CN114283370B - Method and device for identifying uncleaned vehicle and application - Google Patents
Method and device for identifying uncleaned vehicle and application Download PDFInfo
- Publication number
- CN114283370B CN114283370B CN202210148593.0A CN202210148593A CN114283370B CN 114283370 B CN114283370 B CN 114283370B CN 202210148593 A CN202210148593 A CN 202210148593A CN 114283370 B CN114283370 B CN 114283370B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- license plate
- information
- time
- cleaning
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The application provides a method, a device and an application for identifying that a vehicle is not cleaned, wherein the method comprises the following steps: the vehicle target frames are obtained by collecting the vehicle targets for multiple times, the vehicle target frames are sequenced according to the sequence of the collecting time, and if the wheel target frames are not identified, preset data is also filled so as to ensure that each collecting time corresponds to the vehicle identification information. The conclusion of whether the vehicle stays in the cleaning time period can be obtained through the vehicle identification information. And further judging whether the stopped vehicles are the same vehicle or not by combining four conditions of recognizing the license plate. Through the steps, two judgment conditions of vehicles staying in the vehicle cleaning area and the same vehicle as the staying vehicle can be converted into algebraic calculation of the vehicle identification result and the license plate identification result, so that the judgment times are reduced, and the identification speed of the vehicle cleaning result is improved.
Description
Technical Field
The application relates to the technical field of image processing and machine learning, in particular to a method and a device for identifying uncleaned vehicles and application.
Background
The problem that engineering vehicles go on the road with mud in the urban treatment process is always the key point in the treatment, particularly, the arrival of the working season, the engineering vehicles on each construction site are increased increasingly, the engineering vehicles frequently shuttle between the construction site and the urban area, the phenomena of the leakage of the muck along the way and the mud carried by the wheels are visible everywhere, the appearance and the appearance of the city are influenced, and the secondary pollution is caused. The current environmental protection consciousness is a common consensus, road dust brings great negative effects to the production and the life of people, dust pollution generated by road dust on a machineshop truck with mud is closely related to the PM2.5 index of a city, and the dust prevention and treatment of the construction truck are well done to influence the air quality and the appearance of the whole city. The slag car is not cleaned and early-warned, namely, the slag car is effectively managed from the root of the dust emission problem of the slag car, and the influence caused by the road dust emission of the engineering truck is reduced to the maximum extent.
In the prior treatment process, cleaning tasks are carried out on the entering and exiting engineering vehicles mainly by the conscious of construction sites, effective supervision and management are lacked, a plurality of vehicles are not cleaned fully, and the phenomenon that the vehicles still carry mud to go on the road occurs occasionally; with the rapid development of artificial intelligence, the unwashed early warning task of the engineering truck is realized by introducing a computer vision technical means, and the management level of city management in the construction site management process is obviously improved by high-efficiency calculation efficiency and an accurate early warning result.
Therefore, a method for identifying the uncleaned vehicle is needed, which can automatically identify the uncleaned vehicle so as to avoid the engineering vehicle getting on the road with mud.
Disclosure of Invention
The embodiment of the application provides a method and a device for identifying whether a vehicle is not cleaned and application, and aims to solve the problem that the monitoring measures of a machineshop truck on a road with mud cannot be implemented in place due to the fact that the cleaning condition of the vehicle cannot be automatically identified at present and identify whether the vehicle is cleaned in place or not, so that the purpose of early warning that the vehicle is not cleaned is achieved.
In a first aspect, an embodiment of the present application provides a method for identifying that a vehicle is not washed, where the method includes: acquiring a video to be detected corresponding to a vehicle cleaning area, wherein the time length of the video to be detected is set cleaning time length; setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, and inputting all the single-frame images into a trained cleaning action recognition model for recognition to obtain a cleaning action recognition result; setting vehicle sampling duration to acquire a plurality of continuous multiframe images to be detected from the video to be detected, wherein the vehicle sampling duration between two adjacent continuous multiframe images to be detected is less than the cleaning duration; inputting the continuous multiple frames of images to be detected into a trained vehicle detection model and a trained license plate recognition model for detection, and correspondingly obtaining multiple vehicle recognition results and multiple license plate recognition results; and judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
In a second aspect, an embodiment of the present application provides an unwashed vehicle identification apparatus, including: the system comprises a video acquisition module, a video processing module and a video processing module, wherein the video acquisition module is used for acquiring a video to be detected corresponding to a vehicle cleaning area, and the time length of the video to be detected is set cleaning time length; the cleaning action recognition module is used for setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, and inputting all the single-frame images into a trained cleaning action recognition model for recognition to obtain a cleaning action recognition result; the to-be-detected image acquisition module is used for setting vehicle sampling time to acquire a plurality of continuous multiframe to-be-detected images from the to-be-detected video, wherein the vehicle sampling time between two adjacent continuous multiframe to-be-detected images is less than the cleaning time; the vehicle and license plate recognition module is used for inputting the plurality of continuous multi-frame images to be detected into a trained vehicle detection model and a trained license plate recognition model for detection, and obtaining a plurality of vehicle recognition results and a plurality of license plate recognition results correspondingly; and the vehicle uncleaned judging module is used for judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory and a processor, where the memory stores therein a computer program, and the processor is configured to execute the computer program to perform the vehicle uncleaned identification method according to any one of the first aspect.
In a fourth aspect, the present application provides a computer program product comprising software code portions for performing the method for identifying a vehicle uncleaned according to any of the first aspects when the computer program product is run on a computer.
In a fifth aspect, the present application provides a readable storage medium, in which a computer program is stored, the computer program including program code for controlling a process to execute the process, the process including the method for identifying vehicle non-washing according to any one of the first aspect.
The main contributions and innovation points of the embodiment of the application are as follows:
the motion cleaning recognition model and the vehicle detection model input video frames acquired intermittently, so that the recognition speed is obviously higher than the speed of inputting the video frames in the continuous cleaning duration into the model in the conventional means. Based on this, this scheme has still provided a judgement logic that judges to the identification result of the video frame that the discontinuity was gathered: and combining the cleaning action, the vehicle detection result and the license plate detection result to accurately identify the uncleaned vehicle information, and reporting the vehicle information so as to manage the uncleaned vehicle in the follow-up process.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart of main steps of a vehicle uncleaned identification method according to a first embodiment of the present application.
Fig. 2 is a flow chart of unwashed recognition of the work vehicle.
FIG. 3 is a flow chart of an uncleaned determination module.
Fig. 4 is a diagram showing a structure of a washing operation recognition residual.
FIG. 5 is a diagram of a timing characteristics memory module.
