CN111985448A - Vehicle image recognition method and device, computer equipment and readable storage medium - Google Patents
Vehicle image recognition method and device, computer equipment and readable storage medium Download PDFInfo
- Publication number
- CN111985448A CN111985448A CN202010910087.1A CN202010910087A CN111985448A CN 111985448 A CN111985448 A CN 111985448A CN 202010910087 A CN202010910087 A CN 202010910087A CN 111985448 A CN111985448 A CN 111985448A
- Authority
- CN
- China
- Prior art keywords
- data
- image
- network
- preset
- target
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 claims abstract description 59
- 238000012545 processing Methods 0.000 claims description 59
- 238000005286 illumination Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 230000001629 suppression Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 12
- 239000011159 matrix material Substances 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06V20/625—License plates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention discloses a vehicle image recognition method, a vehicle image recognition device, computer equipment and a readable storage medium, which relate to the technical field of image detection and comprise the following steps: initializing a camera, and collecting a current frame picture as an initial image; acquiring target object information in the initial image by adopting a pre-trained detection model based on the initial image; obtaining an image deviation result according to the target object information and preset parameters; the camera is guided and adjusted based on the image deviation result to obtain a target image, and the problems that in the prior art, due to different manual shooting habits, the proportion, the size, the light and the like of obtained partial photos do not meet requirements, the failure rate of the photos is high, an insurance company cannot settle accounts, and the working efficiency is low are solved.
Description
Technical Field
The invention relates to the technical field of image detection, in particular to a vehicle image identification method, a vehicle image identification device, computer equipment and a readable storage medium.
Background
As automobiles become indispensable necessities of people gradually, the problems of traffic accidents and the like are increased day by day, trailer rescue treatment is mostly needed after the traffic accidents happen, in an actual scene, after an automobile owner initiates a rescue request, a third-party rescue company carries out trailer rescue on site, and after the trailer rescue is finished, a corresponding automobile body photo needs to be shot so that a subsequent insurance company can carry out insurance settlement according to the automobile body photo.
However, in the above process, the car body picture is automatically shot by the worker who handles the rescue process, and due to different manual shooting habits, the proportion, the size, the light and the like of the obtained part of the picture do not meet the requirements, the reject ratio of the picture is high, an insurance company cannot settle accounts according to the rejected picture, and needs to shoot again or take other measures to settle accounts, so that the operation is complicated and the efficiency is low.
Disclosure of Invention
The invention aims to provide a vehicle image recognition method, a vehicle image recognition device, computer equipment and a readable storage medium, which are used for solving the problems that in the prior art, due to different manual shooting habits, the proportion, the size, the light ray and the like of an obtained part of photos do not meet requirements, the failure rate of the photos is high, an insurance company cannot settle accounts, and the working efficiency is low.
In order to achieve the above object, the present invention provides a vehicle image recognition method, including:
initializing a camera, and collecting a current frame picture as an initial image;
acquiring target object information in the initial image by adopting a pre-trained detection model based on the initial image;
obtaining an image deviation result according to the target object information and preset parameters;
and guiding and adjusting the camera based on the image deviation result to obtain a target image.
Further, the obtaining target object information in the initial image by using a pre-trained detection model based on the initial image further includes:
performing preset size characteristic recognition on the initial image by adopting a basic network to obtain first data for identifying the position of the vehicle body;
processing the characteristic data output by the conv4_3 convolutional layer in the basic network by adopting a first recognition network to obtain second data for identifying the position of the license plate;
processing the characteristic data output by the conv _7 convolutional layer in the basic network by adopting a second recognition network to obtain third data for identifying the position of the license plate;
outputting an identification result after the first data, the second data and the third data are subjected to non-maximum suppression processing;
and obtaining target object information in the initial image based on the identification result.
Further, processing the feature data output by the conv4_3 convolutional layer in the basic network by using a first recognition network to obtain second data for identifying the position of the license plate, including:
adopting a first volume base layer to reduce the characteristic data output by the conv4_3 volume layer by a first set multiple to obtain first processing data;
determining first body position information with a first classifier based on the first processed data;
searching the license plate position in a first preset range by adopting a first one-hot network according to the first vehicle body position information, and acquiring second processing data;
and marking the second processing data by adopting a first hot spot decision network to obtain second data for identifying the position of the license plate.
Further, marking the second processed data by using a first hot spot decision network to obtain second data for identifying the license plate position, including:
after normalization processing is carried out on the second processing data, marking the data meeting preset conditions;
when the number of marked data in any preset unit in the second processing data exceeds a preset threshold value, marking the unit as target subdata;
and marking each unit in the third processed data one by one to obtain all target subdata and merging the target subdata into second data.
Further, processing the feature data output by the conv _7 convolutional layer in the basic network by using a second recognition network to obtain third data for identifying the position of the license plate, including:
adopting a second volume base layer to reduce the second set multiple processing of the characteristic data output by the conv _7 volume layer to obtain third processing data;
determining second body position information with a second classifier based on the third processed data;
searching the license plate position in a second preset range by adopting a second one-hot network according to the second vehicle body position information, and acquiring fourth processing data;
and marking the fourth processed data by adopting a second hot spot decision network to obtain third data for identifying the position of the license plate.
Further, obtaining an image deviation result according to the target object information and preset parameters, including:
obtaining proportion data and position data corresponding to the target object according to a preset first rule and a preset second rule based on the target object information;
after the initial image with the target object information is subjected to graying processing, calculating illumination data corresponding to the initial image;
comparing the proportion data, the position data and the illumination data with preset parameters to obtain an image deviation result;
the preset parameters comprise a preset proportion parameter, a preset position parameter and a preset illumination parameter.
