CN114445787A - Non-motor vehicle weight recognition method and related equipment - Google Patents
Non-motor vehicle weight recognition method and related equipment Download PDFInfo
- Publication number
- CN114445787A CN114445787A CN202111663748.6A CN202111663748A CN114445787A CN 114445787 A CN114445787 A CN 114445787A CN 202111663748 A CN202111663748 A CN 202111663748A CN 114445787 A CN114445787 A CN 114445787A
- Authority
- CN
- China
- Prior art keywords
- motor vehicle
- information
- human body
- frame
- rider
- 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
Images
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a non-motor vehicle weight identification method, which comprises the following steps: acquiring an image sequence to be processed, wherein the image sequence to be processed comprises a non-motor vehicle; aiming at the image sequence to be processed, carrying out non-motor vehicle detection and human body detection on each frame of image to be processed to obtain first non-motor vehicle information and human body information; if the fact that the human body information and the first non-motor vehicle have a riding relationship is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with attributes of riders; and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle. The non-motor vehicle is identified again through the non-motor vehicle information and the rider's body information, and the rider's body information can be used as the auxiliary information for identifying the weight of the non-motor vehicle, so that the accuracy of identifying the weight of the non-motor vehicle is improved.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a non-motor vehicle weight identification method and related equipment.
Background
With the development of urbanization, the number of motor vehicles and non-motor vehicles in urban traffic is increasing. The non-motor vehicles are used as important components of urban traffic, and the retrieval and tracking of the non-motor vehicles are of great significance for building safe cities and smart cities. Non-motor vehicle re-identification is a technology for automatically identifying a specific non-motor vehicle through multiple surveillance videos in a complex environment. At present, the data collection difficulty of the non-motor vehicle weight recognition aiming at the vision task is high, and the difference of illumination change, visual angle change and resolution ratio can be generated due to the difference of the positions of the cameras, so that the same vehicle can generate self difference under different visual angles, and the accuracy of the non-motor vehicle weight recognition is not high.
Disclosure of Invention
The embodiment of the invention provides a non-motor vehicle weight recognition method and related equipment, wherein the non-motor vehicle is re-recognized through non-motor vehicle information and rider human body information, and the rider human body information can be used as auxiliary information for non-motor vehicle weight recognition, so that the accuracy of non-motor vehicle weight recognition is improved.
In a first aspect, an embodiment of the present invention provides a non-motor vehicle weight identification method, where the method includes:
acquiring an image sequence to be processed, wherein the image sequence to be processed comprises a non-motor vehicle;
carrying out non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information;
when the fact that the human body information and the first non-motor vehicle information have a riding relationship is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with attributes of riders;
and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
Optionally, the first non-motor vehicle information includes a non-motor vehicle detection frame, the human body information includes a human body detection frame, and before the step of associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information having a rider attribute when it is detected that the human body information and the first non-motor vehicle information have a riding relationship, the method includes:
calculating the intersection ratio of the non-motor vehicle detection frame and the human body detection frame, and selecting the non-motor vehicle detection frame with the intersection ratio larger than a preset value to be combined with the human body detection frame to obtain a candidate combination frame, wherein the candidate combination frame comprises the non-motor vehicle detection frame and the human body detection frame;
and performing behavior detection on the images in the candidate combination frame to obtain a behavior detection result of the images in the candidate combination frame, wherein the behavior detection result comprises whether the human body information and the first non-motor vehicle information have a riding relationship.
Optionally, the selecting the non-motor vehicle detection frame with the intersection ratio larger than the preset value to be combined with the human body detection frame to obtain a candidate combination frame includes:
calculating a minimum surrounding frame between the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than a preset value, wherein the minimum surrounding frame covers the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than the preset value at the same time and has the minimum area;
determining the minimum bounding box as the candidate combo box.
Optionally, the performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle includes:
performing target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle;
and re-identifying according to the first track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, the performing target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first trajectory information of the non-motor vehicle includes:
extracting attribute of the rider in the second non-motor vehicle information, extracting human body information corresponding to the rider, and carrying out target tracking on the rider through the human body information;
if the target tracking is successful, obtaining the trajectory information of the rider based on a target tracking result;
according to the track information of the rider, under the condition that the rider has a riding relationship with the non-motor vehicle, the track information of the non-motor vehicle is updated through the track information of the rider, and first track information of the non-motor vehicle is obtained.
Optionally, after the step of performing target tracking on the rider through the human body information, the method further includes:
if the target tracking fails, performing target tracking on the non-motor vehicle through the first non-motor vehicle information to obtain second track information of the non-motor vehicle;
and re-identifying according to the second track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, after obtaining the first trajectory information of the non-motor vehicle, the method includes:
acquiring a first target image sequence of first track information in the image sequence to be processed;
extracting a first identification feature corresponding to the first track information according to the first target image sequence;
calculating feature similarity of all the first track information based on the first identification features, and combining the first track information with the feature similarity larger than a similarity threshold value to obtain combined track information;
and merging the first track information through the combined track information.
In a second aspect, an embodiment of the present invention provides a non-motor vehicle weight recognition apparatus, including:
the device comprises a first acquisition module, a second acquisition module and a processing module, wherein the first acquisition module is used for acquiring an image sequence to be processed, and the image sequence to be processed comprises a non-motor vehicle;
the first detection module is used for carrying out non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information;
the association module is used for associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with the attribute of a rider when the fact that the human body information and the first non-motor vehicle information have a riding relationship is detected;
and the first re-identification module is used for re-identifying based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
In a third aspect, an embodiment of the present invention provides an electronic device, including: the non-motor vehicle weight identification method comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the non-motor vehicle weight identification method provided by the embodiment of the invention.