FIG. 6 is a timing shift diagram of channel complementation.
Fig. 7 is a block diagram showing the configuration of an input format conversion vehicle uncleaned recognition device of text contents according to a second embodiment of the present application.
Fig. 8 is a schematic hardware configuration diagram of an electronic device according to a third embodiment of the present application.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of the present specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of one or more embodiments of the specification, as detailed in the claims which follow.
It should be noted that: in other embodiments, the steps of the corresponding methods are not necessarily performed in the order shown and described herein. In some other embodiments, the method may include more or fewer steps than those described herein. Moreover, a single step described in this specification may be broken down into multiple steps for description in other embodiments; multiple steps described in this specification may be combined into a single step in other embodiments.
As shown in fig. 2, the scheme can be used for identifying the cleaning action of the vehicle, the vehicle target and the license plate information, and combining the three information to make a judgment result on whether the vehicle is cleaned. This result can be used to realize that the vehicle does not wash the early warning to can follow the root cause of vehicle raise dust problem and start, manage the vehicle raise dust condition on the road.
It should be noted that, a computer vision technology is introduced in the present solution, that is, an image is recognized, where the format, size, and source of the image are not limited, for example: the format may be, but is not limited to, JPG, PNG, TIF, BMP. The source of the image may be an image stored in an SD (secure digital memory card); the images can be obtained by shooting through a camera, capturing through picture capturing software and the like.
In addition, the scheme can automatically judge whether various types of vehicles are cleaned or not and give an early warning, in the embodiment, the engineering vehicle is used as an example, and actually, the types of the vehicles can be any one.
The following describes a scenario in which the method of the present application is applied, by taking "acquiring cleaning information of a mobile machinery shop from a real-time video to identify an unwashed vehicle as an example":
firstly, the camera is arranged at a construction site cleaning bayonet to shoot the passing-in and passing-out condition of a vehicle, a shooting area is presented in a camera video picture, the vehicle cleaning area is marked out in the shooting area, when the vehicle enters the vehicle cleaning area, the vehicle is shown in the vehicle cleaning area, the muck car in the cleaning area is found out by detecting the muck car in the cleaning area, meanwhile, the hand cleaning action of workers in the cleaning area is identified, whether the cleaning is normal or not is judged, whether the cleaning time is sufficient or not is judged by combining with time information, and if the requirement is not met, the cleaning is not considered. And recognizing the information of the vehicle, such as a license plate, a vehicle color, a vehicle logo and the like, as alarm information and sending an early warning, so as to realize an early warning task that the muck vehicle is not cleaned.
Fig. 1 is a flowchart of main steps of a vehicle uncleaned identification method according to a first embodiment of the present application.
To achieve this object, as shown in fig. 1, the vehicle uncleaned identification method mainly includes steps S101 to S105 as follows.
Step 101, a video to be detected corresponding to a vehicle cleaning area is obtained, wherein the duration of the video to be detected is set cleaning duration.
In this step, the cleaning time period means a time period sufficient for the engineering truck to be cleaned. For example: the sufficient cleaning time of the engineering truck is 15 minutes, and if the engineering truck only passes through the vehicle cleaning area, the occurrence time of cleaning the engineering truck in the vehicle cleaning area captured by the camera is less than 15 minutes, and the engineering truck is judged to be insufficiently cleaned.
It should be noted that the video to be detected may be acquired in real time, for example, a video acquired by setting a sampling duration from a real-time video stream captured by monitoring is used as the video to be detected. The purpose of setting the vehicle cleaning area is to ensure that the collected video to be detected contains a vehicle cleaning area picture, in some cases, the camera collecting the video to be detected can only collect the vehicle cleaning area picture, in other cases, the camera collecting the video to be detected can collect a picture of a shooting area, and whether a vehicle is in the vehicle cleaning area is determined by marking the vehicle cleaning area in the shooting area.
Step 102, setting an action sampling duration, intercepting a plurality of single-frame images from the video to be detected, inputting all the single-frame images into a trained cleaning action recognition model for recognition, and obtaining a cleaning action recognition result.
In this step, the washing operation refers to an operation of washing the vehicle recognized by the image. For example: in the scene of manually cleaning the vehicle, the model captures the hand movement; under the scene of automatically cleaning the vehicle, the model captures the movement action of the cleaning equipment.
In the step, the cleaning action within the cleaning time is identified to obtain the cleaning action identification result of whether the vehicle cleaning area is cleaned or not.
It is worth mentioning that it is very difficult to automatically identify the washing duration in the prior art, so manual supervision is mostly adopted, or only the staying time of the vehicle in the vehicle washing area is captured, and whether to wash or not is judged according to the staying time. The traditional method brings the following disadvantages: first, if a vehicle is present in the vehicle washing area at a first time, it is necessary to track the stopping condition of each vehicle for a continuous period of time after the first time. That is, the conventional method needs to detect the video frame by frame in a period of time if it is desired to capture the staying time of the vehicle in the vehicle cleaning area. Second, the dwell time of the vehicle in the vehicle wash zone does not indicate that the vehicle must be washed, such as a work vehicle that is slowly moving through the vehicle wash zone and has been in the zone for a period of time, but not washed. The conventional method may erroneously determine that the vehicle has been washed.
Based on this, the solution proposed in this step is: the cleaning time is set, the identification condition of the vehicle is not considered, only the cleaning action is identified, and the identification result is only: there were washed and not washed. This has the following advantages: in the cleaning action identification stage, only the cleaning action is considered, and whether each vehicle is sufficiently cleaned is judged without combining the vehicle, so that the identification result and the calculation speed are very high. Next, based on the feature that the cleaning operation is a continuous operation, even if a certain image is detected by frame skipping, the cleaning operation can be recognized, so that the frame skipping detection operation can be performed without detecting the entire video frame by frame in this step. Specifically, assuming that the hand of the washing person moves from the foot side to the chest side to represent a washing action, the action in the video is represented by the image continuity of a single frame and a single frame, but actually, only the hand of the washing person is collected to be at the foot side at the last moment, and the hand is at the chest at the next moment, so that the washing action caused by the hand movement can be judged.
Based on this, in the step, the action sampling time length frame skipping detection cleaning action is set, so that the beneficial effect of quickly searching the cleaning condition of the engineering truck by means of image processing is achieved. In other words, the object achieved in this step is to recognize only the washing action during the washing duration and to determine whether or not washing is to be performed in the washing area of the vehicle based on the washing action. Compared with the traditional method for judging whether the vehicle is cleaned or not by adopting manual supervision or only capturing the residence time of the vehicle in the vehicle cleaning area, the method has the advantages that the cleaning action is identified quickly, and the logic of whether the vehicle is cleaned or not is not considered when the cleaning action is identified, so that the efficiency is higher.
