0056] When an alarm is triggered, the license plate information cannot be accurately analyzed due to factors such as excessive speed, blurred license plate, unclear captured image, etc., and a distant view of the vehicle passing by will be pushed for user reference, such as Figure 8 As shown in , the license plate information cannot be distinguished at this time. After the distant view is processed by this method, the originally unclear image becomes clear, as shown in Fig. 9 As shown, the license plate photo is now clearly visible. Right-click on the image and select Detail Information to further view detailed information about the vehicle passing by.

[0057] License plate or Vehicle mode: This mode simultaneously detects, analyzes, captures, and processes vehicles and license plates passing through a specified area, and then pushes the obtained vehicle license plate information to the screen. It is mostly used in scenes where the vehicle's appearance is difficult to see clearly in low illumination or night vision.

[0058] This license plate recognition image enhancement method can effectively enhance the useful information in the license plate image, improve the visual effect of the image, and purposefully emphasize the overall or local characteristics of the image for the application of a given image, making the originally unclear image clear or emphasizing certain features of interest, expanding the difference between the features of different objects in the image, and suppressing features of no interest, so as to improve the image quality, enrich the amount of information, and enhance the image interpretation and recognition effect, providing a good foundation for subsequent license plate positioning, segmentation and character recognition, and greatly improving the accuracy and efficiency of vehicle recognition. Moreover, this license plate recognition image enhancement method is simple to operate, and does not require complex training. It can effectively improve the success rate of license plate recognition due to low illumination, excessive speed, blurred license plates, unclear captured images, etc., and has a wide range of application scenarios.

[0059] The above description is only a preferred embodiment of the present invention, but the present invention should not be limited to the contents disclosed in the embodiment and the drawings, so any equivalent or modification completed without departing from the spirit disclosed in the present invention shall fall within the scope of protection of the present invention.

Claims

1. A license plate recognition image enhancement method, characterized in that The specific steps include: S1. Target display brightness adjustment: according to the changes in the light brightness of the current scene, the target brightness standard specified based on human visual comfort, display performance and energy efficiency factors is preset. According to the data provided by the ambient brightness sensor and the preset target brightness standard, the display brightness suitable for the current environment is calculated to improve the visual effect of the image; Specifically, the display brightness suitable for the current environment is calculated in the following way: S11, divided into 16 levels from dark to bright according to different scenes, the display brightness BV changes with the target brightness AETarget; S12, calculate the exposure value EV. For a specified exposure value EV, there is a functional relationship between it and the corresponding shutter speed and aperture size: EV = TV + AV, where TV is the shutter speed value and its relationship with the exposure time T is TV = ; AV is the lens aperture value, and its relationship with the lens aperture A is AV = ; EV is the exposure value. Any given exposure value can be obtained by adjusting the exposure time and aperture size; S13, calculate the display brightness BV, the exposure value EV is a function of the image display brightness BV and the sensitivity SV, and there is a functional relationship EV=f(BV, SV), SV represents the sensitivity, for an image sensor, the sensitivity SV is single, and is obtained by setting the target brightness AE Target, the target brightness AE Target is the brightness that the ISP wants the sensor to output the picture to achieve, for the first time entering the sensor, the ISP detects that the brightness cur_luma of the current frame is lower than the AE Target, it will increase the exposure time for re-exposure until cur_luma is equal to the AE Target, and obtain f in the expression through the metering setting, and then reversely calculate the display brightness BV; S2, sharpening processing, according to the changes in ambient brightness, adjust the sharpness of large edges, small edges, sharpness strength of the Yuv domain, white edge white points, black edge black points, to purposefully emphasize the overall or local features of the image, enhance image clarity, and expand the differences between the features of different objects; S3, image denoising, adjust 3D denoising parameters according to environmental changes, reduce the control parameters of pure time domain filter in darkroom environment, reduce object smear, increase 2D denoising parameters, reduce noise point jump, eliminate noise in the image, and provide clear data for license plate edge detection; image denoising is implemented using 3DNR interface, by adjusting the parameters of pure time domain filter, pure spatial domain filter, and time domain and spatial domain hybrid filter in the tool, image edges and textures are preserved at different gains; S4, AE processing, adjust exposure time, improve the image quality of moving objects, and improve intelligent recognition rate; S5, license plate target detection, detection, analysis, capture, and processing of vehicles passing through the designated area. When the license plate information cannot be accurately analyzed, a license plate or a long-range image of the vehicle passing by is pushed to the display screen, and then steps S1-S4 are executed to perform image enhancement processing on the license plate or the long-range image to accurately obtain the license plate information. The license plate target detection process includes a vehicle priority mode and a license plate or vehicle mode. The vehicle priority mode only detects, analyzes, captures, and processes vehicles passing through the designated area, and then pushes the obtained license plate information of the vehicle to the display screen. When the license plate information cannot be accurately analyzed and an alarm is triggered due to factors such as excessive speed, blurred license plates, and unclear captured images, a long-range image of the vehicle passing by will be pushed for user reference; the license plate or vehicle mode is used for scenes where the appearance of the vehicle is difficult to see clearly in low illumination or night vision. Vehicles and license plates passing through the designated area are detected, analyzed, captured, and processed at the same time, and then the obtained license plate information of the vehicle is pushed to the display screen; When detecting license plate targets, select entrance and exit mode, street mode, and external trigger mode according to the actual usage scenario. The entrance and exit mode means that the camera is set to monitor the entrance and exit, and license plate detection is performed when a vehicle appears. The detection information is compared with the information in the smart library to see if they match, and then a decision is made whether to open the gate to release the vehicle. The street mode means that when a vehicle passes through the detection area and triggers the detection, the license plate information of the detected vehicle will be uploaded and pushed. The external trigger mode means that an I / O alarm will trigger the license plate capture and recognition action, and this mode will continuously push pictures during the alarm process.

Images