In a fourth aspect, the embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements the steps in the non-motor vehicle weight identification method provided by the embodiment of the present invention.
In the embodiment of the invention, an image sequence to be processed is obtained, wherein the image sequence to be processed comprises a non-motor vehicle; aiming at the image sequence to be processed, carrying out non-motor vehicle detection and human body detection on each frame of image to be processed to obtain first non-motor vehicle information and human body information; if the fact that the human body information and the first non-motor vehicle have a riding relationship is detected, the first non-motor vehicle is associated with the human body information to obtain second non-motor vehicle information with attributes of riders; and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle. The non-motor vehicle is identified again through the non-motor vehicle information and the rider's body information, and the rider's body information can be used as the auxiliary information for identifying the weight of the non-motor vehicle, so that the accuracy of identifying the weight of the non-motor vehicle is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a non-motor vehicle weight recognition method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a predictive model training method according to an embodiment of the present invention;
FIG. 3 is a flow chart of a second non-motor vehicle information acquisition provided by an embodiment of the present invention;
FIG. 4 is a flowchart of a target tracking method according to an embodiment of the present invention;
FIG. 5 is a flow chart of another non-motor vehicle weight identification provided by an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a non-motor vehicle weight recognition device according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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.
Referring to fig. 1, fig. 1 is a flowchart of a non-motor vehicle heavy identification method according to an embodiment of the present invention, and as shown in fig. 1, the non-motor vehicle heavy identification method includes the following steps:
101. and acquiring an image sequence to be processed.
In the embodiment of the invention, the road can be shot in real time by the monitoring camera arranged at the designated position to obtain the image sequence to be detected of the non-motor vehicle, or the video of the non-motor vehicle uploaded by the user can be used as the image sequence to be detected.
Further, since the non-motor vehicle needs to be driven by the rider to travel on the road, the sequence of images to be processed includes the non-motor vehicle and the rider.
The image sequence to be detected comprises continuous multi-frame images to be processed, each frame of image to be processed can be used for carrying out target detection on the non-motor vehicles and the human bodies, and the image sequence to be detected can be used for carrying out target tracking on the non-motor vehicles and the human bodies.
102. And performing non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information.
In the embodiment of the invention, the target detection can be performed on each frame of image to be processed in the image sequence to be processed according to time sequence through the target detection model, and the target detection can be non-motor vehicle detection and human body detection.
Further, the first non-motor vehicle information can be obtained by detecting a non-motor vehicle in the image to be processed, and the non-motor vehicle information can include a non-motor vehicle image and a non-motor vehicle detection frame, wherein the non-motor vehicle image can be a small image obtained by cutting out the image to be processed through the non-motor vehicle detection frame.
The human body information can be obtained by performing human body detection on the image to be processed, and the human body information can comprise a human body image and a human body detection frame, wherein the human body image can be a small image obtained by intercepting the image to be processed through the human body detection frame.
It should be noted that the non-motor vehicle detection and the human body detection can be simultaneously performed by a multi-target detection model, and the detection targets of the multi-target detection model include two targets, one target may be a non-motor vehicle, and the other target may be a human body.
The image to be processed is input into the multi-target detection model, a non-motor vehicle detection result and a human body detection result are output, first non-motor vehicle information can be obtained through the non-motor vehicle detection result, and human body information can be obtained through the human body detection result. It should be noted that, the embodiment of the present invention is not limited to the multi-target detection model, and the existing multi-target detection model may be implemented to detect the non-motor vehicle and the human body, for example, the multi-target detection model may be based on the YOLO series detection model.
103. And when the fact that the riding relationship exists between the human body information and the first non-motor vehicle information is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with the attribute of the rider.
In the embodiment of the invention, whether the human body information and the first non-motor vehicle information have the riding relationship can be detected through the human body information and the first non-motor vehicle information, and the riding relationship can be understood as that a rider drives the non-motor vehicle.
Further, the human body information may include posture information, and it is determined whether the posture information matches the first non-motor vehicle information, and when the posture information matches the first non-motor vehicle information, it may be determined that the human body information and the first non-motor vehicle information have a riding relationship.
In a possible embodiment, the human body information includes a human body image, the non-motor vehicle includes a non-motor vehicle image, and the human body image and the non-motor vehicle image can be input into a pre-trained prediction model to predict whether the relationship between the human body and the non-motor vehicle is a riding relationship.
Optionally, the first non-motor vehicle information includes a non-motor vehicle detection frame, the human body information includes a human body detection frame, if it is detected that the human body information and the first non-motor vehicle have a riding relationship, the first non-motor vehicle and the human body information are correlated, and before the step of obtaining second non-motor vehicle information with a riding attribute, a merging ratio of the non-motor vehicle detection frame and the human body detection frame may be calculated, and a non-motor vehicle detection frame with the merging ratio larger than a preset value is selected to be combined with the human body detection frame, so as to obtain a candidate combination frame, where the candidate combination frame includes the non-motor vehicle detection frame and the human body detection frame; and performing behavior detection on the images in the candidate combination frame to obtain a behavior detection result of the images in the candidate combination frame, wherein the behavior detection result comprises whether the human body information and the first non-motor vehicle information have a riding relationship.