103, setting vehicle sampling time to acquire a plurality of continuous multiframe images to be detected from the video to be detected, wherein the vehicle sampling time between two adjacent continuous multiframe images to be detected is less than the cleaning time.
In this step, the sampling duration is less than the cleaning duration, indicating that image data is collected at least twice within the cleaning duration. And the purpose of judging whether the vehicle dwell time achieves sufficient cleaning or not by acquiring image data for multiple times. The explanation is given with two image data acquisitions: when the vehicles are collected for the first time and the vehicles are collected for the second time, and the license plates of the vehicles are judged to be the same for two times, the fact that the vehicles A stay in the vehicle cleaning area for at least the sampling interval duration is shown. Therefore, when the sampling interval duration is set to be the duration that the vehicle can be sufficiently cleaned, the condition that the time that the vehicle stays in the vehicle cleaning area meets the sufficient cleaning duration can be judged through the collected two-time image data.
In addition, it should be noted that, in order to improve the accuracy of the vehicle detection result, each time the step acquires a plurality of continuous frames of images to be detected, that is, whether a vehicle exists in the vehicle cleaning area is determined according to the vehicle detection condition of the plurality of continuous frames of video frames.
And step 104, inputting the continuous multiple frames of images to be detected into the trained vehicle detection model and the trained license plate recognition model for detection, and correspondingly obtaining multiple vehicle recognition results and multiple license plate recognition results.
In this step, it is determined whether there is a vehicle in the vehicle cleaning area based on whether the vehicle target frame is obtained, and if there is no vehicle target frame in the vehicle cleaning area, it indicates that there is no vehicle, and vice versa.
In the step, the license plate recognition result is judged according to whether the obtained license plate data in the license plate target frame are equal or not, namely if the recognized license plate data are the same, the license plate recognition result indicates that the same vehicle stays in the vehicle cleaning area, and if the license plate data are different, the same vehicle is not judged.
And 105, judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
It is worth mentioning that it is generally necessary to combine the characteristics of the color, shape, logo, license plate, etc. of the vehicle while detecting the vehicle to determine whether the vehicle information collected at the first time is the same as the vehicle information collected at the second time. The applicant researches that the license plate numbers in the cleaning time period are different and do not represent that the vehicle is not cleaned, for example, in the actual detection, the following phenomena can occur: the vehicle appears during the first wash interval and the same vehicle appears once during the second wash interval. That is to say, in this case, the license plate information collected each time in the second washing duration is different, but actually, the license plate information of the vehicle should be comprehensively considered by combining the first washing duration and the second washing duration. That is, the manner in which the license plate number is directly recognized to determine whether the same vehicle stays within the vehicle wash area when the vehicle is recognized is liable to cause erroneous determination. Therefore, the scheme has the ingenious points that: when the vehicle target frame is identified, the license plate number is not directly identified and whether the vehicle is the same vehicle or not is judged, and step 104 is used for separately identifying the vehicle identification and the license plate identification to obtain mutually independent results. In the vehicle identification, only whether vehicles exist in a vehicle cleaning area at each acquisition time is judged, only the license plate number is judged in the license plate identification, and the vehicle cleaning judgment error is avoided through the mutually independent detection of the vehicles and the license plates.
Based on this, this scheme will "detect the vehicle and have or not wash the abundant incident" through the mode that detects washing action, vehicle detection frame, license plate detection frame alone and disassemble into the testing process that can carry out the individual detection through three model, and wherein three testing process mutual noninterference has avoided causing the erroneous judgement to the cleaning result. The specific scheme provides two types of embodiments, and the judgment means for judging whether the vehicle is sufficiently cleaned or not by combining the cleaning action, the vehicle identification and the license plate identification results is realized.
In one embodiment, the scheme adopts a judgment method for sequentially judging whether the cleaning action, the vehicle identification result and the license plate identification result meet the cleaning condition to determine whether the vehicle is cleaned. Specifically, in order to save the calculation resources required by judgment to the maximum extent, the cleaning action is judged firstly, if the cleaning action recognition result is not cleaned, the model does not recognize the cleaning action, and the cleaning action is judged to be not cleaned no matter vehicles are collected for several times in the cleaning time period or whether the license plates of the vehicles are the same or not. That is, if the washing operation recognition result indicates that the vehicle is not washed and any of the vehicle recognition results indicates that the vehicle is present, it is determined that the vehicle is not washed.
When the model identifies the cleaning action, whether the conditions of the vehicle and the license plate meet the cleaning condition is judged, specifically, if the cleaning action identification result is that the vehicle is cleaned, and at least two vehicle identification results are that the vehicle exists, and the license plate information of the vehicle existing twice is obtained according to the license plate identification result, the vehicle is judged to be cleaned.
And if the cleaning action recognition result is cleaned, at least two vehicle recognition results are vehicles, and the license plate information is inconsistent according to the license plate recognition result, inquiring the license plate information which is recognized as the vehicle in the license plate recognition results for the first time in a short-time inquiry database, if the license plate information is inquired, considering that the vehicle is cleaned, and if the license plate information is not inquired, judging that the vehicle is not cleaned. And the short-time database stores the license plate information judged to be cleaned by the vehicle within a preset historical time.
Short-term database purpose: the license plate information which is judged to be cleaned by the vehicle in the preset historical time at the previous moment in the database is used for assisting in judging the analyzed vehicle at the current moment, the vehicles which are judged to be cleaned at the previous analysis moment and have too long stay time are filtered, and meanwhile, the condition that the license plate identification is inconsistent in the current analysis time period is further analyzed;
in another embodiment, the scheme numerically represents the recognition condition so as to convert the problem of' whether the same vehicle is cleaned for the cleaning time into an algebraic problem, and compared with the previous embodiment, the calculation amount and the judgment time can be further reduced. In this embodiment, as shown in figure 3,
the vehicle sampling period is set to 1/3 of the wash period, i.e., image data is collected three times during the wash period, defining the collection times as a first time, a second time, and a third time.
Constructing a vehicle matrix A according to the collection sequence of the continuous multi-frame images to be detected and the vehicle identification resultT：
Wherein, 1 represents the presence of a vehicle, and 0 represents the absence of a vehicle; whether the vehicle exists or not is obtained by acquiring a vehicle target frame,
interpretation ATAs shown in the figure, [1, 1]]Is characterized in that vehicles exist at the first moment, the second moment and the third moment, [1, 1, 0 ]]Characterized in that the vehicle exists at the first moment and the second moment and the vehicle does not exist at the third moment, [1, 0, 1]Characterized in that the vehicle exists at the first moment and the third moment, and the vehicle does not exist at the second moment, [0, 1]The vehicle presence at the second time and the third time is indicated, and the vehicle absence at the first time is indicated.
And constructing a license plate matrix B according to the collection sequence of the continuous multi-frame images to be detected and the vehicle identification result:
wherein, the state position of the license plate information corresponding to the state position of the vehicle information in the earliest moment is set as 1 by default, if the corresponding license plates under two moments are the same, the corresponding license plates are set as 1, the license plates are set as-1 by different settings, the license plates of the three are represented by [1, -1, -2] by different settings,
calculating B x ATObtaining a cleaning judgment matrix:
each element in the cleaning judgment matrix represents a state score, and the state score is the product of the vehicle identification result and the license plate identification result at the three acquisition times in the cleaning duration. Specifically, taking the first column of the washing determination matrix as an example, the case where [3, 0, 1, 1, 1] indicates that the vehicle is determined to be unwashed is:
when the vehicle identification result is [1, 1, 1], the license plate matrix is respectively the conditions of [1, 1, 1], [1, -1, -2], [1, -1, 1], [1, 1, -1], [ -1, 1, 1 ].
Namely, when the vehicles are collected at the first moment, the second moment and the third moment, six conditions can occur according to different license plate recognition results, which are respectively as follows:
the license plate information collected for three times is the same;
the license plate information collected for three times is different;
the license plate information at the first moment and the third moment is the same and different from the license plate information at the second moment;
the license plate information at the first moment and the second moment is the same and is different from the license plate information at the third moment;
the license plate information at the second moment and the third moment is the same and different from the license plate information at the first moment.
The above-mentioned (r) - (v) cases correspond to an element value, respectively, and the element value indicates the status score in each case.
And when the state score is greater than 0, judging that the vehicle is cleaned, when the state score is less than or equal to 0, inquiring the license plate information which is firstly identified as the vehicle in the license plate identification result in a short-time inquiry database, if the license plate information is inquired, considering that the vehicle is cleaned, and if the license plate information is not inquired, judging that the vehicle is not cleaned, wherein the license plate information which is judged as the vehicle is cleaned within a preset historical time is stored in the short-time database.
According to the embodiment, the vehicle targets are collected for three times to obtain the plurality of vehicle target frames, the vehicle target frames are sequenced according to the sequence of the collection time, if the vehicle target frames are identified, the vehicle target frames are represented by 1, and if the vehicle target frames are not identified, the vehicle target frames are represented by 0, so that the vehicle matrix is obtained. And the result of whether the vehicle stays in the vehicle cleaning area in the cleaning time period can be obtained through the vehicle matrix. And further judging whether the stopped vehicles are the same vehicle or not by combining the five conditions of the license plate information, and indicating that the vehicles have the stop and the stopped vehicles are the same vehicle on the basis of the state score. That is, according to the embodiment, the two determination conditions of the parked vehicle and the parked vehicle being the same vehicle in the vehicle cleaning area can be converted into algebraic calculations of the vehicle recognition result and the license plate recognition result, so that the number of times of determination is reduced, and the recognition speed of the vehicle cleaning result is increased.
In one possible embodiment, when the vehicle is determined not to be cleaned, reporting vehicle information and sending out an unwashed warning, wherein the vehicle information includes: when the vehicle identification result at each acquisition time is that a vehicle exists, the license plate identification results at each acquisition time are different, and the vehicle information is the license plate information identified at the latest acquisition time before the acquisition time;
when the vehicle identification result is no vehicle at the acquisition moment and the license plate identification results are the same at least two times of acquisition, the vehicle information is the license plate information identified at the acquisition moment corresponding to the same license plate identification result;
and when the vehicle identification result is no vehicle at the collection moment, the license plate identification results are different at each collection moment, and the vehicle information is the earliest license plate information identified at the collection moment.
The embodiment can realize the purposes of identifying the information of the unwashed vehicle under the condition that the vehicle is unwashed, early warning the unwashed event before the vehicle gets on the road and intercepting the vehicle.
Specifically, the video in the cleaning duration is not completely used for recording the cleaning condition of one vehicle, and in most cases, the cleaning information of the previous vehicle and the cleaning information of the next vehicle entering the vehicle cleaning area again after the previous vehicle is cleaned are recorded. Therefore, the situation that different vehicles are identified by several times of license plate identification can occur within a period of cleaning time. Therefore, it is very important to distinguish which vehicle corresponding to the license plate number is not cleaned and report the license plate number information. And the license plate information of the unwashed vehicle to be reported is obtained by combining the vehicle identification result and the license plate identification result.
Reference is again made to fig. 2 to illustrate whether the work vehicle is completely cleaned.
Firstly, a vehicle target frame corresponding to a vehicle cleaning area at each acquisition time is obtained at every sampling interval, wherein the vehicle target frame at each acquisition time is obtained by identifying n continuous video frames.
Specifically, vehicle detection frames of continuous multi-frame monitoring images of a corresponding shooting area at the acquisition time are acquired at intervals of the vehicle sampling time; removing the vehicle detection frame which is not in the vehicle cleaning area to obtain a vehicle determination frame; combining the vehicle determination frames of the continuous multiple frames of the monitoring images to obtain a vehicle determination frame set; and performing overlapping frame removal processing on the vehicle determination frame set to obtain a vehicle target frame.
In the step, firstly, the collected monitoring image is input into a detection model of the engineering vehicle, a vehicle detection frame in the monitoring image is obtained, and then whether the vehicle is in a vehicle cleaning area is determined according to whether the central point of the vehicle detection frame falls in the vehicle cleaning area: if the central point is in the vehicle cleaning area, the vehicle detection frame is reserved, and if the central point is not in the vehicle cleaning area, the vehicle detection frame is removed; and finally combining the vehicle detection frames of the n video frames to obtain a vehicle determination frame set, wherein the vehicle determination frame set represents an accurate set which possibly contains repeated detection frames. And eliminating the vehicle determination frames with the intersection ratio larger than the threshold value at each acquisition time by adopting a non-maximum value inhibition idea, and reserving the remaining vehicle determination frames as vehicle target frames of the vehicle detection model at the acquisition time.
In this step, the main purpose of acquiring n consecutive images during sampling is to reduce the side effects of the convolutional neural network due to pooling. For example: the engineering truck detection algorithm based on the convolutional neural network cannot identify the engineering truck in each frame certainly, even if the images of the engineering trucks are extremely similar, the target can be detected in the former frame, but cannot be identified in the latter frame, and the phenomenon can be relieved by continuously sampling N video frames.
And then inputting the image of the area where the vehicle target frame is located into a license plate recognition model for license plate recognition to obtain a license plate state matrix, wherein the license plate state matrix represents that the license plates collected for several times are the same or different.
In one possible embodiment, the present application provides a cleaning action recognition model for recognizing a cleaning action.
The related cleaning action recognition algorithm comprises the following parts:
(1) constructing a cleaning action recognition data set: dividing each collected cleaning video data into a plurality of video segments, extracting n frames from each segment, processing the video by storing one action image at intervals of t frames, and finally expressing a complete cleaning action by using the n frames of images extracted from one collected segment of video segment, wherein the total number of the divided video segments is D, and D cleaning action training data are finally stored;
(2) designing a cleaning action recognition algorithm: the main network of the cleaning action recognition algorithm adopts a network structure of ResNet50, as shown in FIG. 4, n cleaning action images are input into an action characteristic extraction network to obtain action characteristics, the action characteristics are input into a full connection layer to perform weighted sum on the action characteristics, and then a cleaning action recognition result is obtained through a cleaning action recognizer.
In order to enable the 2D convolution to learn better time sequence information, a module for enhancing time sequence information is designed, specifically referring to fig. 5, the time sequence feature memory module includes two parts, one part of features performs shift calculation to obtain time information of the part of features, the other part of features calculates spatial information of the features and utilizes weighting operation to enhance expression of the spatial information, and finally, the spatial information is obtained by fusion.
Specifically, the method comprises the steps of firstly calculating time sequence information of a convolution layer of 3x3x3 for input features x, then sending extracted feature information into tanh, using the tanh function as output time information of a door controller, and extracting corresponding time information y and space information x-y from original features;
the first part shift calculation includes: the extracted time characteristics are sent to a time sequence displacement module with a complementary channel, the information of adjacent frames is mixed with the information of the current frame, the receptive field of a time channel is enhanced, and meanwhile, the channel information of the displacement time sequence is filled (padding) to remove the characteristic information of the time sequence, a specific filling method is shown in figure 6, the filling method mainly considers that the action correlation of the front time sequence and the rear time sequence of a cleaning action in the identification process is not very high, the operation can be changed to carry out data enhancement, in addition, the time sequence displacement module only carries out displacement in a time domain, no extra calculation burden exists, efficient operation can be carried out as in two-dimensional convolution, and effective modeling is carried out on the time sequence;
the second partial spatial information calculation includes: reducing the spatial dimension of input features in a mean pooling mode of spatial information to obtain x, learning important weight information of each channel in the spatial information by using one-dimensional convolution, re-correcting the spatial feature weight by using a Sigmoid function, and reducing the spatial feature weight to an input scale so as to enhance more spatial features of the characterization information and inhibit secondary weight;
the calculation formula of the whole module is as follows:
wherein, rescale (·) represents a scale transformation function, conv1D represents one-dimensional convolution transformation, Sigmoid (·) function, conv3d represents three-dimensional convolution transformation, y is the selection of time sequence characteristics after input action characteristics pass through tanh, shift represents the time sequence displacement operation of channel complementation, and x ^ is the cleaning action characteristics extracted by the time sequence characteristic memory module;
(3) cleaning action recognition model training:
s1, dividing the acquired cleaning data set into a training set and a testing set according to a certain proportion, and setting input data of an algorithm as X action sequence images;
s2, sending the data into a cleaning recognition network, setting the total iteration number E of training, the batch size B of training and the learning rate lr to be 0.001 for example; during training, the increased learning rate is firstly adjusted to carry out preheating operation of network parameters, and then the normal learning rate is adopted to carry out conventional training, so that the training process is accelerated;
s3, after the cleaning action recognition training data set is iteratively trained for e times, inputting human body verification data into a human body detection model of a training k wheel for prediction to obtain a prediction result, comparing the prediction result with the label marked in the step 1, if the prediction result is consistent with the label marked in the step 1, judging the result to be correct, otherwise, judging the result to be wrong, and calculating the accuracy;
and S4, repeating training iteration, stopping training when the loss function is not reduced and the accuracy rate is not increased any more, and obtaining the trained cleaning action recognition model.
As shown in fig. 4, the method for recognizing a cleaning action using a cleaning action recognition model includes:
inputting all the single-frame images into a backbone network of the cleaning action recognition model for feature extraction to obtain cleaning action features corresponding to each single-frame image; wherein the backbone network is obtained by stacking a plurality of groups of residual modules, each residual module is used for: performing convolution processing on the input features of the single-frame graph to obtain time sequence information, and inputting the input features and the time sequence information into a tanh function to separate time features and space features; inputting the time characteristics into a time sequence displacement module with complementary channels, and extracting time information; pooling the spatial feature mean value into a one-dimensional space, learning important weight information of each channel in the spatial feature by using one-dimensional convolution, re-correcting the weight information, enhancing information expression of the spatial feature by using weighting operation, and extracting spatial information; fusing the time information and the space information to obtain cleaning action characteristics;
and inputting the cleaning action characteristics into a full connection layer of the cleaning action recognition model for classification to obtain a cleaning action recognition result.
The time sequence displacement module mixes the time sequence information of each single frame image with the time sequence information of adjacent frame images, the time sequence displacement module comprises two dimensions of time and an original characteristic channel, the original characteristic channel is divided into a plurality of parts, one part shifts to the upper moment, the other part shifts to the lower moment, the channel information of the shift time sequence of the shift channel characteristics is filled with the channel information of the removal time sequence to obtain a target characteristic channel, and the target characteristic channel is used as the time information.
In summary, the embodiment of the present application provides a method for identifying an unwashed vehicle, which aims to automatically identify an unwashed vehicle, warn an unwashed event before the vehicle gets on the road, and intercept the vehicle.
The vehicle washs regional interior vehicle and detects and obtain vehicle target frame earlier in this scheme, washs regional interior workman's hand washing action to the vehicle and discerns simultaneously, judges whether normally washs, combines "whether same vehicle is gathered many times" again and judges whether the scavenging period is sufficient, if unsatisfied above requirement, then think not the sanitization.
The scheme has the following difference points: the motion cleaning recognition model and the vehicle detection model input video frames acquired intermittently, so that the recognition speed is obviously higher than the speed of inputting the video frames in the continuous cleaning duration into the model in the conventional means. Based on this, this scheme has still provided a judgement logic that judges to the identification result of the video frame that the discontinuity was gathered: and combining the cleaning action, the vehicle detection result and the license plate detection result to accurately identify the uncleaned vehicle information, and reporting the vehicle information so as to manage the uncleaned vehicle in the follow-up process.
In addition, in order to improve the accuracy of cleaning action recognition, the scheme also designs a cleaning action recognition algorithm, and the expression of time sequence characteristics and space characteristics is enhanced in an action characteristic extraction network so as to improve the characteristic extraction effect and facilitate the recognition of the cleaning action.
Fig. 7 is a block diagram of the structure of a vehicle uncleaned recognition device according to a second embodiment of the present application.
As shown in fig. 7, an embodiment of the present application proposes a vehicle uncleaned recognition apparatus including:
the system comprises a video acquisition module, a video acquisition module and a video processing module, wherein the video acquisition module is used for acquiring a video to be detected corresponding to a vehicle cleaning area, and the time length of the video to be detected is set cleaning time length.
And the cleaning action recognition module is used for setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, and inputting all the single-frame images into a trained cleaning action recognition model for recognition to obtain a cleaning action recognition result.
And the to-be-detected image acquisition module is used for setting the vehicle sampling time to acquire a plurality of continuous multiframe to-be-detected images from the to-be-detected video, wherein the vehicle sampling time between two adjacent continuous multiframe to-be-detected images is less than the cleaning time.
And the vehicle and license plate recognition module is used for inputting the plurality of continuous multi-frame images to be detected into the trained vehicle detection model and the trained license plate recognition model for detection, and correspondingly obtaining a plurality of vehicle recognition results and a plurality of license plate recognition results.
And the vehicle uncleaned judging module is used for judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
Fig. 8 is a schematic hardware configuration diagram of an electronic device according to a third embodiment of the present application.
As shown in fig. 8, an electronic device according to an embodiment of the present application includes a memory 804 and a processor 802, where the memory 804 stores a computer program, and the processor 802 is configured to execute the computer program to perform the steps in any of the method embodiments described above.
Specifically, the processor 802 may include a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
The memory 804 may be used to store or cache various data files that need to be processed and/or used for communication, as well as possibly computer program instructions, executed by the processor 802.
The processor 802 implements any of the vehicle uncleaned identification methods in the above embodiments by reading and executing computer program instructions stored in the memory 804.
Optionally, the electronic apparatus may further include a transmission device 806 and an input/output device 808, where the transmission device 806 is connected to the processor 802, and the input/output device 808 is connected to the processor 802.
The transmission device 806 may be used to receive or transmit data via a network. Specific examples of the network described above may include a wired or wireless network provided by a communication provider of the electronic device. In one example, the transmission device includes a Network adapter (NIC) that can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission device 806 can be a Radio Frequency (RF) module, which is used to communicate with the internet in a wireless manner.
The input/output device 808 is used to input or output information. In this embodiment, the input information may be a video, an image frame, or the like, and the output information may be a cleaning action recognition result, a vehicle recognition result, a license plate recognition result, unwashed vehicle information, or the like.
Alternatively, in this embodiment, the processor 802 may be configured to execute the following steps by a computer program:
s101, obtaining a video to be detected corresponding to a vehicle cleaning area, wherein the time length of the video to be detected is set cleaning time length.
S102, setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, inputting all the single-frame images into a trained cleaning action recognition model for recognition, and obtaining a cleaning action recognition result.
S103, setting vehicle sampling time to acquire a plurality of continuous multiframe images to be detected from the video to be detected, wherein the vehicle sampling time between two adjacent continuous multiframe images to be detected is less than the cleaning time.
And S104, inputting the continuous multiple frames of images to be detected into the trained vehicle detection model and the trained license plate recognition model for detection, and correspondingly obtaining multiple vehicle recognition results and multiple license plate recognition results.
And S105, judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
It should be noted that, for specific examples in this embodiment, reference may be made to examples described in the foregoing embodiments and optional implementations, and details of this embodiment are not described herein again.
In general, the various embodiments may be implemented in hardware or special purpose circuits, software, logic or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software which may be executed by a controller, microprocessor or other computing device, although the invention is not limited thereto. While various aspects of the invention may be illustrated and described as block diagrams, flow charts, or using some other pictorial representation, it is well understood that these blocks, apparatus, systems, techniques or methods described herein may be implemented in, as non-limiting examples, hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.
Embodiments of the invention may be implemented by computer software executable by a data processor of the mobile device, such as in a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and/or macros can be stored in any device-readable data storage medium and they include program instructions for performing particular tasks. The computer program product may comprise one or more computer-executable components configured to perform embodiments when the program is run. The one or more computer-executable components may be at least one software code or a portion thereof. Further in this regard it should be noted that any block of the logic flow as in the figures may represent a program step, or an interconnected logic circuit, block and function, or a combination of a program step and a logic circuit, block and function. The software may be stored on physical media such as memory chips or memory blocks implemented within the processor, magnetic media such as hard or floppy disks, and optical media such as, for example, DVDs and data variants thereof, CDs. The physical medium is a non-transitory medium.
It should be understood by those skilled in the art that various features of the above embodiments can be combined arbitrarily, and for the sake of brevity, all possible combinations of the features in the above embodiments are not described, but should be considered as within the scope of the present disclosure as long as there is no contradiction between the combinations of the features.
The above examples are merely illustrative of several embodiments of the present application, and the description is more specific and detailed, but not to be construed as limiting the scope of the present application. 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 application shall be subject to the appended claims.
Claims (11)
1. A method for identifying that a vehicle is not washed is characterized by comprising the following steps:
acquiring a video to be detected corresponding to a vehicle cleaning area, wherein the time length of the video to be detected is set cleaning time length;
setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, and inputting all the single-frame images into a trained cleaning action recognition model for recognition to obtain a cleaning action recognition result;
setting vehicle sampling duration to acquire a plurality of continuous multiframe images to be detected from the video to be detected, wherein the vehicle sampling duration between two adjacent continuous multiframe images to be detected is less than the cleaning duration; wherein the vehicle sampling time length is set as 1/3 of the cleaning time length, and a vehicle matrix A is constructed according to the acquisition sequence of the continuous multi-frame images to be detected and the vehicle identification resultT：
Wherein, 1 represents the presence of a vehicle, and 0 represents the absence of a vehicle;
and constructing a license plate matrix B according to the collection sequence of the continuous multi-frame images to be detected and the vehicle identification result:
wherein, the state position of the license plate information corresponding to the state position of the vehicle information in the earliest moment is set as 1 by default, if the corresponding license plates under two moments are the same, the corresponding license plates are set as 1, the license plates are set as-1 by different settings, the license plates of the three are represented by [1, -1, -2] by different settings,
calculating B x ATObtaining a cleaning judgment matrix:
each element in the cleaning judgment matrix represents a state score, the state score is the product of the vehicle identification result and the license plate identification result at the three-time collection time within the cleaning duration, when the state score is greater than 0, the vehicle is judged to be cleaned, when the state score is less than or equal to 0, the license plate information of the vehicle which is firstly identified as being cleaned in the license plate identification results is inquired in a short-time database, if the license plate information is inquired, the vehicle is considered to be cleaned, and if the license plate information is not inquired, the vehicle is judged not to be cleaned, wherein the license plate information which is judged to be cleaned in the preset historical time is stored in the short-time database;
inputting the continuous multiple frames of images to be detected into a trained vehicle detection model and a trained license plate recognition model for detection, and correspondingly obtaining multiple vehicle recognition results and multiple license plate recognition results;
and judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
2. The method according to claim 1, wherein if the washing operation recognition result is not washed and any one of the vehicle recognition results is a vehicle, it is determined that the vehicle is not washed.
3. The method for identifying whether a vehicle is unwashed according to claim 1, wherein determining whether the vehicle is washed based on the washing action identification result, the plurality of vehicle identification results, and the plurality of license plate identification results comprises:
and if the cleaning action recognition result is that the vehicle is cleaned, at least two vehicle recognition results are that the vehicle exists, and the information of the two license plates is consistent according to the license plate recognition result, judging that the vehicle is cleaned.
4. The method according to claim 1, wherein if the cleaning action recognition result is cleaned, and at least two vehicle recognition results are vehicles, and the license plate information is inconsistent according to the license plate recognition results, the license plate information which is recognized as the vehicle in the license plate recognition results for the first time is inquired in a short-time database, if the license plate information is inquired, the vehicle is considered to be cleaned, and if the license plate information is not inquired, the vehicle is judged to be not cleaned;
and the short-time database stores the license plate information judged to be cleaned by the vehicle within a preset historical time.
5. The method for identifying whether a vehicle is unwashed according to any one of claims 1 to 4, wherein vehicle information is reported and an unwashed warning is issued when the vehicle is determined to be unwashed, wherein the vehicle information includes: when the vehicle identification result at each acquisition time is that a vehicle exists, the license plate identification results at each acquisition time are different, and the vehicle information is the license plate information identified at the latest acquisition time before the acquisition time;
when the vehicle identification result is no vehicle at the acquisition moment and the license plate identification results are the same at least two times of acquisition, the vehicle information is the license plate information identified at the acquisition moment corresponding to the same license plate identification result;
and when the vehicle identification result is no vehicle at the collection moment, the license plate identification results are different at each collection moment, and the vehicle information is the earliest license plate information identified at the collection moment.
6. The method for identifying the uncleaned vehicle according to claim 1, wherein inputting all single-frame images into a trained cleaning action identification model for identification to obtain a cleaning action identification result comprises:
inputting all the single-frame images into a backbone network of the cleaning action recognition model for feature extraction to obtain cleaning action features corresponding to each single-frame image; wherein the backbone network is obtained by stacking a plurality of groups of residual modules, each residual module is used for: performing convolution processing on the input features of the single-frame image to obtain time sequence information, and inputting the input features and the time sequence information into a tanh function to separate time features and space features; inputting the time characteristics into a time sequence displacement module with complementary channels, and extracting time information; pooling the spatial feature mean value into a one-dimensional space, learning important weight information of each channel in the spatial feature by using one-dimensional convolution, re-correcting the weight information, enhancing information expression of the spatial feature by using weighting operation, and extracting spatial information; fusing the time information and the space information to obtain cleaning action characteristics;
and inputting the cleaning action characteristics into a full connection layer of the cleaning action recognition model for classification to obtain a cleaning action recognition result.
7. The method according to claim 6, wherein the time sequence shift module mixes the time sequence information of each single frame image with the time sequence information of the adjacent frame image, the time sequence shift module includes two dimensions of time and an original feature channel, the original feature channel is divided into a plurality of parts, one part is shifted to an upper moment, the other part is shifted to a lower moment, the channel information of the shift time sequence of the shift channel feature is filled with the channel information of the removal time sequence to obtain a target feature channel, and the target feature channel is used as the time information.
8. The method for identifying that a vehicle is not washed according to any one of claims 1 to 4, wherein collecting the vehicle identification result comprises:
the vehicle detection model detects each continuous multiframe image to be detected to correspondingly obtain a plurality of vehicle detection frames;
combining the vehicle detection frames of the continuous multi-frame images to be detected to obtain a vehicle detection frame set;
performing overlapping frame removal processing on the vehicle detection frame set to obtain a vehicle target frame;
and determining a vehicle identification result according to whether the vehicle target frame is detected from each continuous multi-frame image to be detected.
9. An unwashed vehicle identification device characterized by comprising:
the system comprises a video acquisition module, a video processing module and a video processing module, wherein the video acquisition module is used for acquiring a video to be detected corresponding to a vehicle cleaning area, and the time length of the video to be detected is set cleaning time length;
the cleaning action recognition module is used for setting action sampling duration, intercepting a plurality of single-frame images from the video to be detected, and inputting all the single-frame images into a trained cleaning action recognition model for recognition to obtain a cleaning action recognition result;
the to-be-detected image acquisition module is used for setting vehicle sampling time to acquire a plurality of continuous multiframe to-be-detected images from the to-be-detected video, wherein the vehicle sampling time between two adjacent continuous multiframe to-be-detected images is less than the cleaning time; wherein the vehicle sampling time length is set as 1/3 of the cleaning time length, and a vehicle matrix A is constructed according to the acquisition sequence of the continuous multiframe images to be detected and the vehicle identification resultT：
Wherein, 1 represents the presence of a vehicle, and 0 represents the absence of a vehicle;
and constructing a license plate matrix B according to the collection sequence of the continuous multi-frame images to be detected and the vehicle identification result:
wherein, the state position of the license plate information corresponding to the state position of the vehicle information in the earliest moment is set as 1 by default, if the corresponding license plates under two moments are the same, the corresponding license plates are set as 1, the different sets are-1, the license plates of the three are represented by [1, -1, -2] differently,
calculating B x ATObtaining a cleaning judgment matrix:
each element in the cleaning judgment matrix represents a state score, the state score is the product of the vehicle identification result and the license plate identification result at the three-time collection time within the cleaning duration, when the state score is greater than 0, the vehicle is judged to be cleaned, when the state score is less than or equal to 0, the license plate information of the vehicle which is firstly identified as being cleaned in the license plate identification results is inquired in a short-time database, if the license plate information is inquired, the vehicle is considered to be cleaned, and if the license plate information is not inquired, the vehicle is judged not to be cleaned, wherein the license plate information which is judged to be cleaned in the preset historical time is stored in the short-time database;
the vehicle and license plate recognition module is used for inputting the continuous multi-frame images to be detected into a trained vehicle detection model and a trained license plate recognition model for detection, and correspondingly obtaining a plurality of vehicle recognition results and a plurality of license plate recognition results;
and the vehicle uncleaned judging module is used for judging whether the vehicle is cleaned or not based on the cleaning action recognition result, the plurality of vehicle recognition results and the plurality of license plate recognition results.
10. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and the processor is configured to execute the computer program to perform the vehicle uncleaned identification method according to any of claims 1 to 4.
11. A readable storage medium, characterized in that a computer program is stored therein, the computer program comprising program code for controlling a process to execute a process, the process comprising the vehicle uncleaned identification method according to any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148593.0A CN114283370B (en) | 2022-02-18 | 2022-02-18 | Method and device for identifying uncleaned vehicle and application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148593.0A CN114283370B (en) | 2022-02-18 | 2022-02-18 | Method and device for identifying uncleaned vehicle and application |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114283370A CN114283370A (en) | 2022-04-05 |
| CN114283370B true CN114283370B (en) | 2022-06-17 |
Family
ID=80881986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210148593.0A Active CN114283370B (en) | 2022-02-18 | 2022-02-18 | Method and device for identifying uncleaned vehicle and application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114283370B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115273458A (en) * | 2022-06-24 | 2022-11-01 | 富盛科技股份有限公司 | Washing system and method for whether vehicle in construction site uses turbine washing machine or not |
| CN115131737A (en) * | 2022-08-30 | 2022-09-30 | 北京中海住梦科技有限公司 | Vehicle cleaning state identification method and device and electronic equipment |
| CN115880616B (en) * | 2023-03-08 | 2023-05-16 | 城云科技（中国）有限公司 | Method and device for judging specification of cleaning process of large-scale engineering vehicle and application of method and device |
| CN116453071A (en) * | 2023-04-17 | 2023-07-18 | 北京睿芯通量科技发展有限公司 | Identification method and device of vehicle attribute information, electronic equipment and storage medium |
| CN116935308B (en) * | 2023-07-10 | 2024-04-09 | 南京易自助网络科技有限公司 | Car washer safety monitoring system and method based on intelligent identification of car scene AI |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858393A (en) * | 2019-01-11 | 2019-06-07 | 平安科技（深圳）有限公司 | Rule-breaking vehicle recognition methods, system, computer equipment and storage medium |
| CN113221727A (en) * | 2021-05-08 | 2021-08-06 | 杭州鸿泉物联网技术股份有限公司 | Method and system for judging cleaning state of engineering vehicle |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140337319A1 (en) * | 2013-05-13 | 2014-11-13 | Innova Electronics, Inc. | Smart phone application for retrieving and displaying vehicle history report information |
| CN110217205A (en) * | 2019-02-26 | 2019-09-10 | 郑永康 | Vehicle cleaning method and system based on image recognition |
-
2022
- 2022-02-18 CN CN202210148593.0A patent/CN114283370B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858393A (en) * | 2019-01-11 | 2019-06-07 | 平安科技（深圳）有限公司 | Rule-breaking vehicle recognition methods, system, computer equipment and storage medium |
| CN113221727A (en) * | 2021-05-08 | 2021-08-06 | 杭州鸿泉物联网技术股份有限公司 | Method and system for judging cleaning state of engineering vehicle |
Non-Patent Citations (3)
| Title |
|---|
| Pollution Check Control Using License Plate Extraction via Image Processing;Shivani Garg et al.;《Soft Computing: Theories and Applications》;20171125;第133-146页 * |
| 基于图像识别的渣土车监管系统设计及实现;李宇宏 等;《土木工程与管理学报》;20190822;第36卷(第4期);第170-177页 * |
| 时空特征增强机制下基于二维卷积网络的视频行为识别;龚苏明;《中国优秀硕士学位论文全文数据库 信息科技辑》;20220115;第2022年卷(第1期);第I138-2697页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114283370A (en) | 2022-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114283370B (en) | Method and device for identifying uncleaned vehicle and application | |
| CN110807385B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN105744232B (en) | A kind of method of the transmission line of electricity video external force damage prevention of Behavior-based control analytical technology | |
| CN111104903B (en) | Depth perception traffic scene multi-target detection method and system | |
| CN103985182B (en) | A kind of bus passenger flow automatic counting method and automatic counter system | |
| CN111191570B (en) | Image recognition method and device | |
| CN110096975B (en) | Parking space state identification method, equipment and system | |
| CN105519102A (en) | Video monitoring method, video monitoring system and computer program product | |
| CN115620212B (en) | Behavior identification method and system based on monitoring video | |
| CN111429726A (en) | Monitoring video illegal parking vehicle detection and management method and corresponding system | |
| CN114049612B (en) | Highway vehicle charging auditing system based on graph searching technology and driving path dual acquisition and inspection method | |
| CN111914665A (en) | Face shielding detection method, device, equipment and storage medium | |
| CN111626382A (en) | Rapid intelligent identification method and system for cleanliness of vehicle on construction site | |
| CN105243356A (en) | Method of building pedestrian detection model and device and pedestrian detection method | |
| CN110889338A (en) | Unsupervised railway track bed foreign matter detection and sample construction method and unsupervised railway track bed foreign matter detection and sample construction device | |
| CN110889371A (en) | Method and device for detecting throwing of muck truck | |
| CN108109146A (en) | A kind of pavement marker line defect detection device | |
| CN113569756A (en) | Abnormal behavior detection and positioning method, system, terminal equipment and readable storage medium | |
| CN114972911A (en) | Method and equipment for collecting and processing output data of automatic driving perception algorithm model | |
| Javadzadeh et al. | Fast vehicle detection and counting using background subtraction technique and prewitt edge detection | |
| CN111914704B (en) | Tricycle manned identification method and device, electronic equipment and storage medium | |
| CN114067250A (en) | Steal event detection method and device, computer equipment and storage medium | |
| CN113191270A (en) | Method and device for detecting throwing event, electronic equipment and storage medium | |
| CN107730899A (en) | Vehicle identification method and device | |
| CN116805409A (en) | Method for identifying road surface state and evaluating flatness by using driving video |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |