A metal sheet positioning method based on 3D laser depth image and reflection image

Abstract

The present invention relates to the field of intelligent warehousing and radio wave positioning technology, and specifically to a metal plate positioning method based on 3D laser depth images and reflection images, which aims to solve the positioning problem existing in the intelligent warehousing management process of metal plates in the prior art. The method first uses 3D laser depth images and reflection images to obtain the position and size information of the plate, and then accurately extracts the edge contour of the plate through a series of image processing techniques, and finally converts the coordinates of the four corner points of the plate into world coordinates and sends them to the MES production management system. The present invention uses a customized 3D line scan camera, combined with the selection and optimal configuration of the laser, to achieve efficient and accurate measurement of large-size metal plates, which can significantly improve the efficiency and safety of intelligent warehousing management of metal plates.

Description

Technical Field

[0001] The present invention relates to the field of intelligent warehousing and radio wave positioning technology, and in particular to a metal plate positioning method based on 3D laser depth images and reflection images. Background Art

[0002] The information disclosed in the background of the invention is only intended to enhance the understanding of the overall background of the invention and should not be necessarily regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art.

[0003] With the advancement of science and technology and the improvement of innovation capabilities, the technical level of the machinery manufacturing industry has been continuously improved. Especially in the fields of automobile manufacturing, heavy equipment, pressure vessels, etc., the sheet forming process accounts for a large proportion in the production and manufacturing process. The warehousing, outbound, turnover, and transportation of sheets require manual lifting operations. At present, the sheets are generally placed on brackets for storage. When entering the warehouse, it is necessary to manually measure the size of the sheet and record the bracket number, and finally enter it into the material system. In order to save floor space, the brackets are generally stacked layer by layer. When the required sheet is located in the middle bracket, it is necessary to manually stack it up, which is inefficient and has high safety risks. Due to the different types and sizes of metal sheets, there is a high expectation that a highly flexible, low-cost, efficient, and stable intelligent storage solution for metal sheets can be implemented as soon as possible. Among them, the current key technology and difficulty is the positioning method of metal sheets, and there is no mature and reliable visual positioning solution on the market.

[0004] With the development of 3D vision technology, space measurement equipment such as 3D cameras have been widely used in the industrial field. Such sensors can be used to obtain the contour data of the surface of an object. With the help of data algorithms, the truss robots of smart warehousing are equipped with eyes, which can automatically measure the position and size information of the plates, providing basic data support for the positioning of metal plates and their smart warehousing management.

[0005] In summary, studying a metal sheet positioning method based on 3D laser depth images and reflection images is an issue that needs to be solved urgently. Summary of the invention

[0006] In view of the problems existing in the prior art, the purpose of the present invention is to provide a metal plate positioning method based on 3D laser depth images and reflection images. According to the characteristics of metal plate material and positioning detection, the measured metal plate position and size information is sent to the MES production management system to realize intelligent warehousing management of metal plates.

[0007] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0008] A metal plate positioning method based on 3D laser depth image and reflection image comprises the following steps:

[0009] S1, images and data acquisition;

[0010] S2, pretreatment;

[0011] S3, pallet identification;

[0012] S4, edge identification;

[0013] S5, plate identification, the plate identification includes:

[0014] S51, plate pre-selected area extraction;

[0015] S52, plate area extraction;

[0016] S6, data sending.

[0017] Furthermore, in step S1, a split customized 3D line scan camera is used, and its measurement field of view and accuracy can be adjusted according to the installation position of the laser and the camera body; the camera body is driven by a truss robot to move along the length direction of the plate, forming a 3D laser depth image and reflection image of the plate, and calculating the relative installation position of the laser and the camera; respectively, continuously collecting After the acquisition is completed, all images are stitched together in the vertical direction to form a complete depth image and a complete reflection image.

[0018] Furthermore, in step S2, the complete reflection image is processed to eliminate the influence of the dark area, and the corresponding depth information is extracted from the depth image using the extracted area.