In the embodiment of the present invention, the intersection ratio (IoU) refers to a union of the non-vehicle detection frame and the human body detection frame in the intersection ratio of the non-vehicle detection frame and the human body detection frame, and when the intersection ratio is equal to 1, it indicates that the non-vehicle detection frame and the human body detection frame have the same size and overlap in position. Therefore, it can be seen that the larger the intersection ratio of the non-motor vehicle detection frame and the human body detection frame is, the closer the non-motor vehicle detection frame and the human body detection frame are and the more similar the shapes are, and the higher the probability that the person corresponding to the human body detection frame is a rider of the non-motor vehicle corresponding to the non-motor vehicle detection frame is.
And combining the non-motor vehicle detection frame with the intersection ratio larger than the preset value with the human body detection frame to obtain a candidate combination frame, wherein the candidate combination frame comprises the non-motor vehicle detection frame and the human body detection frame, and then capturing a corresponding image from the image to be processed through the candidate combination frame to serve as a candidate image, and the candidate image comprises the non-motor vehicle and the human body. And performing behavior detection on the candidate image, detecting whether a riding behavior line exists in the candidate image, and if the riding behavior line exists, indicating that the first non-motor vehicle information and the human body information have a riding relationship. And if the riding behaviors do not exist, discarding the candidate combination frame, and performing behavior detection on the next candidate combination frame until each non-motor vehicle corresponds to one riding person.
Optionally, in the step of selecting the non-motor vehicle detection frame with the intersection ratio larger than the preset value to be combined with the human body detection frame to obtain the candidate combination frame, a minimum bounding frame between the non-motor vehicle detection frame with the intersection ratio larger than the preset value and the human body detection frame can be calculated, and the minimum bounding frame covers the non-motor vehicle detection frame with the intersection ratio larger than the preset value and the human body detection frame at the same time, and has the smallest area; and determining the minimum bounding box as a candidate combined box. By taking the minimum bounding box of the non-motor vehicle detection frame and the human body detection frame as the candidate combination frame, the image size in the candidate combination frame can be reduced, and the detection accuracy of the riding relationship is improved.
In a possible embodiment, the image to be processed may be a depth image, so that an image intercepted by the candidate combo box is also a depth image, and behavior detection may be performed by including a non-motor vehicle, a human body and the depth image, so that accuracy of behavior detection is higher.
In the embodiment of the invention, the images in the candidate combination frame can be subjected to behavior detection through the pre-trained prediction model, so that whether the images in the candidate combination frame have riding behaviors or not is obtained. The prediction model can be constructed based on the resnet18 network.
Referring to fig. 2, fig. 2 is a flowchart of a predictive model training method according to an embodiment of the present invention, as shown in fig. 2, including the following steps:
201. and collecting riding positive and negative samples.
In the embodiment of the invention, the positive and negative samples of riding are collected to construct a data set, wherein the positive and negative samples are divided into a positive sample and a negative sample, the positive sample is a non-motor vehicle picture with a rider, and the negative sample comprises a non-motor vehicle, a pedestrian and other background pictures without the rider. The sample label for a positive sample may be "yes" and the sample label for a negative sample may be "no".
202. And (6) normalizing the image data.
In the embodiment of the invention, the image data normalization is performed on the images corresponding to the positive and negative samples in the data set, and the image data normalization refers to scaling the image pixel values to be between-1 and 1. Before normalizing the image data, the image corresponding to the positive and negative samples may be resized, such as by resize the image corresponding to the positive and negative samples to a size of 224 × 224 pixels.
203. And (5) training a model.
In the embodiment of the present invention, specifically, positive and negative samples in the data set may be input into the prediction model, and model parameter iteration is performed according to an error between an output result of the prediction model and a sample label, so as to update a model parameter of the prediction model.
204. And judging whether the Loss is less than the threshold value.
In the embodiment of the invention, when the error Loss of the prediction model is less than the preset threshold value, the training of the prediction model is completed, and the step 205 is carried out; when the predicted error Loss is greater than the predetermined threshold, step 206 is entered.
205. The output model is used for prediction.
In the embodiment of the invention, when the error Loss of the prediction model is less than the preset threshold value, the prediction model is output for prediction.
206. And updating the model parameters.
In the embodiment of the invention, when the error Loss of the prediction model is greater than the preset threshold value, model parameter iteration is continued to update the model parameters of the prediction model.
The riding relation of the first non-motor vehicle information and the human body information can be rapidly and accurately obtained by predicting the riding relation of the first non-motor vehicle information and the human body information through the prediction model.
When the first non-motor vehicle information and the human body information have a riding relationship, it is indicated that the person corresponding to the human body information is a rider of the non-motor vehicle corresponding to the first non-motor vehicle information, and at this time, the human body information and the first non-motor vehicle information can be associated to obtain second non-motor vehicle information, wherein the second non-motor vehicle information comprises the first non-motor vehicle information and the human body information having the riding relationship with the first non-motor vehicle.
Further, when the first non-motor vehicle information and the human body information have a riding relationship, corresponding human body images can be extracted from the images to be processed through the human body detection frame in the human body information to perform feature extraction, so that human body attribute features are obtained, and the human body attribute features are added to the first non-motor vehicle information to obtain second non-motor vehicle information.
Referring to fig. 3, fig. 3 is a flowchart of a second non-motor vehicle information acquisition method according to an embodiment of the present invention, as shown in fig. 3, including the following steps:
301. and acquiring monitoring video sequence images.