Further, the guiding adjustment of the camera based on the image deviation result to obtain the target image includes:
generating a corresponding adjustment strategy according to the image deviation result;
judging whether the adjustment strategy is empty, if so, obtaining a target image;
and if not, adjusting the camera according to the adjustment strategy to obtain a target image.
In order to achieve the above object, the present invention also provides a vehicle image recognition apparatus, comprising:
the acquisition module is used for initializing the camera and acquiring a current frame picture as an initial image;
the target object determining module is used for determining target object information in the initial image by adopting a pre-trained detection model based on the initial image;
the image deviation determining module is used for obtaining an image deviation result according to the target object information and preset parameters;
and the execution module is used for guiding and adjusting the camera based on the image deviation result to obtain a target image.
To achieve the above object, the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the vehicle image recognition method when executing the computer program.
In order to achieve the above object, the present invention further provides a computer-readable storage medium comprising a plurality of storage media, each storage medium having stored thereon a computer program, the computer programs stored in the plurality of storage media collectively implementing the steps of the above vehicle image recognition method when executed by a processor.
The technical scheme includes that a trained target detection model is adopted for a current frame portrait to determine a target object in an initial image, whether the target object meets the corresponding requirements of proportion, position or light is judged according to the determined target object information, a corresponding adjusting strategy is generated according to the judgment result, and a camera is adjusted according to the adjusting strategy to obtain a qualified target image.
Drawings
FIG. 1 is a flowchart of a first embodiment of a vehicle image recognition method according to the present invention;
FIG. 2 is a flowchart illustrating a method for recognizing a vehicle image according to an embodiment of the present invention, in which a pre-trained detection model is used to obtain information about a target object in an initial image based on the initial image;
fig. 3 is a flowchart illustrating a first data flow for processing feature data output by a conv4_3 convolutional layer in a base network by using a first recognition network to obtain a second data flow for identifying a license plate position in a first embodiment of a vehicle image recognition method according to the present invention;
fig. 4 is a flowchart of the first embodiment of the vehicle image recognition method according to the present invention, in which the first hot spot decision network is used to mark the second processed data to obtain the second data for identifying the license plate position;
fig. 5 is a flowchart illustrating that a second recognition network is adopted to process feature data output by a conv _7 convolutional layer in a basic network to obtain third data for identifying a license plate position in the first embodiment of the vehicle image recognition method according to the present invention;
fig. 6 is a flowchart illustrating that a second hot spot decision network is used to mark the fourth processed data to obtain third data for identifying the license plate position in the first embodiment of the vehicle image recognition method according to the present invention;
FIG. 7 is a flowchart illustrating an image deviation result obtained according to the target object information and preset parameters according to a first embodiment of the vehicle image recognition method of the present invention;
fig. 8 is a flowchart illustrating guiding adjustment of the camera based on the image deviation result to obtain a target image according to the first embodiment of the vehicle image recognition method of the present invention;
FIG. 9 is a schematic diagram of program modules of a second embodiment of the vehicle image recognition apparatus according to the present invention;
fig. 10 is a schematic diagram of a hardware structure of a computer device according to a third embodiment of the present invention.
Reference numerals:
5. vehicle image recognition device 51, acquisition module 52 and target object determination module
53. Image deviation determination module 531, scale and position data acquisition unit
532. Illumination data acquisition unit 533, result determination unit 54, and execution module
541. Generating unit 542, judging unit 543, and operating unit
6. Computer device 61, memory 62, processor
63. Network interface
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention 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 invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a vehicle image identification method, a vehicle image identification device, computer equipment and a readable storage medium, which are suitable for the field of data analysis of image detection and are based on an acquisition module, a target object determination module, an image deviation determination module and an execution module. The invention collects the current frame picture as the initial image through the collection module, adopts the target determination module to determine the target object in the initial image by adopting the trained target detection model, then judges whether the target object meets the corresponding requirements of proportion, position or light according to the target object information determined by the image deviation determination module, adopts the execution module to generate the corresponding adjustment strategy according to the judgment result, and obtains the qualified target image after adjusting the camera according to the adjustment strategy, thereby solving the problems that in the prior art, the obtained partial picture proportion, size, light and the like do not meet the requirements due to different manual shooting habits, the picture reject ratio is higher, the insurance company can not settle accounts, the working efficiency is lower, and simultaneously, the invention also obtains a new detection model by improving based on the prior SSD target detection model, and deleting part of the search boxes for small target detection, and adding independent detection networks for small targets after the conv4_3 and conv _7 convolutional layers are output, so that the target detection speed is increased, and the accuracy of a target detection result is improved.
Example one
Referring to fig. 1, a vehicle image recognition method of the embodiment is applied to a mobile terminal device with a camera, and includes the following steps:
s100: initializing a camera, and collecting a current frame picture as an initial image;
the mobile terminal device with the camera in the scheme, including but not limited to a smart phone, a tablet, a smart camera, etc., is particularly applied to photographing vehicles and obtaining car body photos for insurance claim settlement, the initialization camera is particularly to set the photographing button on the mobile terminal device to be unavailable, the initialization operation can be manually completed or automatically completed under a preset environment (automatically capturing data after authorization is obtained), such as preset time or preset objects are detected, etc., the photographing process of the car body can be started after the camera is initialized, the current frame picture is the photographed image picture under a normal scene, the analysis and processing are performed based on the image and the camera is guided, the qualification rate of the target image is improved, and the situation that the insurance settlement cannot be performed due to disqualification of the target image is reduced, the method and the device can be applied to smart traffic scenes, and therefore construction of smart cities is promoted.
S200: acquiring target object information in the initial image by adopting a pre-trained detection model based on the initial image;
referring to fig. 2, the step S200 of acquiring the target object information in the initial image includes the following steps:
s210: performing preset size characteristic recognition on the initial image by adopting a basic network to obtain first data for identifying the position of the vehicle body;
in the scheme, the base network is a VGG-16 network in an SSD target detection model, and includes feature maps of different sizes, i.e., conv4_3, conv _7, conv6_2, conv7_2, conv8_2, and conv9_2, and softmax classification and position regression are performed on a plurality of feature maps at the same time, and since a target object in the application scenario is a car or a trailer and the size of the target object is large, the existing VGG-16 network is improved, specifically: the target search box on the conv4_3 network in the existing VGG-16 network is deleted, the number of the search boxes on the conv _7 network is reduced to the original 1/2, the feature identification of a larger size (specific vehicle body outline) is realized through the improvement, and meanwhile, the processing speed is greatly improved because the extraction of small-size features is not needed.
S220: processing the characteristic data output by the conv4_3 convolutional layer in the basic network by adopting a first recognition network to obtain second data for identifying the position of the license plate;
in the scheme, the first recognition network is used for independently processing a target object (such as a license plate) with a smaller size; since the basic network deletes a part of the search box compared to the SSD object detection model in S210, so that the basic network only identifies the object with a larger size, the first identification network and the second identification network in step S230 are added for processing the object with a smaller size, and it should be noted that the first identification network and the second identification network have the same structure and each include a volume base layer, a classifier, a one-hot classifier, and a hot spot decision network layer, which are sequentially arranged.
Specifically, in the step S200, the processing is performed on the feature data output by the conv4_3 convolutional layer in the base network by using the first recognition network to obtain the second data for identifying the license plate position, with reference to fig. 3, where the method includes:
s221: adopting a first volume base layer to reduce the characteristic data output by the conv4_3 volume layer by a first set multiple to obtain first processing data;
it should be noted that, the reduction by the first set factor is a process of reducing the longitudinal thickness by 1/2, and the specific implementation manner is as follows: the first roll of base layer is 1/2 of the conv4_3 convolutional layer, and the first roll of base layer is used for improving the accuracy of the characteristic data output by the conv4_3 convolutional layer.
S222: determining first body position information with a first classifier based on the first processed data;
in the above steps, the first classifier is used to classify the features of the preset size (large size) and further determine the position of the vehicle body, where the position of the vehicle body is represented as a search box containing the vehicle body, and may be implemented by using an SVM classifier, for example.
S223: searching the license plate position in a first preset range by adopting a first one-hot network according to the first vehicle body position information, and acquiring second processing data;
in the above steps, One-Hot is a method of feature extraction, One-Hot coding, also called "One-Hot coding", which uses an N-bit status register to code N states, each state has an independent register bit, and only One of the register bits is valid, i.e. only One state, One-Hot coding is a representation of a categorical variable as a binary vector, which first requires mapping the categorical value to an integer value, and then each integer value is represented as a binary vector, which is a zero value except for the index of the integer, and is labeled as 1. In this embodiment, a first one-hot network is used to search for a small target (i.e., a license plate), and as an example and not by way of limitation, a one-hot coding matrix is output, where each element of the matrix is greater than or equal to 8 and less than 128, and is a license plate area, and other areas are license plate areas, where the first preset range in step S233 is set, as an example, by setting a vehicle body position (x, y), and then the first one-hot network only performs one-hot detection on 0.4x-0.6x in the x direction and 0.3y-0.4y in the y direction of the vehicle body position.
S224: and marking the second processing data by adopting a first hot spot decision network to obtain second data for identifying the position of the license plate.
More specifically, the marking of the second processed data by using the first hot spot decision network to obtain the second data for identifying the license plate position is described with reference to fig. 4, and includes the following steps:
s224-1: after normalization processing is carried out on the second processing data, marking the data meeting preset conditions;
it should be noted that each element of the preset condition matrix is greater than or equal to 8 and less than 128, that is, the license plate area.
S224-2: when the number of marked data in any preset unit in the second processing data exceeds a preset threshold value, marking the unit as target subdata;
specifically, the predetermined unit is (4 × 4).
S224-3: and marking each unit in the third processed data one by one to obtain all target subdata and merging the target subdata into second data.
By way of example and not limitation, the first hotspot decision network is configured to mark data output by the one-hot network after normalization, mark the data with a value greater than or equal to 8 and less than 128 as 0, mark the data with a value greater than 128 as 1, and cumulatively calculate an occupation ratio of 1 in a unit area (4 × 4), and if the occupation ratio of 1 in the unit area exceeds 50%, place a license plate mark in the unit area as Y, and then connect all the unit area areas, where all the connected Y areas are license plate positions.
S230: processing the characteristic data output by the conv _7 convolutional layer in the basic network by adopting a second recognition network to obtain third data for identifying the position of the license plate;
the second recognition network has a structure and a function consistent with those of the first recognition network in step S220, and reference may be made to the implementation process in step S220, where the second data and the third data are both used to identify the license plate region, so that a recognition result with higher accuracy is obtained through the subsequent processing in step S240.
Specifically, the obtaining of the third data for identifying the license plate position in step S230, referring to fig. 5, includes the following steps:
s231: adopting a second volume base layer to reduce the second set multiple processing of the characteristic data output by the conv _7 volume layer to obtain third processing data;
when the second volume base layer in the second identification network is adopted to perform the second setting multiple reduction processing on the characteristic data output by the conv _7 volume layer; the second reduction setting factor is processing of reducing the longitudinal thickness by 1/2, and the specific implementation mode is as follows: 1/2 for the first substrate layer having a thickness conv _7 convolutional layer is set.
S232: determining second body position information with a second classifier based on the third processed data;
consistent with the first classifier described above, the body position is represented as a search box containing the body, which may be implemented, for example, using an SVM classifier.
S233: searching the license plate position in a second preset range by adopting a second one-hot network according to the second vehicle body position information, and acquiring fourth processing data;
the second preset range is similar to the first preset range, for example, if the vehicle body position (x, y) is set, the first one-hot network only performs one-hot detection on an area of the vehicle body position 0.4x-0.6x in the x direction and 0.3y-0.4y in the y direction.
S234: and marking the fourth processed data by adopting a second hot spot decision network to obtain third data for identifying the position of the license plate.
Additionally, the obtaining of the third data for identifying the license plate position in step S234, referring to fig. 6, further includes the following steps:
s234-1: carrying out normalization processing on the fourth processed data and then marking the data meeting preset conditions;
it should be noted that the preset condition is consistent with that in the first hotspot decision network, and is used to mark the data output by the one-hot network after normalization, where the mark is 0 for the data that is greater than or equal to 8 and smaller than 128, and the mark is 1 for the data that is greater than 128.
S234-2: when the number of marked data in any preset unit in the fourth processing data exceeds a preset threshold value, marking the unit as target subdata;
s234-3: and marking each unit in the fourth processing data one by one to obtain all target subdata, and merging the target subdata to obtain second data.
For example, after the step S234-1, the occupation ratio of 1 in the unit area (4 × 4) is calculated in an accumulated manner, if the occupation ratio of the number of 1 in the unit area exceeds 50%, the license plate setting flag in the unit area is Y, and then all the unit area areas are connected, and all the connected Y areas are license plate setting areas, which are the second data.
S240: outputting an identification result after the first data, the second data and the third data are subjected to non-maximum suppression processing;
in this embodiment, a large number of candidate frames may be generated at the same target position, and these candidate frames may overlap each other, and at this time, the non-maximum value is used to suppress and find the best target frame, and eliminate the redundant frame, as an example, a plurality of candidate frames are obtained and the list is built, sorting according to the confidence score of each candidate frame, screening out the candidate frame with the highest confidence to add into the output list, deleting the candidate frame from the list, calculating the areas of other candidate frames except the candidate frame with the highest confidence, calculating IoU (namely dividing the intersection part of the two candidate frames by the union of the two candidate frames) of the candidate frame with the highest confidence and other candidate frames, deleting the candidate frames exceeding the threshold, repeating the process until the list is obtained to be empty, in this embodiment, the second data and the third data may assist the first data in determining the target object.
S250: and obtaining target object information in the initial image based on the identification result.
Specifically, the target object information includes the detected position of the target object, that is, the position of the car body of the car in the present scheme, and if the target object (car) does not exist in the obtained recognition result, the corresponding information is fed back to the user interface, that is, the user is prompted that the car or the trailer does not exist.
In the above embodiment, the target detection model is pre-trained using training data before use; according to the step S210, a part of search boxes in the existing network are deleted, so that only a target object with a larger size is identified, the identification speed is greatly improved, meanwhile, an independent first identification network or second identification network structure is adopted to identify a small target such as a license plate, and the method is different from the mode that the SSD target detection network directly detects the large target and the small target simultaneously in the prior art, the detection speed is improved, meanwhile, the sensitivity of small target detection is improved, the accuracy rate is improved, and the problem that the SSD target detection network is poor in small target adaptability is solved.
It should be noted that the recognition result is represented as an initial image with a target detection frame, and target object information is obtained through position information of the target detection frame.
S300: obtaining an image deviation result according to the target object information and preset parameters;
specifically, the obtaining of the image deviation result according to the target object information and the preset parameter, referring to fig. 7, includes the following steps:
s310: obtaining proportion data and position data corresponding to the target object according to a preset first rule and a preset second rule based on the target object information;
in this scheme, the preset first rule is a1＝(x1/w+y1H)/2, wherein a1Is proportional data; w and h are the length and width of the initial image respectively; x is the number of1、y1The length and the width of the vehicle body corresponding to the proportion data are respectively (can be determined by a target detection frame); the preset second rule is (x)2，y2) (a/w, b/h) wherein (x)2，y2) Is position data; w and h are the length and width of the initial image respectively; a. and b is the central position of the vehicle body corresponding to the position data (determined by the central position of the target detection frame obtained by the identification result).
S320: after the initial image with the target object information is subjected to graying processing, calculating illumination data corresponding to the initial image;
it should be noted that graying allows each pixel in the pixel matrix to satisfy the following relationship: and R ═ G ═ B (the value of the red variable, the value of the green variable, and the value of the blue variable are all equal), and the illumination data is the average value of all pixels in the initial image after the gray scale processing is calculated, and is designated as p.
S330: comparing the proportion data, the position data and the illumination data with preset parameters to obtain an image deviation result;
the preset parameters comprise a preset proportion parameter, a preset position parameter and a preset illumination parameter.
For example, the preset parameter is set to be a ratio data value range {1/2, 2/3}, a value range of x2 in the position data is {2/7, 5/7}, a value range of y2 is {3/7, 5/7}, and a value range of the illumination data is {96, 201}, and the ratio data, the position data and the illumination data are obtained in the steps S31 to S33 and are compared correspondingly, comparison results corresponding to various parameters are obtained, and the comparison results are combined to obtain an image deviation result.
S400: and guiding and adjusting the camera based on the image deviation result to obtain a target image.
Specifically, the above-mentioned guiding adjustment of the camera based on the image deviation result to obtain the target image, referring to fig. 8, includes the following steps:
s410: generating a corresponding adjustment strategy according to the image deviation result;
specifically, the adjustment strategy includes, but is not limited to, the following: if a1<1/2 suggests approach, if a1>2/3 suggests departure; x2<2/7 indicates a right shift, and x2>5/7 indicates a left shift; y2<3/7 indicates upward movement, and y2>5/7 indicates downward movement; if p <96, the prompting light is lightened, and if p >201, the prompting light is lightened, wherein the adjustment strategy is null when the target image information meets each parameter preset range.
S420: judging whether the adjusting strategy is empty;
s430: if yes, obtaining a target image;
s440: and if not, adjusting the camera according to the adjustment strategy to obtain a target image.
In the embodiment, each adjustment strategy can be sent to the user interface for manual adjustment, and the adjustment data can be accurately calculated, so that automatic adjustment is realized by adopting a machine model. In the process of guiding and adjusting the camera, a reference guide label can be preset on the manual adjustment interface, a label to be fitted can be calculated according to target object information and a preset rule, and a photographer is guided by an adjustment strategy to align the label to be fitted with the reference guide label.
The scheme can also comprise uploading the target image to a block chain, and a user can download the target image from the block chain for reference, so that the safety and the fair transparency to the user can be ensured. The user equipment may download the summary information from the blockchain to verify that the priority list is tampered with. The block chain referred by the application is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
According to the technical scheme, a target object in an initial image is determined by adopting a trained target detection model for a current frame portrait, whether the target object meets the corresponding proportion, position or light requirement is judged according to the determined target object information, corresponding adjustment strategies (such as prompting to approach or depart, prompting to move to the right or left, prompting to move up or down, prompting to brighten light and dimming light) are generated according to the judgment result, a qualified target image can be obtained after a camera is adjusted according to the adjustment strategies, and the problem that the qualified target image cannot be used for settlement of insurance companies due to high photo failure rate is solved.
The scheme also provides a new detection model, based on the existing SSD target detection model, a part of search boxes for small target detection are deleted, and independent detection networks (a volume base layer, a classifier, a one-hot classifier and a hot decision network layer, refer to the steps S21-S23) for small targets such as license plates are respectively added after conv4_3 and conv _7 convolutional layers are output, so that the accuracy of target detection results is improved while the target detection speed is improved, and meanwhile, the problem that the SSD target detection model is poor in small target adaptability is solved by monitoring the small targets through the independent detection networks, the sensitivity of the SSD target detection model to the small targets is further improved, and the accuracy of the target detection results is further improved.
Example two:
referring to fig. 9, a vehicle image recognition device 5 of the present embodiment includes:
the acquisition module 51 is used for initializing the camera and acquiring a current frame picture as an initial image;
the mobile terminal device with the camera comprises but not limited to a smart phone, a tablet, a smart camera and the like, is specifically applied to photographing vehicles, and is used for setting a photographing button on the mobile terminal device to be unavailable when a car body photo is obtained for insurance claim settlement.
A target object determination module 52, configured to determine target object information in the initial image based on the initial image by using a pre-trained detection model;
specifically, the detection model comprises a basic network, a first identification network and a second identification network; the basic network is a VGG-16 network in an SSD target detection model, and comprises feature maps with different sizes, namely conv4_3, conv _7, conv6_2, conv7_2, conv8_2 and conv9_2, target search boxes on a conv4_3 network in the existing VGG-16 network are deleted, the number of the search boxes on the conv _7 is reduced to be 1/2, and the first recognition network and the second recognition network have the same structure and respectively comprise a volume base layer, a classifier, a one-hot classifier and a hot point decision network layer which are sequentially arranged.
Preferably, the image deviation determining module 53 further comprises the following:
a ratio and position data obtaining unit 531, configured to obtain ratio data and position data corresponding to the target object according to a preset first rule and a preset second rule based on the target object information;
the preset first rule is a1 ═ x1/w + y1/h)/2, where a1 is the ratio data; w and h are the length and width of the initial image respectively; x1 and y1 are respectively the length and the width of the vehicle body corresponding to the proportional data (which can be determined by a target detection frame); the preset second rule is (x2, y2) ═ a/w, b/h, where (x2, y2) is position data; w and h are the length and width of the initial image respectively; a. and b is the central position of the vehicle body corresponding to the position data respectively.
An illumination data obtaining unit 532, configured to calculate illumination data corresponding to the initial image after graying the initial image with the target object information;
a result determining unit 533, configured to compare the ratio data, the position data, and the illumination data with preset parameters to obtain an image deviation result;
the preset parameters comprise a preset proportion parameter, a preset position parameter and a preset illumination parameter.
And the execution module 54 is configured to perform guiding adjustment on the camera based on the image deviation result to obtain a target image.
Preferably, the execution module 54 includes the following:
a generating unit 541, configured to generate a corresponding adjustment policy according to the image deviation result
The adjustment strategies include, but are not limited to, the following: if a1<1/2 suggests approach, if a1>2/3 suggests departure; x2<2/7 indicates a right shift, and x2>5/7 indicates a left shift; y2<3/7 indicates upward movement, and y2>5/7 indicates downward movement; if p is less than 96, the prompting light is lightened, if p is greater than 201, the prompting light is lightened, and it needs to be noted that when the target image information meets each parameter preset range, the adjustment strategy is null;
a judging unit 542, configured to judge whether the adjustment policy is empty;
an operation unit 543 for obtaining a target image when the determination result is yes; and if not, adjusting the camera according to the adjustment strategy to obtain a target image.
The technical scheme is based on image classification in image detection, a current frame of image is collected by a collection module to serve as an initial image, a target determination module is adopted to determine a target object in the initial image by adopting a trained target detection model, then whether the target object information meets preset parameters or not is judged according to target object information determined by a proportion and position data acquisition unit and a lighting data acquisition unit in an image deviation determination module and a result judgment unit, a generation unit in an execution module is adopted to generate a corresponding adjustment strategy according to the judgment result after an image deviation result is obtained, a qualified target image can be obtained after a camera is adjusted according to the adjustment strategy based on the judgment unit and an operation unit, and the proportion, the proportion and the ratio of partial pictures, which are obtained and are caused by different manual shooting habits in the prior art are reduced, The size, the light and the like do not meet the requirements, the failure rate of the photos is high, the insurance company cannot settle accounts, and the working efficiency is low.
In the scheme, in the detection model processing process in a target determination module, a basic network is adopted to perform preset size feature recognition on an initial image to obtain first data for identifying the position of a vehicle body, the first recognition network and the second recognition network are respectively adopted to process feature data output by conv4_3 and conv _7 rolling layers in the basic network to obtain second data and third data for identifying the position of a license plate, finally target object information is obtained based on the first data, the second data and the third data, part of a search frame for small target detection is deleted based on the existing SSD target detection model, meanwhile, the first recognition network and the second recognition network which are independent and used for small targets are respectively added after the conv4_3 and conv _7 rolling layers are output to perform secondary detection on the small targets such as the license plate and the like, the accuracy of a target detection result is improved while the target detection speed is improved, meanwhile, the small target is monitored through the independent detection network, the problem that the adaptability of the SSD target detection model to the small target is poor is solved, the sensitivity to the small target is further improved, and the accuracy of a target detection result is further improved.
Example three:
in order to achieve the above object, the present invention further provides a computer device 6, where the computer device 6 includes a plurality of computer devices 6, and the components of the vehicle image recognition apparatus 5 according to the second embodiment may be distributed in different computer devices, and the computer devices may be a smart phone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server, or a rack server (including an independent server or a server cluster formed by a plurality of servers) that executes programs, and the like. The computer device of the embodiment at least includes but is not limited to: a memory 61, a processor 62, a network interface 63, and the vehicle image recognition device 5, which are communicably connected to each other through a system bus, as shown in fig. 10. It should be noted that fig. 10 only shows a computer device with components, but it should be understood that not all of the shown components are required to be implemented, and more or fewer components may be implemented instead.
In the present embodiment, the memory 61 (i.e., a readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the memory 61 may be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. In other embodiments, the memory 61 may also be an external storage device of the computer device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the computer device. Of course, the memory 61 may also include both internal and external storage devices of the computer device. In this embodiment, the memory 61 is generally used for storing an operating system and various types of application software installed in a computer device, such as a program code of the vehicle image recognition apparatus according to the first embodiment. Further, the memory 61 may also be used to temporarily store various types of data that have been output or are to be output.
Processor 62 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 62 is typically used to control the overall operation of the computer device. In this embodiment, the processor 62 is configured to operate the program codes stored in the memory 61 or process data, for example, operate the vehicle image recognition device, so as to implement the vehicle image recognition method according to the first embodiment.
The network interface 63 may comprise a wireless network interface or a wired network interface, and the network interface 63 is typically used to establish a communication connection between the computer device 6 and other computer devices 6. For example, the network interface 63 is used to connect the computer device 6 to an external terminal via a network, establish a data transmission channel and a communication connection between the computer device 6 and the external terminal, and the like. The network may be a wireless or wired network such as an Intranet (Intranet), the Internet (Internet), a Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), a 4G network, a 5G network, Bluetooth (Bluetooth), Wi-Fi, and the like.
It is noted that fig. 10 only shows the computer device 6 with components 61-63, but it is to be understood that not all shown components are required to be implemented, and that more or less components may be implemented instead.
In the present embodiment, the vehicle image recognition apparatus 5 stored in the memory 61 may also be divided into one or more program modules, which are stored in the memory 61 and executed by one or more processors (in the present embodiment, the processor 62) to complete the present invention.
Example four:
to achieve the above objects, the present invention also provides a computer-readable storage system including a plurality of storage media, such as a flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application store, etc., on which a computer program is stored, which when executed by a processor 62, implements corresponding functions. The computer readable storage medium of the present embodiment is used for storing a vehicle image recognition device, and when being executed by the processor 62, the vehicle image recognition method of the first embodiment is implemented.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A vehicle image recognition method, characterized by comprising:
initializing a camera, and collecting a current frame picture as an initial image;
acquiring target object information in the initial image by adopting a pre-trained detection model based on the initial image;
obtaining an image deviation result according to the target object information and preset parameters;
and guiding and adjusting the camera based on the image deviation result to obtain a target image.
2. The vehicle image recognition method according to claim 1, wherein the obtaining target object information in the initial image based on the initial image by using a pre-trained detection model further comprises:
performing preset size characteristic recognition on the initial image by adopting a basic network to obtain first data for identifying the position of the vehicle body;
processing the characteristic data output by the conv4_3 convolutional layer in the basic network by adopting a first recognition network to obtain second data for identifying the position of the license plate;
processing the characteristic data output by the conv _7 convolutional layer in the basic network by adopting a second recognition network to obtain third data for identifying the position of the license plate;
outputting an identification result after the first data, the second data and the third data are subjected to non-maximum suppression processing;
and obtaining target object information in the initial image based on the identification result.
3. The vehicle image recognition method according to claim 2, wherein the processing the feature data output by the conv4_3 convolutional layer in the base network by using the first recognition network to obtain the second data for identifying the position of the license plate comprises:
adopting a first volume base layer to reduce the characteristic data output by the conv4_3 volume layer by a first set multiple to obtain first processing data;
determining first body position information with a first classifier based on the first processed data;
searching the license plate position in a first preset range by adopting a first one-hot network according to the first vehicle body position information, and acquiring second processing data;
and marking the second processing data by adopting a first hot spot decision network to obtain second data for identifying the position of the license plate.
4. The vehicle image recognition method of claim 3, wherein the marking the second processed data with the first hot spot decision network to obtain second data for identifying a license plate location comprises:
after normalization processing is carried out on the second processing data, marking the data meeting preset conditions;
when the number of marked data in any preset unit in the second processing data exceeds a preset threshold value, marking the unit as target subdata;
and marking each unit in the third processed data one by one to obtain all target subdata and merging the target subdata into second data.
5. The vehicle image recognition method according to claim 2, wherein the processing the feature data output by the conv _7 convolutional layer in the base network by using the second recognition network to obtain third data for identifying the position of the license plate comprises:
adopting a second volume base layer to reduce the second set multiple processing of the characteristic data output by the conv _7 volume layer to obtain third processing data;
determining second body position information with a second classifier based on the third processed data;
searching the license plate position in a second preset range by adopting a second one-hot network according to the second vehicle body position information, and acquiring fourth processing data;
and marking the fourth processed data by adopting a second hot spot decision network to obtain third data for identifying the position of the license plate.
6. The vehicle image recognition method according to claim 1, wherein the obtaining of the image deviation result according to the target object information and preset parameters includes:
obtaining proportion data and position data corresponding to the target object according to a preset first rule and a preset second rule based on the target object information;
after the initial image with the target object information is subjected to graying processing, calculating illumination data corresponding to the initial image;
comparing the proportion data, the position data and the illumination data with preset parameters to obtain an image deviation result;
the preset parameters comprise a preset proportion parameter, a preset position parameter and a preset illumination parameter.
7. The vehicle image recognition method according to claim 1, wherein the performing guidance adjustment on a camera based on the image deviation result to obtain a target image comprises:
generating a corresponding adjustment strategy according to the image deviation result;
judging whether the adjustment strategy is empty, if so, obtaining a target image;
and if not, adjusting the camera according to the adjustment strategy to obtain a target image.
8. A vehicle image recognition apparatus, characterized by comprising:
the acquisition module is used for initializing the camera and acquiring a current frame picture as an initial image;
the target object determining module is used for determining target object information in the initial image by adopting a pre-trained detection model based on the initial image;
the image deviation determining module is used for obtaining an image deviation result according to the target object information and preset parameters;
and the execution module is used for guiding and adjusting the camera based on the image deviation result to obtain a target image.
9. A computer arrangement, characterized in that the computer arrangement comprises a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the vehicle image recognition method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium comprising a plurality of storage media, each storage medium having a computer program stored thereon, wherein the computer programs stored in the plurality of storage media, when executed by a processor, collectively implement the steps of the vehicle image recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010910087.1A CN111985448A (en) | 2020-09-02 | 2020-09-02 | Vehicle image recognition method and device, computer equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010910087.1A CN111985448A (en) | 2020-09-02 | 2020-09-02 | Vehicle image recognition method and device, computer equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111985448A true CN111985448A (en) | 2020-11-24 |
Family
ID=73447893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010910087.1A Pending CN111985448A (en) | 2020-09-02 | 2020-09-02 | Vehicle image recognition method and device, computer equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111985448A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114895832A (en) * | 2022-05-17 | 2022-08-12 | 网易（杭州）网络有限公司 | Object adjustment method and device, electronic equipment and computer readable medium |
| CN115601977A (en) * | 2022-10-27 | 2023-01-13 | 天津华慧智能科技有限公司（Cn） | Mobile phone automatic-guide shooting judgment accurate parking system and use method |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005295406A (en) * | 2004-04-02 | 2005-10-20 | Casio Comput Co Ltd | Imaging device, and program |
| CN106139564A (en) * | 2016-08-01 | 2016-11-23 | 纳恩博（北京）科技有限公司 | Image processing method and device |
| CN106740841A (en) * | 2017-02-14 | 2017-05-31 | 驭势科技（北京）有限公司 | Method for detecting lane lines, device and mobile unit based on dynamic control |
| CN107316001A (en) * | 2017-05-31 | 2017-11-03 | 天津大学 | Small and intensive method for traffic sign detection in a kind of automatic Pilot scene |
| CN107729808A (en) * | 2017-09-08 | 2018-02-23 | 国网山东省电力公司电力科学研究院 | A kind of image intelligent acquisition system and method for power transmission line unmanned machine inspection |
| CN109325488A (en) * | 2018-08-31 | 2019-02-12 | 阿里巴巴集团控股有限公司 | For assisting the method, device and equipment of car damage identification image taking |
| CN109376518A (en) * | 2018-10-18 | 2019-02-22 | 深圳壹账通智能科技有限公司 | Privacy leakage method and relevant device are prevented based on recognition of face |
| CN110285793A (en) * | 2019-07-08 | 2019-09-27 | 中原工学院 | A kind of Vehicular intelligent survey track approach based on Binocular Stereo Vision System |
| CN110633629A (en) * | 2019-08-02 | 2019-12-31 | 广东电网有限责任公司清远供电局 | Power grid inspection method, device, equipment and storage medium based on artificial intelligence |
| CN110738192A (en) * | 2019-10-29 | 2020-01-31 | 腾讯科技（深圳）有限公司 | Human motion function auxiliary evaluation method, device, equipment, system and medium |
| CN111046917A (en) * | 2019-11-20 | 2020-04-21 | 南京理工大学 | Object-based enhanced target detection method based on deep neural network |
| CN111079506A (en) * | 2019-10-11 | 2020-04-28 | 深圳壹账通智能科技有限公司 | Augmented reality-based information acquisition method and device and computer equipment |
| WO2020140772A1 (en) * | 2019-01-02 | 2020-07-09 | 腾讯科技（深圳）有限公司 | Face detection method, apparatus, device, and storage medium |
-
2020
- 2020-09-02 CN CN202010910087.1A patent/CN111985448A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005295406A (en) * | 2004-04-02 | 2005-10-20 | Casio Comput Co Ltd | Imaging device, and program |
| CN106139564A (en) * | 2016-08-01 | 2016-11-23 | 纳恩博（北京）科技有限公司 | Image processing method and device |
| CN106740841A (en) * | 2017-02-14 | 2017-05-31 | 驭势科技（北京）有限公司 | Method for detecting lane lines, device and mobile unit based on dynamic control |
| CN107316001A (en) * | 2017-05-31 | 2017-11-03 | 天津大学 | Small and intensive method for traffic sign detection in a kind of automatic Pilot scene |
| CN107729808A (en) * | 2017-09-08 | 2018-02-23 | 国网山东省电力公司电力科学研究院 | A kind of image intelligent acquisition system and method for power transmission line unmanned machine inspection |
| CN109325488A (en) * | 2018-08-31 | 2019-02-12 | 阿里巴巴集团控股有限公司 | For assisting the method, device and equipment of car damage identification image taking |
| CN109376518A (en) * | 2018-10-18 | 2019-02-22 | 深圳壹账通智能科技有限公司 | Privacy leakage method and relevant device are prevented based on recognition of face |
| WO2020140772A1 (en) * | 2019-01-02 | 2020-07-09 | 腾讯科技（深圳）有限公司 | Face detection method, apparatus, device, and storage medium |
| CN110285793A (en) * | 2019-07-08 | 2019-09-27 | 中原工学院 | A kind of Vehicular intelligent survey track approach based on Binocular Stereo Vision System |
| CN110633629A (en) * | 2019-08-02 | 2019-12-31 | 广东电网有限责任公司清远供电局 | Power grid inspection method, device, equipment and storage medium based on artificial intelligence |
| CN111079506A (en) * | 2019-10-11 | 2020-04-28 | 深圳壹账通智能科技有限公司 | Augmented reality-based information acquisition method and device and computer equipment |
| CN110738192A (en) * | 2019-10-29 | 2020-01-31 | 腾讯科技（深圳）有限公司 | Human motion function auxiliary evaluation method, device, equipment, system and medium |
| CN111046917A (en) * | 2019-11-20 | 2020-04-21 | 南京理工大学 | Object-based enhanced target detection method based on deep neural network |
Non-Patent Citations (7)
| Title |
|---|
| JOSHUA J. NORDBERG 等: "Practical Aspects of Adjusting Digital Cameras", METHODS IN CELL BIOLOGY, pages 151 - 162 * |
| MIN-KOOK CHOI 等: "Co-Occurrence Matrix Analysis-Based Semi-Supervised Training for Object Detection", 2018 25TH IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), pages 1333 - 1337 * |
| TIANCAI WANG 等: "Learning Rich Features at High-Speed for Single-Shot Object Detection", 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), pages 1971 - 1980 * |
| WEIQIANG LI 等: "A Single-Shot Object Detector with Feature Aggregation and Enhancement", 2019 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), pages 3910 - 3914 * |
| 刘继丹 等: "改进SSD卷积神经网络在钢板边裂纹检测中的应用", 冶金自动化, vol. 44, no. 04, pages 43 - 47 * |
| 叶召元 等: "基于自动驾驶场景的目标检测算法DFSSD", 计算机工程与应用, vol. 56, no. 16, pages 139 - 147 * |
| 徐科 等: "冷轧带钢表面质量自动检测系统的在线应用研究", 冶金自动化, pages 51 - 53 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114895832A (en) * | 2022-05-17 | 2022-08-12 | 网易（杭州）网络有限公司 | Object adjustment method and device, electronic equipment and computer readable medium |
| CN114895832B (en) * | 2022-05-17 | 2023-08-08 | 网易（杭州）网络有限公司 | Object adjustment method, device, electronic equipment and computer readable medium |
| CN115601977A (en) * | 2022-10-27 | 2023-01-13 | 天津华慧智能科技有限公司（Cn） | Mobile phone automatic-guide shooting judgment accurate parking system and use method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107194398B (en) | Vehicle damages recognition methods and the system at position | |
| WO2019205376A1 (en) | Vehicle damage determination method, server, and storage medium | |
| CN106599792B (en) | Method for detecting hand driving violation behavior | |
| CN108009543A (en) | A kind of licence plate recognition method and device | |
| WO2021143063A1 (en) | Vehicle damage assessment method, apparatus, computer device, and storage medium | |
| CN112862702B (en) | Image enhancement method, device, equipment and storage medium | |
| CN112287912B (en) | Deep learning-based lane line detection method and device | |
| CN111985448A (en) | Vehicle image recognition method and device, computer equipment and readable storage medium | |
| CN113822247B (en) | Method and system for identifying illegal building based on aerial image | |
| CN110633289A (en) | License plate number determining method, server and parking management system | |
| WO2021027157A1 (en) | Vehicle insurance claim settlement identification method and apparatus based on picture identification, and computer device and storage medium | |
| CN112132216B (en) | Vehicle type recognition method and device, electronic equipment and storage medium | |
| CN114038004A (en) | Certificate information extraction method, device, equipment and storage medium | |
| CN112581522A (en) | Method and device for detecting position of target object in image, electronic equipment and storage medium | |
| CN111860219B (en) | High-speed channel occupation judging method and device and electronic equipment | |
| CN111507957B (en) | Identity card picture conversion method and device, computer equipment and storage medium | |
| CN112036488A (en) | Event identification method, device and equipment based on image identification and storage medium | |
| CN112861567B (en) | Vehicle type classification method and device | |
| CN115810134A (en) | Image acquisition quality inspection method, system and device for preventing car insurance from cheating | |
| CN111695615A (en) | Vehicle damage assessment method and device based on artificial intelligence, electronic equipment and medium | |
| CN112597995B (en) | License plate detection model training method, device, equipment and medium | |
| CN112733652B (en) | Image target recognition method, device, computer equipment and readable storage medium | |
| CN112966687B (en) | Image segmentation model training method and device and communication equipment | |
| CN113160272B (en) | Target tracking method and device, electronic equipment and storage medium | |
| CN114898306B (en) | Method and device for detecting target orientation and electronic equipment |
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 |