[0019] Further, in step S3, the depth threshold of the tray is first used and , perform threshold processing on the complete depth image; through the extracted ROI area and the threshold Compare and determine whether the processed plate is on the pallet. If it is greater than the threshold , the processed panels are located on pallets.

[0020] Further, in step S4, the complete reflection image is first scaled to a size of , the purpose is to balance including but not limited to computer computing power, hardware resources and processing time, and then use the edge extraction neural network model to perform semantic segmentation on the reflection image to extract the edge contour of the image.

[0021] Furthermore, in step S51, the following steps are included:

[0022] S511, edge contour extraction:

[0023] S5111, the edge contour extracted in step S4 is converted into the edge contour using the contour length ,area Screen and filter out interference areas;

[0024] S5112, converting the edge contour processed in step S5111 into a sub-pixel contour, merging the collinear contours and closing the adjacent contours using the straight line fitting principle;

[0025] S5113, the rectangular structure element of the contour is The closing operation fills the gaps in the contour and forms multiple rectangular closed areas;

[0026] S512, Filtering and Screening:

[0027] S5121, the edge contours in the plurality of rectangular closed areas processed in step S5113 are processed using the contour length ,area ,width and height Screen and filter out areas that do not meet the conditions;

[0028] S5122, calculate the edge area of ​​each rectangular closed area respectively And the area of ​​the black hole in the region , and calculate , filter out Interference area less than 5.0;

[0029] S513, to ensure the accuracy of edge contour extraction, the edge contour extracted in step S5111 is intersected with the edge contour selected in step S5122 to eliminate the influence of morphological operation when selecting the edge contour;

[0030] S514, performing a complement operation on the edge contour processed in step S513, thereby converting the edge analysis problem into the analysis problem of the area segmented by the edge contour, that is, converting the pre-selected edge contour area of ​​the plate into the pre-selected area of ​​the plate;

[0031] S515: The pre-selected area of ​​the plate in step S514 is rectified by the rectangular and convexity The operation is performed to filter out the interference area;

[0032] S516, calculating the width of the pre-selected areas of the plate processed in step S515 ,high and contour length , and find the ratio of the perimeter of the region to the length of the contour , then filter Pre-selected area of ​​the plate with a value greater than 0.85;

[0033] S517, the plates are stacked in multiple layers, and the neighboring areas are searched for all pre-selected areas. pixels, only the area with the highest height is selected from the neighboring area, which is the pre-selected area of ​​the topmost board;

[0034] S518, classifying the plate pre-selection areas in the horizontal direction, and selecting the plate pre-selection areas with the highest height in the same area;

[0035] S519, non-board area screening: set the minimum threshold of board height for RGV and pallet for screening and , eliminating the interference of fake plate pre-selected area.

[0036] Further, in step S52, the coordinates and depth values ​​of the four corner points of the pre-selected area of ​​the plate in step 519 are respectively obtained according to the following formulas. The depth value is extracted from the depth image using the corner point area, and the image coordinates are converted into world coordinates:

[0037] ;

[0038] ;

[0039] ;

[0040] in , , is the world coordinate, is the column where the pixel is located, is the row where the pixel is located, is the pixel depth value, is the scaling parameter of the laser line along the horizontal direction, is the offset parameter of the laser line in the horizontal direction, is the scaling parameter of the laser line along the moving direction, is the scaling parameter of the laser line along the vertical direction, is the bias parameter of the laser line along the vertical direction;

[0041] In step S6, the world coordinates of the four corner points of all the plates in step S52 are sent to the MES production management system.

[0042] Furthermore, in step S1, the relative installation positions of the laser and the camera are calculated by the following formula:

[0043] ;

[0044] ;

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] in, is the chip width, is the chip height, is the number of pixel rows on the chip, is the camera focal length, is the vertical angle of view, is the horizontal angle of view, is the number of rows on the upper boundary of the vertical angle of the field of view, is the number of rows below the vertical boundary of the field of view, It is the angle above the vertical center line of the field of view ROI, It is the angle below the vertical center line of the field of view ROI. is the target field of view width, is the target field of view, is the distance of the camera chip relative to the target, is the distance between the laser and the camera, is the camera installation angle, is the fan angle of the laser, is the installation height of the laser, is the installation height of the camera.

[0052] Furthermore, in step S1, the pulse signal of the encoder is connected to the camera body, so that the image information of the camera body is bound to the encoder position:

[0053] ;

[0054] Among them, the encoder position resolution , The maximum height of a single image taken by the camera. is the length of the scanning area of ​​the camera body.

[0055] Further, in step S3, if it is less than the threshold , using the depth threshold of RGV and Perform threshold processing on the complete depth image, and the area of ​​the extracted ROI area is larger than the threshold , the plate to be processed is located on the RGV, otherwise there is no ROI area of ​​the plate to be processed;

[0056] After the plate ROI area is extracted, the complete reflection image is extracted using this area, further reducing the optional area for processing the plate and the amount of data for subsequent image processing.

[0057] The present invention has at least the following beneficial effects:

[0058] (1) It provides a reliable and stable solution for visual scanning and imaging of metal sheets and provides the hardware foundation for visual positioning of sheets.

[0059] (2) The characteristics of the depth image and reflection image of the metal sheet are fully utilized and combined to extract the area of ​​the top metal sheet, fully considering the various problems encountered in the process of metal sheet area extraction.

[0060] (3) It provides a highly flexible, low-cost, high-efficiency, and high-stability visual positioning solution for the intelligent warehousing industry of metal sheets. It liberates the labor force, improves the labor shortage problem, and saves the company's warehousing costs and space. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the algorithm flow of the present invention;

[0062] Figure 2 The figure is a schematic diagram of the overall scheme;

[0063] Figure 3 This is a schematic diagram of camera field of view mapping;

[0064] Figure 4 This is a schematic diagram of the plate placement;

[0065] Figure 5 Schematic diagram of the plate edge profile. DETAILED DESCRIPTION

[0066] The present invention will be further described below in conjunction with specific embodiments and drawings, taking this embodiment as an example.

[0067] The inventors of the present application have discovered that in the intelligent warehousing technology industry for sheet metal, there has been no mature and reliable visual positioning method due to the wide variety of metal sheet materials and different sizes, resulting in the fact that fully automated and intelligent warehousing solutions have not been put into practical use. At present, the processes of sheet metal entry, exit, turnover, and transportation require manual lifting operations, and the measurement, coding, and entry of sheet metal information also rely mainly on manual labor. In order to save floor space, brackets are generally stacked layer by layer. When the required sheet metal is located in the middle bracket, manual stacking operations are required, which results in low efficiency and high safety risks.

[0068] Considering the difficulties and technical defects in the above industries, the inventors of this application, combined with relevant experience in industrial automation, have proposed a metal sheet positioning method and device based on 3D laser depth images and reflection images in terms of metal sheet visual positioning. Figures 1 to 5 shown.

[0069] Since the field of view of metal plate measurement is relatively large, the width of the field of view is about 2000mm, and the field of view length (the length of the camera body scanning area) is about 7500mm. Conventional cameras can no longer meet the measurement application, such as small field of view, large influence of ambient light, and inability to obtain height information, so a split customized 3D line scan camera is used. The advantage is that the laser can be selected according to the characteristics of the plate. Tests have found that a laser with a wavelength of 638nm and a power of 2W is more suitable. The measurement field of view and accuracy can be adjusted according to the installation position of the laser and the camera body. Since the intensity of the laser emitted by the laser is much higher than the surrounding ambient light, the interference of the surrounding ambient light can be eliminated during measurement. The truss robot is used to drive the camera body to move along the length direction of the plate, and a 3D laser depth image and reflection image can be formed for the plate. The depth image contains the distance from each point on the surface of the object to the camera, and the reflection image contains the intensity of the laser reflected back from the surface of the object.

[0070] The laser and camera are located directly above the center of the plate width, according to the formula:

[0071] ;

[0072] ;

[0073] ;

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] The relative installation position of the laser and the camera can be calculated, where the chip width is , the chip height is , the number of pixel rows of the chip is , the camera focal length is , the vertical angle of view is , the horizontal angle of view is , the number of rows on the upper boundary of the vertical angle of the field of view is , the number of rows below the vertical angle of the field of view is , the angle above the vertical center line of the field of view ROI is , the angle below the vertical center line of the field of view ROI is , target field of view width , target field of view depth , the camera is installed at an angle of , the fan angle of the laser is .

[0080] The distance between the camera chip and the target is obtained as , the distance between the laser and the camera is , the installation height of the laser is , the camera is installed at a height of .

[0081] The above installation dimensions can meet the requirements of plate width 2000mm and depth of field 300mm. The pulse signal of the encoder needs to be connected to the camera so that the camera's image information is bound to the encoder position. The encoder position resolution , the maximum height of a single image taken by the camera .

[0082] The image acquisition module needs to continuously acquire the depth image and the reflection image respectively. images are required to cover the entire metal sheet area, where , find the smallest positive integer After the acquisition is completed, all images need to be stitched vertically to form a complete image. At this time, a complete depth image and a complete reflection image are obtained. The general image size is 2560×23280.

[0083] The preprocessing module needs to process the reflected image first, eliminate the influence of the dark area, and use the extracted area to extract the available depth information in the depth image.

[0084] The pallet recognition module first uses the depth threshold of the pallet and , threshold processing is performed on the depth image, and the area of ​​the extracted ROI area is greater than a certain threshold , it can be considered that the processed plate is located on the pallet. When the processed ROI area is smaller than this threshold, the depth threshold of RGV is used. and , threshold processing is performed on the depth image, and the area of ​​the extracted ROI area is greater than the threshold , it can be considered that the plate to be processed is located on the RGV, otherwise there is no plate ROI area to be processed. When the plate ROI area is extracted, this area can be used to extract the reflection image, further reducing the optional area of ​​the plate to be processed and the amount of data in the subsequent image processing.

[0085] The edge recognition module needs to scale the above reflection image to a suitable size first. ,in , , which must be a multiple of 16. This size is determined by balancing factors such as computer computing power, hardware resources, and processing time. Then, a dedicated edge extraction neural network model (such as the Compact model and the Enhanced model) is used to perform semantic segmentation on the reflection image and extract the edge contour of the image. Compared with traditional algorithms, such as image edge detection based on the Canny algorithm, and regional segmentation schemes based on regional clustering and regional growth, the recognition accuracy, stability, and accuracy are much higher.

[0086] Main steps of plate identification module:

[0087] The edge contour extracted above is analyzed using the contour length ,area Screening is performed to filter out smaller interference areas.

[0088] The above edge contours are converted into sub-pixel contours, and the collinear contours are merged and the adjacent contours are closed using the straight line fitting principle, such as least squares straight line fitting, gradient descent straight line fitting and Hough transform straight line fitting.

[0089] The rectangular structure element of the contour is The closing operation is used to fill the small gaps within the contour to form rectangular closed areas.

[0090] For the above edge contour, use the contour length ,area ,width ,high Filter out areas that do not meet the conditions.

[0091] Calculate the edge area of ​​each region separately And the area of ​​the black hole in the region , and calculate , filter out Interference area, generally the black hole area of ​​the pre-selected area of ​​the plate is much larger than the edge area, then the remaining edge contour area is the pre-selected edge contour area of ​​the plate.

[0092] In order to ensure the accuracy of edge contour extraction, the edge contour extracted from the smaller interference area is filtered out. The intersection operation is performed on the edge contours screened by the interference area, which can eliminate the influence of morphological operations when screening the edge contours.

[0093] By performing a complement operation on the edge contour, the edge analysis problem is converted into an analysis problem of the area segmented by the edge contour, that is, the pre-selected edge contour area of ​​the plate is converted into the pre-selected area of ​​the plate.

[0094] The rectangularity of the pre-selected area of ​​the plate is , convexity Screen and filter out interference areas.

[0095] Calculate the width of the pre-selected area of ​​the plate ,high , contour length , find the ratio of the perimeter of the region to the length of the contour ,filter Preselected area of ​​the sheet.

[0096] The plates are stacked in multiple layers. When a small plate is stacked on a large plate, the edge contour of the small plate may divide the area of ​​the large plate into multiple rectangular areas. At this time, the divided area of ​​the large plate is invalid. You can search for neighboring areas for all pre-selected areas. pixels, only the area with the highest height is selected from the neighboring area, which is the pre-selected area of ​​the topmost board.

[0097] Only one stack of plates will be placed horizontally in the same area on the RGV or pallet. Sometimes, there will be multiple non-adjacent pre-selected areas in the horizontal direction due to the problem of plate stacking. The above plate pre-selected areas are classified in the horizontal direction, and the plate pre-selected areas in the same area are screened with the highest height.

[0098] Due to the structural characteristics of RGV and pallet, some edge contours on them will also be extracted when the edge contour of the plate is extracted. The above plate pre-selection area may contain RGV and pallet areas. The minimum threshold of plate height can be set for RGV and pallet respectively for screening, which can effectively eliminate the interference of false plate pre-selection areas, such as , .

[0099] The coordinates and depth values ​​of the four corner points of the pre-selected area of ​​the plate are obtained respectively. The depth value can be extracted from the depth image using the corner point area according to the formula:

[0100] ;

[0101] ;

[0102] ;

[0103] Convert image coordinates to world coordinates, where , , is the world coordinate, and the column where the pixel is located is , the behavior of the pixel , the pixel depth value is , the scaling parameter of the laser line along the horizontal direction is , the offset parameter of the laser line in the horizontal direction is , the scaling parameter of the laser line along the moving direction is , which is the resolution of the encoder, and the scaling parameter of the laser line along the vertical direction is , the bias parameter of the laser line along the vertical direction is .

[0104] The world coordinates of the four corner points of all the above-mentioned plates are sent to the MES production management system through the Http protocol of the data sending module.

[0105] Multiple RGV workstations can be arranged according to the production rhythm, and the truss robot can carry the camera and shuttle back and forth between different RGV workstations.

[0106] For some highly reflective metal plates, if they are not treated, their diffuse reflection performance to laser is poor, which will affect the scanning imaging. At this time, you can spray the developer to increase the surface roughness of the plate and improve the imaging quality.

[0107] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced to each other.

[0108] The terms "upper", "lower", "outer side", "inner side", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish the relative relationship in position if they exist, and do not need to be qualitative. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0109] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A metal sheet positioning method based on 3D laser depth image and reflection image, characterized in that: The steps include: S1, images and data acquisition; S2, pretreatment; S3, pallet identification; S4, edge identification; S5, plate identification, the plate identification includes: S51, plate pre-selected area extraction; S52, plate area extraction; S6, data sending; In step S1, a split customized 3D line scan camera is used, and its measurement field of view and accuracy can be adjusted according to the installation position of the laser and the camera body; the truss robot is used to drive the camera body to move along the length direction of the plate, to form a 3D laser depth image and reflection image of the plate, and to calculate the relative installation position of the laser and the camera; respectively, continuous acquisition A depth image and a reflection image are collected to cover the entire metal sheet area. After the collection is completed, all images are stitched in sequence in the vertical direction to form a complete depth image and a complete reflection image. In step S4, the complete reflection image is first scaled to a size of , the purpose is to balance including but not limited to computer computing power, hardware resources and processing time, and then use the edge extraction neural network model to perform semantic segmentation on the reflection image to extract the edge contour of the image; In step S51, the following steps are included: S511, edge contour extraction: S5111, the edge contour extracted in step S4 is converted into the edge contour using the contour length ,area Screen and filter out interference areas; S5112, converting the edge contour processed in step S5111 into a sub-pixel contour, merging the collinear contours and closing the adjacent contours using the straight line fitting principle; S5113, the rectangular structure element of the contour is The closing operation fills the gaps in the contour and forms multiple rectangular closed areas; S512, Filtering and Screening: S5121, the edge contours in the plurality of rectangular closed areas processed in step S5113 are processed using the contour length ,area ,width and height Screen and filter out areas that do not meet the conditions; S5122, calculate the edge area of ​​each rectangular closed area respectively And the area of ​​the black hole in the region , and calculate , filter out Interference area less than 5.0; S513, to ensure the accuracy of edge contour extraction, the edge contour extracted in step S5111 is intersected with the edge contour selected in step S5122 to eliminate the influence of morphological operation when selecting the edge contour; S514, performing a complement operation on the edge contour processed in step S513, thereby converting the edge analysis problem into the analysis problem of the area segmented by the edge contour, that is, converting the pre-selected edge contour area of ​​the plate into the pre-selected area of ​​the plate; S515: The pre-selected area of ​​the plate in step S514 is rectified by the rectangular and convexity The operation is performed to filter out the interference area; S516, calculating the width of the pre-selected areas of the plate processed in step S515 ,high and contour length , and find the ratio of the perimeter of the region to the length of the contour , then filter Pre-selected area of ​​the plate with a value greater than 0.85; S517, the plates are stacked in multiple layers, and the neighboring areas are searched for all pre-selected areas. pixels, only the area with the highest height is selected from the neighboring area, which is the pre-selected area of ​​the topmost board; S518, classifying the plate pre-selection areas in the horizontal direction, and selecting the plate pre-selection areas with the highest height in the same area; S519, non-board area screening: set the minimum threshold of board height for RGV and pallet for screening and , eliminating the interference of fake plate pre-selected area.

2. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 1, characterized in that: In step S2, the complete reflection image is processed to eliminate the influence of the dark area, and the corresponding depth information is extracted from the depth image using the extracted area.

3. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 2, characterized in that: In step S3, the depth threshold of the pallet is first used and , perform threshold processing on the complete depth image; through the extracted ROI area and the threshold Compare and determine whether the processed plate is on the pallet. If it is greater than the threshold , the processed panels are located on pallets.

4. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 1, characterized in that: In step S52, the coordinates and depth values ​​of the four corner points of the plate pre-selected area in step 519 are respectively obtained according to the following formulas. The depth value is extracted from the depth image using the corner point area, and the image coordinates are converted into world coordinates: ； in , , is the world coordinate, The column where the pixel is located, is the row where the pixel is located, is the pixel depth value, is the scaling parameter of the laser line along the horizontal direction, is the offset parameter of the laser line in the horizontal direction, is the scaling parameter of the laser line along the moving direction, is the scaling parameter of the laser line along the vertical direction, is the bias parameter of the laser line along the vertical direction; In step S6, the world coordinates of the four corner points of all the plates in step S52 are sent to the MES production management system.

5. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 1, characterized in that: In step S1, the relative installation positions of the laser and the camera are calculated by the following formula: ； ； ； ； ； ； ； ； in, is the chip width, is the chip height, is the number of pixel rows on the chip, is the camera focal length, is the vertical angle of view, is the horizontal angle of view, is the number of rows on the upper boundary of the vertical angle of the field of view, is the number of rows at the lower boundary of the vertical angle of the field of view, It is the angle above the vertical center line of the field of view ROI, It is the angle below the vertical center line of the field of view ROI. is the target field of view width, is the target field of view, is the distance of the camera chip relative to the target, is the distance between the laser and the camera, is the camera installation angle, is the fan angle of the laser, is the installation height of the laser, is the installation height of the camera.

6. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 1, characterized in that: In step S1, the pulse signal of the encoder is connected to the camera body so that the image information of the camera body is bound to the encoder position; ； Among them, the encoder position resolution , The maximum height of a single image taken by the camera. is the length of the scanning area of ​​the camera body.

7. The metal sheet positioning method based on 3D laser depth image and reflection image as claimed in claim 6, characterized in that: In step S3, if it is less than the threshold , using the depth threshold of RGV and Perform threshold processing on the complete depth image, and the area of ​​the extracted ROI area is larger than the threshold , the plate to be processed is located on the RGV, otherwise there is no ROI area of ​​the plate to be processed; After the plate ROI area is extracted, the complete reflection image is extracted using this area, further reducing the optional area for processing the plate and the amount of data for subsequent image processing.

Images