In an embodiment of the present invention, the monitoring video sequence image is the image sequence to be processed.
302. Non-motor vehicles and human body detection.
In the embodiment of the invention, non-motor vehicle and human body detection can be carried out on each frame of image in the monitoring video sequence image to obtain a non-motor vehicle detection frame and a human body detection frame.
303. Non-motorized vehicles and human bodies IoU.
In an embodiment of the present invention, IoU of the non-motor vehicle detection box and the human detection box are calculated. The reference IoU denotes that the intersection of the non-motor vehicle detection frame and the human body detection frame is compared with the union of the non-motor vehicle detection frame and the human body detection frame, and when the intersection ratio is equal to 1, it indicates that the non-motor vehicle detection frame and the human body detection frame have the same size and are overlapped in position. Therefore, it can be seen that the larger the intersection ratio of the non-motor vehicle detection frame and the human body detection frame is, the closer the non-motor vehicle detection frame and the human body detection frame are and the more similar the shapes are, and the higher the probability that the person corresponding to the human body detection frame is a rider of the non-motor vehicle corresponding to the non-motor vehicle detection frame is.
304. A determination IoU is made as to whether the threshold is satisfied.
In the embodiment of the invention, when IoU is greater than the threshold, the probability that the person corresponding to the human body detection frame is the non-motor vehicle rider corresponding to the non-motor vehicle detection frame is higher.
305. Combining non-motor vehicles and human targets, and covering the two targets by taking a minimum bounding box.
In an embodiment of the invention, the two objects covered by the minimum bounding box are non-motor vehicles and human bodies, respectively.
306. And intercepting the image by the smallest surrounding frame, and inputting the image into a CNN network for prediction.
In the embodiment of the invention, the minimum bounding box intercepted image comprises a non-motor vehicle image and a human body image, and the CNN convolutional neural network can be used as a prediction model to carry out riding prediction on the minimum bounding box intercepted image.
307. And judging whether to ride the bicycle.
In the embodiment of the present invention, if the bicycle is ridden, the process proceeds to step 308, and if the bicycle is not ridden, the process proceeds to step 310.
308. And (5) extracting characteristics of the rider.
In the embodiment of the invention, the characteristics of the rider can be extracted through a characteristic extraction technology.
309. Non-motorized vehicles add to the attributes of the rider.
In an embodiment of the present invention, a rider characteristic may be added to the first non-motor vehicle information as a rider attribute of the first non-motor vehicle information, thereby obtaining the second non-motor vehicle information.
310. The human target is discarded.
In the embodiment of the invention, when the human body is not the rider, the human body is discarded, namely the human body information which is not the rider is discarded, so that the non-motor vehicle weight recognition is conveniently carried out according to the human body information.
The corresponding human body image can be extracted from the image to be processed through the human body detection frame in the human body information to perform feature extraction, so that human body attribute features are obtained, and the human body attribute features are added into the first non-motor vehicle information to obtain second non-motor vehicle information.
104. And re-identifying based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
In an embodiment of the present invention, the second non-motor vehicle information includes the first non-motor vehicle information and the human body information of the rider, and further, the second non-motor vehicle information includes the first non-motor vehicle information and the human body characteristics of the participant.
And carrying out target tracking on the second non-motor vehicle information through a target tracking algorithm so as to obtain the track information of the non-motor vehicle, and carrying out non-motor vehicle weight identification according to the track information of the non-motor vehicle. The target tracking algorithm tld, deep sort, etc.
Optionally, target tracking may be performed on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle; and re-identifying according to the first track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
The first non-motor vehicle information and the human body information in the second non-motor vehicle information can be subjected to target tracking through target tracking algorithms such as tld and deep sort, so that the first track information of the non-motor vehicle is obtained. The first trajectory information is obtained by performing target tracking on the second non-motor vehicle information.
Optionally, in the step of performing target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain the first track information of the non-motor vehicle, the attribute of the rider in the second non-motor vehicle information may be extracted to extract human body information of the corresponding rider, and the target tracking is performed on the rider through the human body information; if the target tracking is successful, obtaining the track information of the rider based on the target tracking result; according to the track information of the rider, under the condition that the rider has a riding relationship with the non-motor vehicle, the track information of the non-motor vehicle is updated through the track information of the rider, and first track information of the non-motor vehicle is obtained.
The non-motor vehicle is tracked by tracking the rider, the track of the non-motor vehicle is obtained to identify the weight of the non-motor vehicle, and the accuracy of identifying the weight of the non-motor vehicle can be higher.
Optionally, if the target tracking fails, performing target tracking on the non-motor vehicle through the first non-motor vehicle information to obtain second track information of the non-motor vehicle; and re-identifying according to the second track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
When the target tracking is performed on the rider, the target may be lost due to occlusion or other reasons, and the target tracking may fail. And after the target tracking fails, tracking the non-motor vehicle through the first non-motor vehicle information to obtain the track of the non-motor vehicle, so that the accuracy rate of re-identification of the non-motor vehicle is further improved. The second trajectory information may be trajectory information obtained by performing target tracking using the first non-motor vehicle information, and the second trajectory information is different from the first trajectory information in that target tracking by the rider is not used.
Referring to fig. 4, fig. 4 is a flowchart of a target tracking method according to an embodiment of the present invention, as shown in fig. 4, including the following steps:
401. non-motor vehicle detection frame.
