CN110619330A - Recognition model training method and device, computer equipment and recognition method - Google Patents
Recognition model training method and device, computer equipment and recognition method Download PDFInfo
- Publication number
- CN110619330A CN110619330A CN201910880226.8A CN201910880226A CN110619330A CN 110619330 A CN110619330 A CN 110619330A CN 201910880226 A CN201910880226 A CN 201910880226A CN 110619330 A CN110619330 A CN 110619330A
- Authority
- CN
- China
- Prior art keywords
- identification code
- vehicle identification
- image
- code area
- character
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a training method and device for a recognition model of a vehicle recognition code, computer equipment and a recognition method. The training method comprises the steps of obtaining a plurality of vehicle identification code area images, and marking the positions and the types of characters in the vehicle identification code area images; performing data expansion operation on the marked vehicle identification code area images to generate expanded images corresponding to the vehicle identification code area images, and determining the positions and the types of characters in the expanded images; generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the expansion images; and training a vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image. The technical problem that the vehicle identification code cannot be accurately detected in the traditional technology is solved.
Description
Technical Field
The present application relates to the field of vehicle detection, and in particular, to a method and an apparatus for training a recognition model of a vehicle identification code, a computer device, and a recognition method.
Background
With the continuous development of social economy and the continuous improvement of the living standard of people, the quantity of motor vehicles in cities is rapidly increased. The workload of annual inspection of motor vehicles is also rapidly increased. The detection and identification of the vehicle identification code are one of important items for vehicle annual inspection.
The Vehicle Identification Number (VIN) is a multi-character code serving as a unique Identification code of the Vehicle, and includes information such as a manufacturer, a year, a Vehicle type, a Vehicle body type and code, an engine code, an assembly location, and the like of the Vehicle. Typically, the vehicle identification code consists of a 17-digit character, comprising letters and numbers, colloquially known as a seventeen-digit code.
However, the conventional technology has the technical problem that the vehicle identification code cannot be accurately identified.
Disclosure of Invention
Therefore, it is necessary to provide a training method, an apparatus and an identification method for an identification model of a vehicle identification code, aiming at the technical problem that the vehicle identification code cannot be accurately identified in the conventional technology.
A method of training a recognition model of a vehicle identification code, the method comprising: acquiring a plurality of vehicle identification code area images, and marking the position and the category of each character in each vehicle identification code area image; performing data expansion operation on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and determining the position and the category of each character in each expanded image; generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the expansion images; and training the vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image.
According to the training method for the recognition model of the vehicle recognition code, due to the fact that the positions and the types of the characters in the image of the vehicle recognition code region are marked, effective recognition of the same adjacent characters in the image to be detected of the vehicle recognition code is achieved, the technical problem that the vehicle recognition code cannot be accurately detected in the traditional technology is solved, and the accuracy of detection of the vehicle recognition code is improved.
Drawings
FIG. 1 is a diagram illustrating an exemplary environment in which a method for training a vehicle identification code recognition model may be implemented;
FIG. 2a is a schematic flow chart illustrating a method for training a vehicle identification code recognition model according to an embodiment;
FIG. 2b is a schematic illustration of a vehicle identification code according to one embodiment;
FIG. 3a is a schematic diagram illustrating the flow of a layout transformation operation in one embodiment;
FIG. 3b is a diagram illustrating a layout transformation operation in one embodiment;
FIG. 4a is a schematic flow chart of a bend transform operation in one embodiment;
FIG. 4b is a schematic diagram of a bend transformation operation in one embodiment;
FIG. 4c is a schematic illustration of a rotation operation in one embodiment;
FIG. 5 is a diagram of a network structure of a character instance segmentation model in one embodiment;
FIG. 6 is a flow chart illustrating the process of recognizing the image of the vehicle identification code region to be recognized by using the character instance segmentation model in one embodiment;
FIG. 7 is a diagram illustrating a vehicle identification code region in an image to be detected, in accordance with an embodiment;
FIG. 8 is a schematic flow chart diagram illustrating a method for training a vehicle identification code recognition model according to one embodiment;
FIG. 9 is a block diagram showing an example of a training apparatus for a vehicle identification code recognition model;
FIG. 10 is a diagram showing an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Referring to fig. 1, a schematic diagram of an application environment according to an embodiment of the present application is provided. The application environment may include: a first computer device 110, a second computer device 120, and an image acquisition device 130. The first Computer device 110 and the second Computer device 120 refer to electronic devices with strong data storage and computation capabilities, for example, the first Computer device 110 and the second Computer device 120 may be a PC (Personal Computer) or a server. Specifically, the image acquisition device 130 acquires an image of the vehicle identification code of the vehicle to obtain an image to be detected of the vehicle identification code, and sends the image to be detected of the vehicle identification code to the first computer device 110 through network connection. Before detecting the image to be detected, a technician is required to construct the target detection model on the second computer device 120, and train the constructed target detection model through the second computer device 120. The object detection model refers to a machine learning model that segments an object of interest (such as a vehicle identification code) from an image to be detected. For example, the target detection model may be a deep learning-based SSD (single shot multi-box detection) target detection algorithm model, and the SSD may detect the vehicle identification code region in the image to be detected through a rectangular frame by using a single deep neural network.
The technician may also build a vehicle id recognition model on the second computer device 120, where the vehicle id recognition model may employ a character instance segmentation model, and train the built vehicle id recognition model through the second computer device 120. The training of the vehicle identification code recognition model comprises the following steps: the second computer device 120 acquires a plurality of vehicle identification code area images and labels the position and the category of each character in each vehicle identification code area image; performing data expansion operation on the marked vehicle identification code area images to generate expanded images corresponding to the vehicle identification code area images; generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the expansion images; and training a vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image.
The trained object detection model, the vehicle identification code recognition model, may be published from the second computer device 120 into the first computer device 110. The first computer device 110 may use the target detection model to detect an image to be detected of the vehicle identification code provided by the user, determine whether a region image of the vehicle identification code to be identified exists in the image to be detected, and if so, extract the region image of the vehicle identification code to be identified from the image to be detected; the vehicle identification code recognition model is adopted to recognize and segment the image of the vehicle identification code area to be recognized, so that the recognition and segmentation result of the vehicle identification code is obtained, and the character string corresponding to the image of the vehicle identification code area to be recognized is generated according to the recognition and segmentation result of the vehicle identification code; then, comparing a character string corresponding to the image of the vehicle identification code area to be identified with a reference character string of the vehicle identification code; and recording the recognition result of the vehicle identification code according to the comparison result. It is understood that the first computer device 110 may also take the form of a terminal, which may be an electronic device such as a cell phone, a tablet, an e-book reader, a multimedia player device, a wearable device, a PC, etc. And the terminal finishes the detection work of the image to be detected of the vehicle identification code through the target detection model and the character instance segmentation model. Of course, the first computer device 110 may also be integrated with the image acquisition device 130.
In one embodiment, as shown in fig. 2a, a training method for a vehicle identification code recognition model is provided, which is illustrated by applying the method to the second computer device 120 in fig. 1, and includes the following steps:
s210, obtaining a plurality of vehicle identification code area images, and labeling the position and the category of each character in each vehicle identification code area image.
The Vehicle Identification Number (VIN) is a multi-character code serving as a unique Identification code of the Vehicle, and includes information such as a manufacturer, a year, a Vehicle type, a Vehicle body type and code, an engine code, an assembly location, and the like of the Vehicle. The vehicle identification code region image is a partial image corresponding to the vehicle identification code extracted from the vehicle identification code image. The vehicle identification code image can be obtained by shooting the vehicle identification code through the image acquisition equipment. The vehicle identification code image can be detected by using the target detection model, whether the vehicle identification code image has the vehicle identification code area image or not is detected, and if the vehicle identification code image has the vehicle identification code area image, the image to be detected is extracted or intercepted to obtain the vehicle identification code area image. And storing the detected vehicle identification code area image in a server local to the second computer equipment or in communication connection with the second computer equipment.
Specifically, the second computer device locally stores a plurality of vehicle identification code area images in advance, and the plurality of vehicle identification code area images are locally and directly acquired from the second computer device. Alternatively, the plurality of vehicle identification code region images are stored in a server communicatively connected to the second computer device, from which the plurality of vehicle identification code region images are obtained. And then, labeling the acquired vehicle identification code area images by using an open-source Labelme labeling tool. Since each character of the same category needs to be recognized, each character in the vehicle identification code region image needs to be respectively subjected to category labeling and position labeling during labeling. In particular, the vehicle VIN code has 34 characters of "0-9", "A-N", "P" and "R-Z", and the category (class) can be represented by 0-33. When the same characters or continuous same characters exist in the vehicle VIN codes, the same characters in the vehicle VIN codes are distinguished by marking the positions and the categories of the characters.
S220, performing data expansion operation on the marked vehicle identification code area images to generate expanded images corresponding to the vehicle identification code area images, and determining the positions and the types of characters in the expanded images.
The data expansion refers to a means for increasing the number of training sets by image deformation or noise increase. The data augmentation operation includes at least one of a layout transformation, a bending transformation, a rotation, and a translation. Specifically, on one hand, the labeling work of the vehicle identification code is relatively time-consuming and the labeling cost is high, and on the other hand, in an actual situation, the probability of encountering the double-row frame numbers and the arc-shaped frame numbers is also low, so that the construction of the sample set with the double-row frame numbers and the arc-shaped frame numbers is difficult to achieve. For these reasons, data expansion operation is performed on the marked vehicle identification code region images to generate expanded images corresponding to the vehicle identification code region images, and one vehicle identification code region image is expanded into a preset number of different forms of pictures, for example, the original one picture is expanded into 10 different forms of pictures. It can be understood that, by labeling each vehicle identification code region image, a tag file (such as a JSON file) corresponding to each vehicle identification code region image is obtained, and a tag map corresponding to the vehicle identification code region image is generated according to each tag file. When the data expansion operation is performed on the vehicle identification code region image, the same data expansion operation needs to be performed on the tag map corresponding to the vehicle identification code region image to obtain the tag map corresponding to the expanded image, and the tag map corresponding to the expanded image can be converted into a tag file (such as a JSON file), so that the position and the category of each character in each expanded image are determined through the tag map corresponding to the expanded image.
And S230, generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the extended images.
Wherein the sample set is a training set of the vehicle identification code recognition model for training. Specifically, a plurality of vehicle identification code region images are acquired, and each acquired vehicle identification code region image can be used as a training set of a vehicle identification code recognition model. And performing data expansion operation on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, wherein each expanded image obtained by the data expansion operation can also be used as a training set of a vehicle identification code identification model, namely each acquired vehicle identification code area image and each expanded image obtained by expansion are constructed into a sample set of model training.
S240, training a vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the type of each character in each vehicle identification code area image and the position and the type of each character in each extended image.
And training the vehicle identification code recognition model by using the label values corresponding to the sample set and the sample set, so that the vehicle identification code recognition model has the capability of recognizing each character in the vehicle identification code area image. Specifically, a sample set composed of each vehicle identification code region image and each extended image obtained through extension is used as a training set of the model. And acquiring the position and the type of each character in each vehicle identification code area image through marking, and determining the position and the type of each character in each extended image through a tag graph corresponding to each extended image, so that a vehicle identification code recognition model is trained by utilizing the sample set of the vehicle identification code area image, the position and the type of each character in each vehicle identification code area image and the position and the type of each character in each extended image. Further, according to the difference between the result predicted by the vehicle identification code recognition model and the label, the model parameters of the vehicle identification code recognition model are adjusted and training is continued until the training stopping condition is met, and then the training is finished. Wherein the training stop condition is a condition for ending the model training. The training stopping condition may be that a preset number of iterations is reached, or that the performance index of the vehicle identification code recognition model after the model parameters are adjusted reaches a preset index.
For example, the difference between the present embodiment and the conventional technology is that the vehicle VIN code generally consists of 17-bit characters, and the characters constituting the vehicle VIN code include 34 characters of "0-9", "a-N", "P", and "R-Z". When these characters are arranged and combined to form the vehicle VIN code, a plurality of identical characters often appear in succession. As shown in fig. 2b, the vehicle VIN code is displayed as: LA9BAGGV0FHXCF 049. For the vehicle VIN code containing continuous same characters (such as GG), the conventional technology is used for identification, and the identification result is as follows: LA9BAGV0FHXCF 049. The identification result of the traditional technology is compared with the vehicle VIN code, so that the traditional technology cannot identify the vehicle VIN code containing continuous same characters, and the vehicle identification code cannot be accurately detected. In this embodiment, since the positions and the types of the characters in the vehicle identification code region image are labeled, the same adjacent characters in the vehicle identification code region image are identified by the vehicle identification code identification model, and the identification result is: LA9BAGGV0 FXCCF 049, the recognition result of this embodiment is compared with the vehicle VIN code, and it can be seen that the embodiment accurately recognizes the vehicle VIN code containing continuous same characters, so that the vehicle identification code can be accurately detected.
In the implementation, a plurality of vehicle identification code area images are obtained, and the positions and the types of characters in the vehicle identification code area images are marked; performing data expansion operation on the marked vehicle identification code area images to generate expanded images corresponding to the vehicle identification code area images, and determining the positions and the types of characters in the expanded images; generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the expansion images; and training a vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image. The method and the device realize effective recognition of the same adjacent characters in the image to be detected of the vehicle identification code, thereby solving the technical problem that the vehicle identification code cannot be accurately detected in the traditional technology and improving the accuracy of vehicle identification code detection.
In one embodiment, the data augmentation operation includes a layout transformation. As shown in fig. 3a, the data expansion operation is performed on each labeled vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and the method includes the following steps:
s310, respectively intercepting the images of the vehicle identification code areas to obtain background images which do not contain all characters;
and S320, adjusting the size of the background image.
Specifically, in order to keep agreement with the real scene of the vehicle identification code, a partial image is cut out from any position in the vehicle identification code region image that does not include each character, and the cut-out image that does not include each character is used as a background image of the extended image. In order to make the size of the background image consistent with the preset size of the expanded image, the size of the background image needs to be further adjusted. The size of the background image can be adjusted according to the size data input by the user, and the size of the background image can also be adjusted according to the pixels of the background image. Illustratively, the size of the background image is adjusted by the resize function.
S330, according to the position of each character in each vehicle identification code area image, the vehicle identification codes in each vehicle identification code area image are cut off at a preset position, and at least two parts of vehicle identification code images are obtained.
Specifically, since the positions of the characters in the vehicle identification code region images are labeled and the characters correspond to the circumscribed rectangle frame, the coordinates of the four vertices of the circumscribed rectangle frame of each character are known. The preset position can be arranged between two adjacent characters of the vehicle identification code, and the preset position is arranged between the two adjacent characters according to the actual condition. Selecting at least one group of adjacent characters, determining corresponding dividing lines between the selected adjacent characters, and cutting the vehicle identification codes in the vehicle identification code area images through the dividing lines to obtain at least two parts of vehicle identification code images.
Illustratively, through statistical analysis, the probability of occurrence of the double-row vehicle VIN codes is relatively high, and it is assumed that the vehicle identification codes in the vehicle identification code area image are divided into two segments. For any vehicle identification code area image, the vehicle identification code in the vehicle identification code area image consists of M characters, coordinates of four vertexes of a respectively circumscribed rectangular frame of the known M characters are marked, and a preset position is arranged between the Nth (1 < N < M) character and the (N + 1) th character. Determining a dividing line by using the upper right vertex and the lower right vertex of the character N circumscribing the rectangular frame; or determining a dividing line by using the upper left vertex and the lower left vertex of the (N + 1) -th character circumscribed rectangle frame; or determining a dividing line by using the upper right vertex of the circumscribed rectangle frame of the Nth character and the lower left vertex of the circumscribed rectangle frame of the (N + 1) th character; or a division line is determined by the lower right vertex of the N-th character circumscribed rectangular frame and the upper left vertex of the (N + 1) -th character circumscribed rectangular frame; and cutting off the vehicle identification code in the vehicle identification code area image by using the segmentation line to obtain two parts of vehicle identification code images. It should be noted that the vehicle identification code in the vehicle identification code region image may be cut into three or more parts by the same method in combination with the actual situation, which is not described herein again.
S340, overlapping at least two parts of vehicle identification code images to the background image after size adjustment, and generating an expanded image after layout transformation operation.
Specifically, the vehicle identification code is cut off at a preset position to obtain at least two parts of vehicle identification code images, the at least two parts of vehicle identification code images are subjected to layout transformation and are superposed on the background image after size adjustment, and an expanded image after layout transformation operation is generated. For example, in the case where the vehicle identification code includes 17 characters, after statistically analyzing the picture data of a large number of double-row vehicle identification codes, it is found that the first row of the double-row vehicle identification codes is generally 9 characters and the second row is 8 characters. Acquiring a single-row vehicle identification code area image, as shown in fig. 3b, truncating the single-row vehicle identification code according to the coordinates of the lower right vertex of the circumscribed rectangular frame of the 9 th character and the coordinates of the upper left vertex of the circumscribed rectangular frame of the next character to obtain two partial images, and pasting the two partial images on the background image after size adjustment to obtain an expanded image comprising double rows of vehicle identification codes.
In this embodiment, the layout of the vehicle identification code in the vehicle identification code region image is changed through the layout transformation operation, at least two obtained vehicle identification code images are superimposed on the background image after the size adjustment to generate the expanded image after the layout transformation operation, and the expanded image obtained through the layout transformation operation is used for training the vehicle identification code recognition model, so that the generalization capability of the vehicle identification code recognition model is improved, and more types of vehicle identification codes can be recognized.
In one embodiment, the generating the expanded image after the layout transformation operation by superimposing at least two parts of the vehicle identification code image onto the resized background image comprises: respectively adjusting the sizes of at least two parts of vehicle identification code images; and superposing the at least two parts of vehicle identification code images after the size adjustment to the background image after the size adjustment to generate an expanded image after the layout transformation operation.
Specifically, after the vehicle identification codes in the respective vehicle identification code area images are cut, the sizes of the acquired respective portions of the vehicle identification code images need to be adjusted. When the size of each part of the vehicle identification code image is adjusted, new pixel points are inevitably introduced. In order to avoid introducing new pixel values into the label graph, a nearest neighbor interpolation mode should be selected when the interpolation operation is performed. And superposing the vehicle identification code images of all the parts after the size adjustment on the background image after the size adjustment to generate an expanded image after the layout transformation operation.
In one embodiment, the data expansion operation includes a warping transformation. As shown in fig. 4a, the data expansion operation is performed on each marked vehicle identification code region image to generate an expanded image corresponding to each vehicle identification code region image, and the method includes the following steps:
s410, respectively determining circumscribed rectangular frames surrounding the vehicle identification codes in the vehicle identification code area images according to the positions of the first character and the last character in the vehicle identification code area images;
and S420, acquiring coordinates of each vertex of each circumscribed rectangular frame and central coordinates of each vehicle identification code area image.
The image distortion correction method comprises the steps of performing curvature transformation operation on an image, wherein the curvature transformation operation is based on the idea of image distortion correction, wherein the distortion correction is actually to place pixel points in the image at positions where the pixel points should be theoretically located, and the image is corrected by solving the mapping relation between the distance between the pixel points in the image and the center of a lens and the actual distance. Specifically, the positions of the characters in the vehicle identification code region images are marked, and the characters correspond to the circumscribed rectangle frame, so that the coordinates of four vertexes of the circumscribed rectangle frame of each character can be known. For any vehicle identification code area image, the circumscribed rectangular frame surrounding the vehicle identification code can be determined according to the coordinates of the top left vertex and the top left vertex of the circumscribed rectangular frame of the first character and the coordinates of the top right vertex and the top right vertex of the circumscribed rectangular frame of the last character in the vehicle identification code area image, and each character is located in the circumscribed rectangular frame surrounding the vehicle identification code. And four vertexes of the circumscribed rectangular frame surrounding the vehicle identification code are respectively the coordinates of the upper left vertex and the upper left vertex of the circumscribed rectangular frame of the first character and the coordinates of the upper right vertex of the circumscribed rectangular frame of the last character. In this embodiment, the center of the lens is recorded as the center coordinate of the vehicle identification code region image, and the center coordinate of the vehicle identification code region image is acquired for the purpose of performing the bending transformation operation.
And S430, determining a bending coefficient corresponding to the bending transformation according to the actual radian of the arc-shaped vehicle identification code.
Specifically, bending operation is carried out on the vehicle identification code area image by trying different bending coefficients, the bending amplitude obtained by the bending operation is compared with the actual radian of the arc-shaped vehicle identification code, and a group of coefficients with the bending amplitude most similar to the actual arc-shaped vehicle frame number are selected as the bending coefficients.
And S440, performing bending transformation on the vehicle identification code area images according to the coordinates and the center coordinates of the vertexes and the bending coefficients to generate the extended images after the bending transformation operation.
Specifically, as shown in fig. 4b, after coordinates of each vertex of the circumscribed rectangular frame surrounding the vehicle identification code and center coordinates of the vehicle identification code region image are acquired for any vehicle identification code region image, the vehicle identification code region image may be subjected to bending transformation according to the coordinates and the center coordinates of each vertex and a bending coefficient, so as to generate an extended image after the bending transformation operation. Furthermore, when the bending change of the pixel points is carried out, border crossing protection is needed, namely the pixel points after the bending change cannot exceed the size range of the image of the vehicle identification code area, so that incomplete characters are avoided.
In this embodiment, the shape of the vehicle identification code in the vehicle identification code region image is changed through the bending transformation operation, and the expanded image after the bending transformation operation is used for training the vehicle identification code identification model, so that the generalization capability of the vehicle identification code identification model can be improved, and the vehicle identification code with more shapes can be identified.
In one embodiment, the data augmentation operation includes at least one of rotation, translation. Performing data expansion operation on the marked vehicle identification code area images to generate expanded images corresponding to the vehicle identification code area images, wherein the expanded images comprise: rotating the marked vehicle identification code region images by a preset angle to generate an expanded image after rotating operation; and/or translating the marked images of the vehicle identification code areas according to a preset direction to generate an expanded image after translation operation.
Specifically, as shown in fig. 4c, since the model has rotation invariance, data expansion can be performed by performing small-angle rotation on each labeled vehicle identification code region image. In order to avoid overfitting of the model to the learning of the character position information, translation operations in different directions are carried out on the marked vehicle identification code region images to carry out data expansion.
In one embodiment, the vehicle identification code recognition model employs a character instance segmentation model. Fig. 5 shows a network structure diagram of a character instance segmentation model. The character instance segmentation model calls Resnet-101 as a base network, uses an FPN (FeaturePyrenamid networks) structure, and uses a semantic segmentation function. The method comprises the following steps of adopting a character instance segmentation model to identify an image of a vehicle identification code area to be identified, wherein the character instance segmentation model needs to finish three things: character detection, character classification, and character segmentation on a pixel level hierarchy. Specifically, as shown in fig. 5, the vehicle identification code region image to be recognized is input to the input layer of the character instance segmentation model for some processing, and the output of the input layer is transmitted to the base network connected to the input layer. And extracting a Feature image (Feature Map) corresponding to the vehicle identification code area image to be identified through the base network. Feature images output by the base network are transmitted to a target detection candidate frame generation network and an ROI (Region of Interest) Align layer, and the ROI Align layer is connected to the target detection candidate frame generation network. Two branches are connected behind the ROI Align layer, and the two branches are respectively a first branch and a second branch. Wherein the first branch comprises a first sub-branch and a second sub-branch. The first subbranch comprises a frame regression layer, and the target detection is carried out on the image of the vehicle identification code area to be identified through the first subbranch, so that each character and the position of each character in the image of the vehicle identification code area to be identified are determined. The second subbranch comprises a classifier, and the target classification is carried out on the image of the vehicle identification code area to be recognized through the second subbranch, so that the category of each character in the image of the vehicle identification code area to be recognized is determined. The second branch comprises a Mask layer (also called a Mask layer), and the pixel-level target segmentation is carried out on the vehicle identification code area image to be recognized through the second branch to obtain a plurality of pixel-level segmentation objects. Since the pixel-level segmentation object includes each character constituting the vehicle identification code and other backgrounds, each character corresponding to the vehicle identification code is selected from the plurality of pixel-level segmentation objects according to each recognized character and the category of each character. And sequencing the recognized characters according to the positions and the types of the characters in the vehicle identification code area image to be recognized to generate a character string corresponding to the vehicle identification code area image, wherein the generated character string is the recognized vehicle identification code.
In one embodiment, the application of the vehicle identification code recognition model is illustrated as an example of a character instance segmentation model and applied in the first computer device 110 in FIG. 1. As shown in fig. 6, the method for identifying the vehicle identification code in the image of the vehicle identification code area to be identified by using the character instance segmentation model comprises the following steps:
s610, obtaining the area image of the vehicle identification code to be identified and the reference character string of the vehicle identification code.
The to-be-identified vehicle identification code region image is a partial image which is extracted from the to-be-detected image and corresponds to the vehicle identification code. The image to be detected is an image which is obtained by shooting the vehicle identification code area through image acquisition equipment and needs to be subjected to vehicle identification code detection. The image to be detected can only comprise the image of the vehicle identification code area to be identified, can also not comprise the image of the vehicle identification code area to be identified and only comprises some background images, and can also simultaneously comprise the image of the vehicle identification code area to be identified and some background images. The method comprises the steps of carrying out image acquisition on a vehicle identification code of a vehicle through image acquisition equipment, sending an image to be detected of the vehicle identification code to first computer equipment in a wired connection mode or a wireless connection mode, and obtaining the image to be detected of the vehicle identification code through the first computer equipment. The image to be detected of the vehicle identification code can also be stored in advance in the first computer device locally or in a server in communication connection with the first computer device, and the first computer device obtains the image to be detected from the server locally or in communication connection with the first computer device. Since the vehicle identification code is composed of several bit characters, a reference character string of the vehicle identification code, which is a criterion for judging whether the vehicle identification code in the image to be detected passes the detection, is stored in advance. It will be appreciated that the reference string of vehicle identification codes may be pre-stored locally on the first computer device or on a server communicatively connected to the first computer device. Specifically, as shown in fig. 7, the target detection model is used to detect whether the image to be detected has the image of the vehicle identification code region to be identified, and if so, the image of the vehicle identification code region to be identified is extracted from the image to be detected. The reference string of vehicle identification codes may be obtained from a server local to the first computer device or communicatively coupled to the first computer device.
And S620, identifying and segmenting the image of the vehicle identification code area to be identified through the character instance segmentation model.
The character instance segmentation model is a machine learning model for recognizing the outline of each character in the vehicle identification code at a pixel level. Specifically, inputting a to-be-recognized vehicle identification code area image detected by a target detection model into a character instance segmentation model, performing target detection on the to-be-recognized vehicle identification code area image through the character instance segmentation model, and determining each character and the position of each character in the to-be-recognized vehicle identification code area image; carrying out target classification on the vehicle identification code area image to be identified through a character instance segmentation model, and determining the category of each character in the vehicle identification code area image to be identified; performing pixel-level target segmentation on the vehicle identification code region image to be identified through a character instance segmentation model to obtain a plurality of pixel-level segmentation objects; and selecting each character corresponding to the vehicle identification code from the plurality of pixel level segmentation objects according to each character and the category of each character in the vehicle identification code area image.
And S630, generating a character string corresponding to the vehicle identification code area image to be identified according to the identification segmentation result of the vehicle identification code area image to be identified.
Specifically, the vehicle identification code area image to be identified is identified by using the character instance segmentation model, and the type of characters and the positions of the characters in the vehicle identification code area image to be identified are identified, so that the characters are sequenced according to the positions and types of the characters in the vehicle identification code area image to be identified, and the character string corresponding to the vehicle identification code area image is generated.
S640, comparing a character string corresponding to the image of the vehicle identification code area to be identified with a reference character string of the vehicle identification code;
and S650, recording the identification result of the vehicle identification code according to the comparison result.
Specifically, each character of the vehicle identification code in the vehicle identification area image to be identified is identified by using the character instance segmentation model, and a character string corresponding to the vehicle identification code area image is generated. In order to determine whether the image to be detected passes the audit, it is necessary to determine whether the character string corresponding to the image of the vehicle identification code region to be identified is consistent with the reference character string of the vehicle identification code. And comparing the character string corresponding to the image of the vehicle identification code area to be identified with the reference character string of the vehicle identification code, and recording the identification result of the vehicle identification code according to the comparison result. If the two are consistent, the recognition result of the vehicle identification code is recorded as passing detection, and if the two are not consistent, the recognition result of the vehicle identification code is recorded as not passing detection.
In the embodiment, the adjacent same characters in the image of the vehicle identification code region can be identified through the character embodiment segmentation model, the technical problem that the vehicle identification code cannot be accurately identified in the traditional technology is solved, and the accuracy of vehicle identification code detection is improved.
In one embodiment, as shown in fig. 8, after performing a data expansion operation on each labeled vehicle identification code region image to generate an expanded image corresponding to each vehicle identification code region image, the method further includes the following steps:
s810, detecting whether the extended image contains incomplete characters;
and S820, if yes, deleting the extended image containing the incomplete character.
Specifically, after performing data expansion operations such as layout conversion, bending conversion, rotation, translation, and the like on the vehicle identification code region image, partial characters in the expanded image may be incomplete, that is, incomplete characters. If the sample set comprises the image samples containing the incomplete characters, the training of the subsequent vehicle identification code recognition model is adversely affected, and therefore the recognition effect is affected. Therefore, it is necessary to detect whether the extended image includes the missing character, and if so, delete the extended image including the missing character.
In one embodiment, the present application provides a training method for a vehicle identification code recognition model, and the vehicle identification code recognition model employs a character instance segmentation model. The method comprises the following steps:
s802, obtaining a plurality of vehicle identification code area images, and labeling the position and the type of each character in each vehicle identification code area image.
S804, respectively intercepting the images of the vehicle identification code areas to obtain background images which do not contain all characters.
And S806, adjusting the size of the background image.
And S808, according to the position of each character in each vehicle identification code area image, truncating the vehicle identification code in each vehicle identification code area image at a preset position to obtain at least two parts of vehicle identification code images.
And S810, respectively adjusting the sizes of at least two parts of vehicle identification code images.
And S812, overlapping the at least two parts of vehicle identification code images after the size adjustment to the background image after the size adjustment, and generating an expanded image after the layout conversion operation.
S814, determining circumscribed rectangular frames surrounding the vehicle identification codes in the vehicle identification code area images respectively according to the positions of the first character and the last character in the vehicle identification code area images.
And S816, obtaining the coordinates of each vertex of each circumscribed rectangular frame and the center coordinates of each vehicle identification code area image.
And S818, determining a bending coefficient corresponding to the bending transformation according to the actual radian of the arc-shaped vehicle identification code.
And S820, according to the coordinates and the center coordinates of each vertex, combining the bending coefficients to perform bending transformation on each vehicle identification code area image, and generating an extended image after the bending transformation operation.
And S822, rotating the marked vehicle identification code area images by a preset angle to generate an expanded image after rotating operation. And/or
S824, translating the marked vehicle identification code region images according to a preset direction to generate an expanded image after translation operation.
S826, detecting whether the expanded image after layout transformation, bending transformation, rotation and translation operation contains incomplete characters.
And S828, if yes, deleting the extended image containing the incomplete character.
And S830, generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the reserved expansion images.
S832, training a vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image.
Labeling each vehicle identification code region image to obtain a tag file (such as a JSON file) corresponding to each vehicle identification code region image, and generating a tag map corresponding to the vehicle identification code region image according to each tag file. When performing data expansion operations such as layout transformation, bending transformation, rotation, translation and the like on the vehicle identification code region image, the same data expansion operation needs to be performed on the tag map corresponding to the vehicle identification code region image to obtain the tag map corresponding to the expanded image, and the tag map corresponding to the expanded image can be converted into a tag file (such as a JSON file), so that the position and the category of each character in each expanded image are determined through the tag map corresponding to the expanded image.
S834, acquiring the area image of the vehicle identification code to be identified and the reference character string of the vehicle identification code.
And S836, identifying and segmenting the vehicle identification code area image to be identified through the character instance segmentation model.
And S838, generating a character string corresponding to the vehicle identification code area image to be identified according to the identification and segmentation result of the vehicle identification code area image to be identified.
And S840, comparing the character string corresponding to the vehicle identification code area image to be identified with the reference character string of the vehicle identification code.
And S842, recording the identification result of the vehicle identification code according to the comparison result.
It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, the steps are not necessarily executed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the above embodiments may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or the stages is not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a part of the sub-steps or the stages of other steps.
In one embodiment, the present application provides a method of identifying a vehicle identification code, the method comprising: acquiring a vehicle identification code area image to be identified; and inputting the area image of the vehicle identification code to be identified into the identification model of the vehicle identification code obtained by training in any one of the embodiments for identification, and outputting the identification result of the area image of the vehicle identification code to be identified.
In one embodiment, the present application provides a training apparatus for a vehicle identification code recognition model, as shown in fig. 9, the training apparatus 900 includes an obtaining module 910, a data expansion module 920, a sample set generating module 930, and a model training module 940. Wherein:
an obtaining module 910, configured to obtain a plurality of vehicle identification code region images, and label a position and a category of each character in each vehicle identification code region image;
the data expansion module 920 is configured to perform data expansion operation on the marked vehicle identification code region images, generate expanded images corresponding to the vehicle identification code region images, and determine positions and categories of characters in the expanded images;
a sample set generating module 930, configured to generate a sample set of the vehicle identification code region image according to each vehicle identification code region image and each extended image;
and the model training module 940 is used for training the vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image, and the position and the category of each character in each extended image.
For specific definition of the training device for the recognition model of the vehicle identification code, reference may be made to the above definition of the training method for the recognition model of the vehicle identification code, and details are not repeated here. The modules in the training device for the recognition model of the vehicle identification code can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 10. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of training a recognition model of a vehicle identification code. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method steps of the above embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the method steps of the above-mentioned embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (11)
1. A method for training a recognition model of a vehicle identification code, the method comprising:
acquiring a plurality of vehicle identification code area images, and marking the position and the category of each character in each vehicle identification code area image;
performing data expansion operation on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and determining the position and the category of each character in each expanded image;
generating a sample set of the vehicle identification code area images according to the vehicle identification code area images and the expansion images;
and training the vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image.
2. The method of claim 1, wherein the data augmentation operation comprises a layout transformation; the data expansion operation is performed on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and the method comprises the following steps:
respectively intercepting each vehicle identification code area image to obtain a background image which does not contain each character;
adjusting the size of the background image;
according to the position of each character in each vehicle identification code area image, the vehicle identification code in each vehicle identification code area image is cut off at a preset position, and at least two parts of vehicle identification code images are obtained;
at least two parts of vehicle identification code images are superposed to the background image after the size adjustment, and an expanded image after the layout transformation operation is generated.
3. The method of claim 2, wherein the superimposing at least two portions of the vehicle identification code image onto the resized background image to generate the expanded image after the layout transformation operation comprises:
respectively adjusting the sizes of at least two parts of vehicle identification code images;
and superposing the at least two parts of vehicle identification code images after the size adjustment to the background image after the size adjustment to generate an expanded image after the layout transformation operation.
4. The method of claim 1, wherein the data augmentation operation comprises a warping transformation; the data expansion operation is performed on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and the method comprises the following steps:
determining a circumscribed rectangular frame surrounding the vehicle identification code in each vehicle identification code area image according to the positions of the first character and the last character in each vehicle identification code area image;
acquiring coordinates of each vertex of each circumscribed rectangular frame and central coordinates of each vehicle identification code area image;
determining a bending coefficient corresponding to bending transformation according to the actual radian of the arc-shaped vehicle identification code;
and performing bending transformation on each vehicle identification code area image by combining the bending coefficient according to the coordinates of each vertex and the central coordinates to generate an expanded image after the bending transformation operation.
5. The method of claim 1, wherein the data augmentation operation comprises at least one of rotation, translation; the data expansion operation is performed on each marked vehicle identification code area image to generate an expanded image corresponding to each vehicle identification code area image, and the method comprises the following steps:
rotating each marked vehicle identification code area image by a preset angle to generate an expanded image after rotating operation; and/or
And translating the marked vehicle identification code area images according to a preset direction to generate an expanded image after translation operation.
6. The method of claim 1, wherein the vehicle identification code recognition model employs a character instance segmentation model.
7. The method of claim 6, wherein identifying the vehicle identification code in the image of the vehicle identification code region to be identified by using the character instance segmentation model comprises:
acquiring the area image of the vehicle identification code to be identified and a reference character string of the vehicle identification code;
identifying and segmenting the vehicle identification code area image to be identified through the character instance segmentation model;
generating a character string corresponding to the vehicle identification code area image to be identified according to the identification segmentation result of the vehicle identification code area image to be identified;
comparing a character string corresponding to the to-be-identified vehicle identification code area image with a reference character string of the vehicle identification code;
and recording the recognition result of the vehicle identification code according to the comparison result.
8. The method according to claim 1, wherein after performing a data expansion operation on each of the labeled vehicle id code region images to generate an expanded image corresponding to each of the vehicle id code region images, the method further comprises:
detecting whether the extended image contains incomplete characters;
if yes, deleting the extended image containing the incomplete character.
9. A method for identifying a vehicle identification code, the method comprising:
acquiring a vehicle identification code area image to be identified;
inputting the vehicle identification code area image to be identified into the identification model of the vehicle identification code obtained by training according to any one of claims 1 to 8 for identification, and outputting the identification result of the vehicle identification code area image to be identified.
10. An apparatus for training a recognition model of a vehicle identification code, the apparatus comprising:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring a plurality of vehicle identification code area images and marking the positions and the types of characters in the vehicle identification code area images;
the data expansion module is used for performing data expansion operation on each marked vehicle identification code area image, generating an expanded image corresponding to each vehicle identification code area image, and determining the position and the type of each character in each expanded image;
the sample set generating module is used for generating a sample set of the vehicle identification code area image according to each vehicle identification code area image and each expansion image;
and the model training module is used for training the vehicle identification code recognition model according to the sample set of the vehicle identification code area image, the position and the category of each character in each vehicle identification code area image and the position and the category of each character in each extended image.
11. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 9 are implemented when the computer program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910880226.8A CN110619330A (en) | 2019-09-18 | 2019-09-18 | Recognition model training method and device, computer equipment and recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910880226.8A CN110619330A (en) | 2019-09-18 | 2019-09-18 | Recognition model training method and device, computer equipment and recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110619330A true CN110619330A (en) | 2019-12-27 |
Family
ID=68923316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910880226.8A Pending CN110619330A (en) | 2019-09-18 | 2019-09-18 | Recognition model training method and device, computer equipment and recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110619330A (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160342A (en) * | 2019-12-30 | 2020-05-15 | 上海眼控科技股份有限公司 | Vehicle reflective mark verification method and device, computer equipment and storage medium |
| CN111259902A (en) * | 2020-01-13 | 2020-06-09 | 上海眼控科技股份有限公司 | Arc-shaped vehicle identification number detection method and device, computer equipment and medium |
| CN111292396A (en) * | 2020-01-16 | 2020-06-16 | 武汉轻工大学 | Image sample set generation method, device and apparatus and storage medium |
| CN111429516A (en) * | 2020-03-23 | 2020-07-17 | 上海眼控科技股份有限公司 | Corner point positioning method and device of frame number, computer equipment and storage medium |
| CN111488883A (en) * | 2020-04-14 | 2020-08-04 | 上海眼控科技股份有限公司 | Vehicle frame number identification method and device, computer equipment and storage medium |
| CN111598103A (en) * | 2020-06-18 | 2020-08-28 | 上海眼控科技股份有限公司 | Frame number identification method and device, computer equipment and storage medium |
| CN112418214A (en) * | 2020-11-06 | 2021-02-26 | 北京五八信息技术有限公司 | Vehicle identification code identification method and device, electronic equipment and storage medium |
| CN113191426A (en) * | 2021-04-28 | 2021-07-30 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification model creation method, vehicle identification method and related components |
| CN113468906A (en) * | 2021-07-12 | 2021-10-01 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, recognition method, device, equipment and medium |
| CN113989592A (en) * | 2021-10-28 | 2022-01-28 | 三一建筑机器人(西安)研究院有限公司 | Expansion method and device for semantically segmenting image sample and electronic equipment |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150302611A1 (en) * | 2014-04-16 | 2015-10-22 | Xerox Corporation | Vehicle dimension estimation from vehicle images |
| CN105894004A (en) * | 2014-12-16 | 2016-08-24 | 中防通用电信技术有限公司 | Complement technology of locating incomplete license plate aiming at license plate identification system |
| CN107563380A (en) * | 2017-09-08 | 2018-01-09 | 上海理工大学 | A kind of vehicle license plate detection recognition method being combined based on MSER and SWT |
| CN108875722A (en) * | 2017-12-27 | 2018-11-23 | 北京旷视科技有限公司 | Character recognition and identification model training method, device and system and storage medium |
| CN108898137A (en) * | 2018-05-25 | 2018-11-27 | 黄凯 | A kind of natural image character identifying method and system based on deep neural network |
| CN109344835A (en) * | 2018-09-20 | 2019-02-15 | 上海眼控科技股份有限公司 | Altering detecting method based on vehicle VIN code character position |
| CN109447076A (en) * | 2018-09-20 | 2019-03-08 | 上海眼控科技股份有限公司 | A kind of vehicle VIN code recognition detection method for vehicle annual test |
| CN109598246A (en) * | 2018-12-07 | 2019-04-09 | 广东亿迅科技有限公司 | Vehicle enters and leaves detection method, device, computer equipment and storage medium |
| CN109740547A (en) * | 2019-01-04 | 2019-05-10 | 平安科技（深圳）有限公司 | A kind of image processing method, equipment and computer readable storage medium |
| CN109753914A (en) * | 2018-12-28 | 2019-05-14 | 安徽清新互联信息科技有限公司 | A kind of license plate character recognition method based on deep learning |
-
2019
- 2019-09-18 CN CN201910880226.8A patent/CN110619330A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150302611A1 (en) * | 2014-04-16 | 2015-10-22 | Xerox Corporation | Vehicle dimension estimation from vehicle images |
| CN105894004A (en) * | 2014-12-16 | 2016-08-24 | 中防通用电信技术有限公司 | Complement technology of locating incomplete license plate aiming at license plate identification system |
| CN107563380A (en) * | 2017-09-08 | 2018-01-09 | 上海理工大学 | A kind of vehicle license plate detection recognition method being combined based on MSER and SWT |
| CN108875722A (en) * | 2017-12-27 | 2018-11-23 | 北京旷视科技有限公司 | Character recognition and identification model training method, device and system and storage medium |
| CN108898137A (en) * | 2018-05-25 | 2018-11-27 | 黄凯 | A kind of natural image character identifying method and system based on deep neural network |
| CN109344835A (en) * | 2018-09-20 | 2019-02-15 | 上海眼控科技股份有限公司 | Altering detecting method based on vehicle VIN code character position |
| CN109447076A (en) * | 2018-09-20 | 2019-03-08 | 上海眼控科技股份有限公司 | A kind of vehicle VIN code recognition detection method for vehicle annual test |
| CN109598246A (en) * | 2018-12-07 | 2019-04-09 | 广东亿迅科技有限公司 | Vehicle enters and leaves detection method, device, computer equipment and storage medium |
| CN109753914A (en) * | 2018-12-28 | 2019-05-14 | 安徽清新互联信息科技有限公司 | A kind of license plate character recognition method based on deep learning |
| CN109740547A (en) * | 2019-01-04 | 2019-05-10 | 平安科技（深圳）有限公司 | A kind of image processing method, equipment and computer readable storage medium |
Non-Patent Citations (2)
| Title |
|---|
| FU YI 等: "A pattern-based directional element extraction method for vehicle license plate detection in complex background", 《2011 IEEE INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING, COMMUNICATIONS AND COMPUTING (ICSPCC)》 * |
| 邹明明 等: "基于改进模板匹配的车牌字符识别算法实现", 《国外电子测量技术》 * |
Cited By (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11797304B2 (en) | 2018-02-01 | 2023-10-24 | Tesla, Inc. | Instruction set architecture for a vector computational unit |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11983630B2 (en) | 2018-09-03 | 2024-05-14 | Tesla, Inc. | Neural networks for embedded devices |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11908171B2 (en) | 2018-12-04 | 2024-02-20 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| CN111160342A (en) * | 2019-12-30 | 2020-05-15 | 上海眼控科技股份有限公司 | Vehicle reflective mark verification method and device, computer equipment and storage medium |
| CN111259902A (en) * | 2020-01-13 | 2020-06-09 | 上海眼控科技股份有限公司 | Arc-shaped vehicle identification number detection method and device, computer equipment and medium |
| CN111292396B (en) * | 2020-01-16 | 2023-08-29 | 武汉轻工大学 | Image sample set generation method, device, apparatus and storage medium |
| CN111292396A (en) * | 2020-01-16 | 2020-06-16 | 武汉轻工大学 | Image sample set generation method, device and apparatus and storage medium |
| CN111429516A (en) * | 2020-03-23 | 2020-07-17 | 上海眼控科技股份有限公司 | Corner point positioning method and device of frame number, computer equipment and storage medium |
| CN111488883A (en) * | 2020-04-14 | 2020-08-04 | 上海眼控科技股份有限公司 | Vehicle frame number identification method and device, computer equipment and storage medium |
| CN111598103A (en) * | 2020-06-18 | 2020-08-28 | 上海眼控科技股份有限公司 | Frame number identification method and device, computer equipment and storage medium |
| CN112418214A (en) * | 2020-11-06 | 2021-02-26 | 北京五八信息技术有限公司 | Vehicle identification code identification method and device, electronic equipment and storage medium |
| CN113191426A (en) * | 2021-04-28 | 2021-07-30 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification model creation method, vehicle identification method and related components |
| CN113468906A (en) * | 2021-07-12 | 2021-10-01 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, recognition method, device, equipment and medium |
| CN113468906B (en) * | 2021-07-12 | 2024-03-26 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, identification device, equipment and medium |
| CN113989592A (en) * | 2021-10-28 | 2022-01-28 | 三一建筑机器人(西安)研究院有限公司 | Expansion method and device for semantically segmenting image sample and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110619330A (en) | Recognition model training method and device, computer equipment and recognition method | |
| CN110659647B (en) | Seal image identification method and device, intelligent invoice identification equipment and storage medium | |
| US10936911B2 (en) | Logo detection | |
| CN109285105B (en) | Watermark detection method, watermark detection device, computer equipment and storage medium | |
| CN110598687A (en) | Vehicle identification code detection method and device and computer equipment | |
| CN108961315B (en) | Target tracking method and device, computer equipment and storage medium | |
| CN110516541B (en) | Text positioning method and device, computer readable storage medium and computer equipment | |
| CN113469092B (en) | Character recognition model generation method, device, computer equipment and storage medium | |
| CN110210480B (en) | Character recognition method and device, electronic equipment and computer readable storage medium | |
| CN111832561B (en) | Character sequence recognition method, device, equipment and medium based on computer vision | |
| CN110163864B (en) | Image segmentation method and device, computer equipment and storage medium | |
| CN112101386A (en) | Text detection method and device, computer equipment and storage medium | |
| CN111753782A (en) | False face detection method and device based on double-current network and electronic equipment | |
| CN111046755A (en) | Character recognition method, character recognition device, computer equipment and computer-readable storage medium | |
| CN111275743A (en) | Target tracking method, device, computer readable storage medium and computer equipment | |
| CN113592881A (en) | Image reference segmentation method and device, computer equipment and storage medium | |
| CN112580499A (en) | Text recognition method, device, equipment and storage medium | |
| CN112766275A (en) | Seal character recognition method and device, computer equipment and storage medium | |
| CN112418033A (en) | Landslide slope surface segmentation and identification method based on mask rcnn neural network | |
| CN113468906A (en) | Graphic code extraction model construction method, recognition method, device, equipment and medium | |
| CN112434698A (en) | Character recognition method, character recognition device, electronic equipment and storage medium | |
| CN111144294A (en) | Target identification method and device, computer equipment and readable storage medium | |
| JP2014232373A (en) | Collation object extraction system, collation object extraction method, and collation object extraction program | |
| CN112800801B (en) | Method and device for recognizing pattern in image, computer equipment and storage medium | |
| CN115359490A (en) | Text direction identification method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230228 |