In the embodiment of the invention, the non-motor vehicle detection frame is input, and the non-motor vehicle detection frame can be a non-motor vehicle detection frame in the first non-motor vehicle information or a non-motor vehicle detection frame in the second non-motor vehicle information.
402. Whether or not there is a rider.
In the embodiment of the invention, whether the non-motor vehicle corresponding to the non-motor vehicle detection frame has a rider or not is judged, specifically, whether the non-motor vehicle information contains the riding attribute or not is judged, and if the non-motor vehicle information contains the riding attribute or not, the non-motor vehicle corresponding to the non-motor vehicle detection frame can be determined to have the rider.
403. And tracking the rider.
In an embodiment of the invention, the rider may be tracked by a target tracking algorithm.
404. Whether the tracing is successful.
In the embodiment of the present invention, it is determined whether the tracking of the rider is successful, and if so, the process proceeds to step 405, and if not, the process proceeds to step 406.
405. The non-motor vehicle and rider detection box updates the non-motor vehicle trajectory.
In the embodiment of the invention, the track of the non-motor vehicle is updated through the non-motor vehicle and rider detection frame (also called a candidate combination frame) to obtain the first track information of the non-motor vehicle.
406. Non-motor vehicle tracking.
In the embodiment of the invention, the non-motor vehicle can be tracked through a target tracking algorithm.
407. Whether the tracing is successful.
In the embodiment of the invention, whether the non-motor vehicle is tracked successfully or not is judged, if so, the step 409 is carried out, and if not, the step 408 is carried out.
408. A new non-motor vehicle trajectory is created.
In embodiments of the present invention, unsuccessful tracking of a non-motor vehicle indicates a potential new non-motor vehicle target, so a new non-motor vehicle trajectory may be created.
409. The non-motor vehicle detection frame updates the non-motor vehicle track.
In the embodiment of the invention, the non-motor vehicle track is updated through the non-motor vehicle detection frame to obtain the second track information of the non-motor vehicle.
410. Non-motor vehicle trajectory.
In the embodiment of the invention, the first track information and the second track information are both taken as the non-motor vehicle track information.
In the embodiment of the invention, each non-motor vehicle track corresponds to one identification characteristic, the identification characteristic can be id of the non-motor vehicle track, and the problems of shielding, mirror crossing and the like in the tracking process are considered, so that a target is easy to interrupt, and a plurality of tracks are stored in the same non-motor vehicle. It is therefore desirable to merge the collected non-motor vehicle trajectories.
Optionally, a first target image sequence of the first track information in the image sequence to be processed is obtained; extracting a first identification feature corresponding to the first track information according to the first target image sequence; calculating the feature similarity of all the first track information based on the first identification features, and combining the first track information with the feature similarity larger than a similarity threshold value to obtain combined track information; and combining the first track information by combining the track information.
Further, a second target image sequence of second track information in the image sequence to be processed is obtained; extracting a second identification feature corresponding to the second track information according to the second target image sequence; calculating the feature similarity of all second track information based on the second identification features, and combining the second track information with the feature similarity larger than a similarity threshold value to obtain combined track information; and combining the second track information by combining the track information.
Further, calculating the feature similarity of the combined second track information and the combined first track information, and combining the track information with the feature similarity larger than a similarity threshold value to obtain combined track information; by combining the track information, all the track information is merged.
Specifically, the identification feature may be an id of the non-motor vehicle track, and the feature of all pictures in the non-motor vehicle track may be extracted first, and an average value of the features is taken as the id feature of the non-motor vehicle track. And then calculating the feature similarity among all the id features. And the id features with the feature similarity larger than the threshold are considered to be possible to be the same non-motor vehicle, are divided into a class and are combined, and a corresponding id combination list is output. And finally, providing the recommended id combination list for a labeling person to manually confirm whether the ids in the combination are the same non-motor vehicle or not.
By the process, a large amount of non-motor vehicle id data can be rapidly collected and can be used for training a non-motor vehicle picture feature extraction network model. In the actual use stage, the trained model is used, a non-motor vehicle picture is input, and the non-motor vehicle depth feature is output and used for searching the target non-motor vehicle in the mass data.
Referring to fig. 5, fig. 5 is a flowchart of another non-motor vehicle weight recognition method according to an embodiment of the present invention, as shown in fig. 5, including the following steps:
501. and (4) capturing images of the non-motor vehicle.
In an embodiment of the present invention, the non-motor vehicle captured image refers to an image captured during a non-motor vehicle weight recognition process. The non-motor vehicle snapshot comprises the step of snapshot of the non-motor vehicle.
502. And (6) image normalization.
In the embodiment of the invention, the image data normalization is carried out on the non-motor vehicle snapshot image, and the image data normalization refers to the scaling of the image pixel value to be between-1 and 1. Prior to normalizing the image data, the non-motor vehicle snap-shot image may be resized, such as to a 224 x 224 pixel size.
503. And (5) extracting a model from the features.
In the embodiment of the invention, the non-motor vehicle snapshot image subjected to image normalization is input to the feature extraction model.
504. A non-motor vehicle feature is snap-shot.
In the embodiment of the invention, the non-motor vehicle characteristics are extracted through the characteristic extraction model.
505. The characteristic cosine maximum similarity S.
In the embodiment of the invention, the cosine similarity between the target non-motor vehicle feature vector obtained in the step 508 and the snap-shot non-motor vehicle feature extracted by the feature extraction model is calculated to obtain the cosine maximum similarity S between the snap-shot non-motor vehicle feature and the target non-motor vehicle feature.
506. S > threshold is true.
In the embodiment of the present invention, if S > threshold is true, the process proceeds to step 507, and if S > threshold is true, the process proceeds to step 509.
507. Is a search target.
In the embodiment of the invention, if the threshold value S is satisfied, the target non-motor vehicle is similar to the snapshot non-motor vehicle.
508. Target non-motor vehicle feature vectors.
In the embodiment of the present invention, the target non-motor vehicle feature vector is extracted from the image sequence to be processed according to the non-motor vehicle trajectory in the steps 101 to 104, and specifically, may be features of all pictures under the non-motor vehicle trajectory, and an average value of the features is taken as an id feature of the non-motor vehicle trajectory.
509. Not the search target.
In the embodiment of the invention, if the S > threshold value is not satisfied, the target non-motor vehicle is not similar to the snapshot non-motor vehicle.
510. And (6) ending.
In the embodiment of the invention, an image sequence to be processed is obtained, wherein the image sequence to be processed comprises a non-motor vehicle; aiming at the image sequence to be processed, carrying out non-motor vehicle detection and human body detection on each frame of image to be processed to obtain first non-motor vehicle information and human body information; if the fact that the human body information and the first non-motor vehicle have a riding relationship is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with attributes of riders; and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle. The non-motor vehicle is identified again through the non-motor vehicle information and the rider's body information, and the rider's body information can be used as the auxiliary information for identifying the weight of the non-motor vehicle, so that the accuracy of identifying the weight of the non-motor vehicle is improved.
It should be noted that the non-motor vehicle weight recognition method provided by the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. capable of performing non-motor vehicle weight recognition.
Optionally, referring to fig. 6, fig. 6 is a schematic structural diagram of a non-motor vehicle weight recognition device according to an embodiment of the present invention, and as shown in fig. 6, the device includes:
a first obtaining module 601, configured to obtain a to-be-processed image sequence, where the to-be-processed image sequence includes a non-motor vehicle;
a first detection module 602, configured to perform non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed, so as to obtain first non-motor vehicle information and human body information;
the association module 603 is configured to, when it is detected that the human body information and the first non-motor vehicle information have a riding relationship, associate the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with a rider attribute;
and a first re-identification module 604, configured to perform re-identification based on the second non-motor vehicle information, so as to obtain a re-identification result of the non-motor vehicle.
Optionally, before the associating module 603, the apparatus includes:
the calculation module is used for calculating the intersection ratio of the non-motor vehicle detection frame and the human body detection frame, selecting the non-motor vehicle detection frame with the intersection ratio larger than a preset value and combining the non-motor vehicle detection frame with the human body detection frame to obtain a candidate combination frame, and the candidate combination frame comprises the non-motor vehicle detection frame and the human body detection frame;
and the second detection module is used for carrying out behavior detection on the images in the candidate combination frame to obtain a behavior detection result of the images in the candidate combination frame, wherein the behavior detection result comprises whether the human body information and the first non-motor vehicle information have a riding relationship or not.
Optionally, the calculating module includes:
the calculation submodule is used for calculating a minimum enclosing frame between the non-motor vehicle detection frame and the human body detection frame, the intersection ratio of which is greater than a preset value, and the minimum enclosing frame covers the non-motor vehicle detection frame and the human body detection frame, the intersection ratio of which is greater than the preset value, and the area of the minimum enclosing frame is the smallest;
a determining submodule, configured to determine the minimum bounding box as the candidate combo box.
Optionally, the re-identifying module 604 includes:
the tracking submodule is used for carrying out target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle;
and the re-identification submodule is used for re-identifying according to the first track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, the tracking sub-module includes:
the extraction unit is used for extracting the attribute of the rider in the second non-motor vehicle information, extracting the human body information corresponding to the rider and carrying out target tracking on the rider through the human body information;
the tracking unit is used for obtaining the trajectory information of the rider based on a target tracking result if the target tracking is successful;
and the updating unit is used for updating the track information of the non-motor vehicle according to the track information of the rider under the condition that the rider has a riding relationship with the non-motor vehicle, so as to obtain first track information of the non-motor vehicle.
Optionally, the apparatus further comprises:
the tracking module is used for tracking the target of the non-motor vehicle through the first non-motor vehicle information to obtain second track information of the non-motor vehicle if the target tracking fails;
and the second re-identification module is used for re-identifying according to the second track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, the apparatus includes:
the second acquisition module is used for acquiring a first target image sequence of the first track information in the image sequence to be processed;
the extraction module is used for extracting a first identification feature corresponding to the first track information according to the first target image sequence;
the combination module is used for calculating the feature similarity of all the first track information based on the first identification features and combining the first track information with the feature similarity larger than a similarity threshold value to obtain combined track information;
and the merging module is used for merging the first track information through the combined track information.
The non-motor vehicle weight recognition device provided by the embodiment of the invention can be applied to equipment such as a smart phone, a computer and a server which can perform non-motor vehicle weight recognition.
The non-motor vehicle weight recognition device provided by the embodiment of the invention can realize each process realized by the non-motor vehicle weight recognition method in the method embodiment, and can achieve the same beneficial effects. To avoid repetition, further description is omitted here.
Referring to fig. 7, fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, as shown in fig. 7, including: a memory 702, a processor 701 and a computer program of a non-motor vehicle weight recognition method stored on the memory 702 and executable on the processor 701, wherein:
the processor 701 is configured to call the computer program stored in the memory 702, and perform the following steps:
acquiring an image sequence to be processed, wherein the image sequence to be processed comprises a non-motor vehicle;
carrying out non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information;
when the fact that the human body information and the first non-motor vehicle information have a riding relationship is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with attributes of riders;
and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
Optionally, the first non-motor vehicle information executed by the processor 701 includes a non-motor vehicle detection frame, the human body information includes a human body detection frame, and before the step of associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with a property of a rider when it is detected that the human body information and the first non-motor vehicle information have a riding relationship, the method includes:
calculating the intersection ratio of the non-motor vehicle detection frame and the human body detection frame, and selecting the non-motor vehicle detection frame with the intersection ratio larger than a preset value to be combined with the human body detection frame to obtain a candidate combination frame, wherein the candidate combination frame comprises the non-motor vehicle detection frame and the human body detection frame;
and performing behavior detection on the images in the candidate combination frame to obtain a behavior detection result of the images in the candidate combination frame, wherein the behavior detection result comprises whether the human body information and the first non-motor vehicle information have a riding relationship.
Optionally, the selecting, performed by the processor 701, the non-motor vehicle detection frame with the intersection ratio larger than the preset value is combined with the human body detection frame to obtain a candidate combination frame, where the selecting and combining ratio is larger than the preset value, and the candidate combination frame includes:
calculating a minimum surrounding frame between the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than a preset value, wherein the minimum surrounding frame covers the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than the preset value at the same time and has the minimum area;
determining the minimum bounding box as the candidate combo box.
Optionally, the performing, by the processor 701, re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle includes:
carrying out target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle;
and re-identifying according to the first track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, the performing, by the processor 701, target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle includes:
extracting attributes of the rider in the second non-motor vehicle information, extracting human body information corresponding to the rider, and carrying out target tracking on the rider through the human body information;
if the target tracking is successful, obtaining the trajectory information of the rider based on a target tracking result;
according to the track information of the rider, under the condition that the rider has a riding relationship with the non-motor vehicle, the track information of the non-motor vehicle is updated through the track information of the rider, and first track information of the non-motor vehicle is obtained.
Optionally, after the step of performing target tracking on the rider through the human body information, the processor 701 further includes:
if the target tracking fails, performing target tracking on the non-motor vehicle through the first non-motor vehicle information to obtain second track information of the non-motor vehicle;
and re-identifying according to the second track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
Optionally, after obtaining the first trajectory information of the non-motor vehicle, the processor 701 executes the method including:
acquiring a first target image sequence of first track information in the image sequence to be processed;
extracting a first identification feature corresponding to the first track information according to the first target image sequence;
calculating feature similarity of all the first track information based on the first identification features, and combining the first track information with the feature similarity larger than a similarity threshold value to obtain combined track information;
and merging the first track information through the combined track information.
The electronic equipment provided by the embodiment of the invention can realize each process realized by the non-motor vehicle weight identification method in the method embodiment, and can achieve the same beneficial effects. To avoid repetition, further description is omitted here.
The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements each process of the non-motor vehicle re-identification method or the application-side non-motor vehicle re-identification method provided in the embodiment of the present invention, and can achieve the same technical effect, and in order to avoid repetition, the computer program is not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a computer readable storage medium and executed by a computer to implement the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above disclosure is only for the purpose of illustrating the preferred embodiments of the present invention, and it is therefore to be understood that the invention is not limited by the scope of the appended claims.
Claims (10)
1. A non-motor vehicle weight recognition method is characterized by comprising the following steps:
acquiring an image sequence to be processed, wherein the image sequence to be processed comprises a non-motor vehicle;
performing non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information;
when the fact that the human body information and the first non-motor vehicle information have a riding relationship is detected, associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with attributes of riders;
and performing re-identification based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
2. The method of claim 1, wherein the first non-motor vehicle information comprises a non-motor vehicle detection box, the human body information comprises a human body detection box, and prior to the step of associating the first non-motor vehicle with the human body information when the human body information is detected to have a riding relationship with the first non-motor vehicle information, resulting in second non-motor vehicle information having rider attributes, the method comprises:
calculating the intersection ratio of the non-motor vehicle detection frame and the human body detection frame, and selecting the non-motor vehicle detection frame with the intersection ratio larger than a preset value to be combined with the human body detection frame to obtain a candidate combination frame, wherein the candidate combination frame comprises the non-motor vehicle detection frame and the human body detection frame;
and performing behavior detection on the images in the candidate combination frame to obtain a behavior detection result of the images in the candidate combination frame, wherein the behavior detection result comprises whether the human body information and the first non-motor vehicle information have a riding relationship.
3. The method of claim 2, wherein the selecting the non-motor vehicle detection frame with the intersection ratio greater than a preset value to be combined with the human body detection frame to obtain a candidate combination frame comprises:
calculating a minimum surrounding frame between the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than a preset value, wherein the minimum surrounding frame covers the non-motor vehicle detection frame and the human body detection frame with the intersection ratio larger than the preset value at the same time and has the minimum area;
and determining the minimum bounding box as the candidate combined box.
4. The method of claim 3, wherein said re-identifying based on said second non-motor vehicle information to obtain a re-identification result of said non-motor vehicle comprises:
carrying out target tracking on the image sequence to be processed according to the second non-motor vehicle information to obtain first track information of the non-motor vehicle;
and re-identifying according to the first track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
5. The method of claim 4, wherein the target tracking the sequence of images to be processed according to the second non-motor vehicle information to obtain first trajectory information of the non-motor vehicle comprises:
extracting attribute of the rider in the second non-motor vehicle information, extracting human body information corresponding to the rider, and carrying out target tracking on the rider through the human body information;
if the target tracking is successful, obtaining the trajectory information of the rider based on a target tracking result;
according to the track information of the rider, under the condition that the rider has a riding relationship with the non-motor vehicle, the track information of the non-motor vehicle is updated through the track information of the rider, and first track information of the non-motor vehicle is obtained.
6. The method of claim 5, wherein after the step of target tracking the rider via the body information, the method further comprises:
if the target tracking fails, performing target tracking on the non-motor vehicle through the first non-motor vehicle information to obtain second track information of the non-motor vehicle;
and re-identifying according to the second track information of the non-motor vehicle to obtain a re-identification result of the non-motor vehicle.
7. The method of claim 4, wherein after obtaining the first trajectory information for the non-motor vehicle, the method comprises:
acquiring a first target image sequence of first track information in the image sequence to be processed;
extracting a first identification feature corresponding to the first track information according to the first target image sequence;
calculating feature similarity of all the first track information based on the first identification features, and combining the first track information with the feature similarity larger than a similarity threshold value to obtain combined track information;
and merging the first track information through the combined track information.
8. A non-motor vehicle weight identification device, the device comprising:
the device comprises a first acquisition module, a second acquisition module and a processing module, wherein the first acquisition module is used for acquiring an image sequence to be processed, and the image sequence to be processed comprises a non-motor vehicle;
the first detection module is used for carrying out non-motor vehicle detection and human body detection on each frame of image to be processed in the image sequence to be processed to obtain first non-motor vehicle information and human body information;
the association module is used for associating the first non-motor vehicle with the human body information to obtain second non-motor vehicle information with the attribute of a rider when the human body information and the first non-motor vehicle information are detected to have a riding relationship;
and the first re-identification module is used for re-identifying based on the second non-motor vehicle information to obtain a re-identification result of the non-motor vehicle.
9. An electronic device, comprising: memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the non-motor vehicle weight identification method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps in the non-motor vehicle weight recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111663748.6A CN114445787A (en) | 2021-12-31 | 2021-12-31 | Non-motor vehicle weight recognition method and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111663748.6A CN114445787A (en) | 2021-12-31 | 2021-12-31 | Non-motor vehicle weight recognition method and related equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114445787A true CN114445787A (en) | 2022-05-06 |
Family
ID=81366630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111663748.6A Pending CN114445787A (en) | 2021-12-31 | 2021-12-31 | Non-motor vehicle weight recognition method and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114445787A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114444895A (en) * | 2021-12-31 | 2022-05-06 | 深圳云天励飞技术股份有限公司 | Cleaning quality evaluation method and related equipment |
-
2021
- 2021-12-31 CN CN202111663748.6A patent/CN114445787A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114444895A (en) * | 2021-12-31 | 2022-05-06 | 深圳云天励飞技术股份有限公司 | Cleaning quality evaluation method and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hassaballah et al. | Vehicle detection and tracking in adverse weather using a deep learning framework | |
| Shah et al. | CADP: A novel dataset for CCTV traffic camera based accident analysis | |
| CN109558823B (en) | Vehicle identification method and system for searching images by images | |
| US20060067562A1 (en) | Detection of moving objects in a video | |
| CN108009466B (en) | Pedestrian detection method and device | |
| CN114155284A (en) | Pedestrian tracking method, device, equipment and medium based on multi-target pedestrian scene | |
| CN112434566B (en) | Passenger flow statistics method and device, electronic equipment and storage medium | |
| CN111881322B (en) | Target searching method and device, electronic equipment and storage medium | |
| CN112381132A (en) | Target object tracking method and system based on fusion of multiple cameras | |
| CN110610123A (en) | Multi-target vehicle detection method and device, electronic equipment and storage medium | |
| CN110533661A (en) | Adaptive real-time closed-loop detection method based on characteristics of image cascade | |
| CN111401308A (en) | Fish behavior video identification method based on optical flow effect | |
| CN112836683A (en) | License plate recognition method, device, equipment and medium for portable camera equipment | |
| CN113033523B (en) | Method and system for constructing falling judgment model and falling judgment method and system | |
| CN114445787A (en) | Non-motor vehicle weight recognition method and related equipment | |
| CN113112479A (en) | Progressive target detection method and device based on key block extraction | |
| CN117423157A (en) | Mine abnormal video action understanding method combining migration learning and regional invasion | |
| CN116311166A (en) | Traffic obstacle recognition method and device and electronic equipment | |
| CN111862147A (en) | Method for tracking multiple vehicles and multiple human targets in video | |
| CN116912763A (en) | Multi-pedestrian re-recognition method integrating gait face modes | |
| CN113869163B (en) | Target tracking method and device, electronic equipment and storage medium | |
| CN115953744A (en) | Vehicle identification tracking method based on deep learning | |
| CN115393755A (en) | Visual target tracking method, device, equipment and storage medium | |
| CN115311680A (en) | Human body image quality detection method and device, electronic equipment and storage medium | |
| CN114494355A (en) | Trajectory analysis method and device based on artificial intelligence, terminal equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |