CN112949486B - Intelligent traffic data processing method and device based on neural network - Google Patents
Intelligent traffic data processing method and device based on neural network Download PDFInfo
- Publication number
- CN112949486B CN112949486B CN202110227091.2A CN202110227091A CN112949486B CN 112949486 B CN112949486 B CN 112949486B CN 202110227091 A CN202110227091 A CN 202110227091A CN 112949486 B CN112949486 B CN 112949486B
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- result
- image
- data
- network
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The utility model relates to an intelligent traffic data processing method and device based on neural network, the method includes responding to the intelligent traffic data acquisition command, triggering the camera to acquire the image data, the shooting area of the image data forms the target area; carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result; obtaining corresponding pedestrian position data according to each pedestrian in the pedestrian identification result; acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment, and acquiring each indoor positioning data in the indoor positioning results; and uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space. The storage scheme can be more reasonable by utilizing the embodiment of the disclosure.
Description
Technical Field
The present disclosure relates to the field of data processing, and in particular, to an intelligent traffic data processing method and apparatus based on a neural network.
Background
With the development of science and technology, more and more related services using geographic position data as core data are provided, such as a map service, a navigation service, a data analysis service, and even other personalized customization services with high added value brought to users based on the data analysis service. The role and importance of the geographic location data in the application service are self-evident, and how to properly acquire, store and apply the geographic location data needs to be researched.
Disclosure of Invention
The disclosure provides an intelligent traffic data processing method and device based on a neural network. The technical scheme of the disclosure is as follows:
the intelligent traffic data processing method based on the neural network comprises the following steps:
responding to an intelligent traffic data acquisition instruction, triggering a camera to acquire image data, wherein a shooting area of the image data forms a target area;
carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result;
obtaining corresponding pedestrian position data according to each pedestrian in the pedestrian identification result;
acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment, and acquiring each indoor positioning data in the indoor positioning results;
and uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space.
Preferably, there is no overlap between the shooting areas of different cameras, and the union of the shooting areas of the cameras is the target area.
Preferably, the neural network may be a preset pedestrian recognition model, and the training method of the pedestrian recognition model includes:
acquiring an image sample set, wherein each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing a region with a pedestrian from a region without the pedestrian, and the labeling result of the image sample represents the pedestrian in the image sample;
inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result, obtaining an image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into the target identification network to obtain a pedestrian identification result; and obtaining image recognition loss according to the pedestrian recognition result and the labeling result, and adjusting parameters of the feature extraction network and the target recognition network according to the image recognition loss until the preset convergence condition is reached.
Preferably, the method further comprises determining a time interval formed by the first N minutes and the last N minutes of the intelligent traffic data acquisition instruction acquisition time as the time slice.
Preferably, the uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space includes:
obtaining effective position data of the target area, wherein the effective position data comprises the pedestrian position data and the indoor positioning data which fall into the target area;
carrying out grid division on the target area, and numbering the divided grids according to a preset sequence;
according to the relation between the effective position data in the target area and the grids, carrying out grid growth according to the grids to obtain grid growth results;
and storing the effective position data on cloud nodes of a cloud storage space according to a grid growing result.
Intelligent traffic data processing device based on neural network includes:
the intelligent traffic data acquisition module is used for responding to an intelligent traffic data acquisition instruction and triggering a camera to acquire image data, and a shooting area of the image data forms a target area;
the pedestrian recognition module is used for carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result;
the pedestrian position data acquisition module is used for acquiring corresponding pedestrian position data according to each pedestrian in the pedestrian identification result;
the indoor positioning data acquisition module is used for acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment and acquiring each indoor positioning data in the indoor positioning results;
and the storage module is used for uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space.
Preferably, the method further comprises the following steps:
the system comprises an image sample set acquisition module, a pedestrian detection module and a pedestrian detection module, wherein the image sample set acquisition module is used for acquiring an image sample set, each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing a region with a pedestrian from a region without the pedestrian, and the annotation result of the image sample represents the pedestrian in the image sample;
the first training module is used for inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result, obtaining an image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
the second training module is used for inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into the target identification network to obtain a pedestrian identification result; obtaining image identification loss according to the pedestrian identification result and the labeling result, and adjusting parameters of the feature extraction network and the target identification network according to the image identification loss until the preset convergence condition is reached
Preferably, the storage module includes:
a valid position data determination unit, configured to acquire valid position data of the target area, where the valid position data includes the pedestrian position data and the indoor positioning data that fall within the target area;
the numbering unit is used for carrying out grid division on the target area and numbering the divided grids according to a preset sequence;
the growth unit is used for carrying out grid growth according to the grid according to the relation between the effective position data in the target area and the grid to obtain a grid growth result;
and the storage unit is used for storing the effective position data on cloud nodes of the cloud storage space according to the grid growth result.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:
according to the intelligent traffic data processing method based on the neural network, intelligent traffic data can be obtained through an artificial intelligence means and an infinite positioning means, and the intelligent traffic data can be uniformly stored in the cloud storage nodes, so that data inclination is avoided, the utilization rate of a data storage space is high, and the storage scheme is more reasonable.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.
Fig. 1 is a schematic diagram illustrating a neural network-based intelligent traffic data processing method according to an exemplary embodiment.
FIG. 2 is a flow diagram illustrating a method of training a pedestrian recognition model in accordance with an exemplary embodiment;
FIG. 3 is a schematic flow diagram illustrating the uniform storage of the pedestrian location data and the indoor location data in cloud storage space according to an exemplary embodiment;
FIG. 4 is a schematic diagram illustrating a process for storing the valid location data on cloud nodes of a cloud storage space according to grid growth results, according to an example embodiment;
FIG. 5 is a flow diagram illustrating a visual representation of a population distribution based on the management table in accordance with an exemplary embodiment;
fig. 6 is a block diagram illustrating a neural network-based intelligent traffic data processing apparatus according to an exemplary embodiment.
Detailed Description
In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
Referring to fig. 1, fig. 1 illustrates an intelligent traffic data processing method based on a neural network according to an exemplary embodiment, and as shown in fig. 1, the method includes:
s101, responding to an intelligent traffic data acquisition instruction, triggering a camera to acquire image data, wherein a shooting area of the image data forms a target area.
In the embodiment of the present disclosure, there is no overlapping in the shooting areas of different cameras, and the union of the shooting areas of the cameras is the target area.
And S102, carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result.
In the embodiment of the present disclosure, the neural network may be a preset pedestrian recognition model, and a training method of the pedestrian recognition model is shown in fig. 2, and includes:
s1021, an image sample set is obtained, each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing an area where a pedestrian exists and an area where the pedestrian does not exist, and the labeling result of the image sample represents the pedestrian in the image sample.
S1022, inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result, obtaining an image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
s1023, inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into the target identification network to obtain a pedestrian identification result; and obtaining image recognition loss according to the pedestrian recognition result and the labeling result, and adjusting parameters of the feature extraction network and the target recognition network according to the image recognition loss until the preset convergence condition is reached.
The image samples of steps S1023 and S1022 may be the same or different, in the embodiment of the present invention, in step S1022, the image segmentation network is trained with fixed parameters, and the network after the fixed parameters is applied to the training of the target identification network of step S1023, the paths of parameter feedback adjustment in the two steps are different, and the timing sequence is also different, which is obviously different from the related art.
In the embodiment of the invention, the segmentation network is arranged in front of the target recognition network, and higher semantic features for recognition can be provided for the target recognition network according to the information output by the segmentation network in practical application, so that a more accurate pedestrian recognition result is obtained. In the related art, the segmentation is usually positioned after the identification network, and the segmentation is finally finished according to the identification result.
And S1024, determining the feature extraction network, the image segmentation network and the target identification network which are subjected to parameter adjustment as the pedestrian identification model.
And S103, obtaining corresponding pedestrian position data according to each pedestrian in the pedestrian recognition result.
The shooting area of each camera is known, and corresponding pedestrian position data can be obtained for each pedestrian in the image corresponding to the shooting area.
S104, acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment, and acquiring each indoor positioning data in the indoor positioning results.
In the embodiment of the invention, a time interval formed by the first N minutes and the last N minutes of the intelligent traffic data acquisition instruction acquisition time can be determined as the time segment, and the embodiment of the invention is not limited by N.
In this disclosed embodiment, the indoor positioning data includes information such as user identification, user location data, user application program state, and wherein user location data and user identification are the fixed fields of the indoor positioning data, the indoor positioning data may also include any other fields, and no further description is given here.
Usually, when the user is located in the building, the GPS signal is difficult to accurately position the user, so that the indoor positioning request of the user can be collected, the indoor positioning request is obtained, indoor positioning service is provided for the user, a corresponding indoor positioning response is fed back, and indoor positioning data reported by the user according to the indoor positioning response is obtained.
Specifically, the obtaining of the indoor positioning request provides an indoor positioning service for the user, and then feeds back a corresponding indoor positioning response, including:
s1, signal information in the indoor positioning request is obtained, wherein the signal information is composed of at least one communication item, and the communication item comprises a signal source identification of a wireless communication hotspot and signal source strength.
S2, acquiring a position record set, wherein the position record in the position record set represents the corresponding relation between signal information and user position data.
And S3, calculating the correlation between the signal information in the indoor positioning request and the signal information of the position record in the position record set.
In a possible embodiment, for two pieces of signal information, a common signal source identifier and a signal source strength corresponding to the common signal source identifier in each piece of signal information may be extracted, and a signal strength difference corresponding to the common signal source identifier is calculated according to the signal source strength corresponding to each piece of signal information, so as to obtain a correspondence between each common signal source identifier and the corresponding signal strength difference. Sequencing according to the ascending sequence of the signal intensity difference to obtain a common signal source sequence, and based on the common signal source sequence, according to a formula sigma qiθiAnd calculating the correlation between the two signal information. i is the serial number of each signal source of the signal source sequence, qiIs a weight value, which can be set in a preset relevancy strategy, thetaiThe similarity value corresponding to the intensity difference is obtained by calculating the corresponding relation between the similarity value and the intensity difference in the correlation strategy. In the relevancy strategy, the higher the ranking is, the higher the weight is; the smaller the intensity difference, the higher the similarity value.
Illustratively, if signal source a is present for both signal information, with intensities of-30 dbm and-40 dbm, respectively, the difference in intensity is | (-30) - (-40) | ═ 10 dbm.
And S4, taking the user position data corresponding to the signal information in the position record set with the highest correlation as target user position data, and correspondingly generating an indoor positioning response.
And S105, uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space.
Specifically, in order to ensure the density uniformity of data distribution in the cloud storage space, the pedestrian position data and the indoor positioning data may be stored with the user position data in the pedestrian position data and the indoor positioning data as a reference.
Specifically, the uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space, as shown in fig. 3, includes:
s1051, effective position data of the target area are obtained, and the effective position data comprise the pedestrian position data and the indoor positioning data which fall into the target area.
And S1052, carrying out grid division on the target area, and numbering the divided grids according to a preset sequence.
In the embodiment of the present invention, a specific division method of the grids is not limited, and the grids may be numbered in the sequence of rows or columns, for example, if there are 5 rows and 5 columns of grids, the first row is numbered 1 to 5, the second row is numbered 6 to 10, and so on.
And S1053, according to the relation between the effective position data in the target area and the grids, carrying out grid growth according to the grids to obtain grid growth results.
Specifically, the grid growing according to the grid to obtain a grid growing result includes: randomly distributing a preset number of growing seeds in the target area, and growing the growing seeds in the target area, wherein the growing process of the growing seeds is a process of covering grids in a growing interval corresponding to the growing seeds, until the total number difference of the effective position data in each growing interval is smaller than a preset difference threshold after the growing is finished.
And S1054, storing the effective position data on cloud nodes of a cloud storage space according to a grid growth result.
Specifically, the storing the effective position data on a cloud node of a cloud storage space according to a grid growth result, as shown in fig. 4, includes:
s10541, generating a grid growth record according to the grid growth result, and storing the grid growth record in a preset management table, wherein the grid growth record comprises a time slice, a boundary of each growth interval and a cloud storage serial number corresponding to each growth interval.
Specifically, the time slice is determined in step S104, and the boundary of the growth interval is determined in step S1053, and the method for acquiring the cloud storage sequence number corresponding to each growth interval in the embodiment of the present invention includes:
s105411, extracting each label in each growth interval, and taking the ascending ordering result of each label as the temporary code of the growth interval.
S105412, taking the temporary coded hash result as the cloud storage serial number of the growth interval.
When data is queried, a time slice can be determined according to time in a query condition, the grid growth record is queried in the management table, a corresponding growth interval is queried according to position information in the query condition, a cloud storage serial number is obtained, and data meeting the query condition can be queried in a cloud storage node pointed by the cloud storage serial number.
In the embodiment of the invention, the basic cloud storage node corresponds to the first cloud storage serial number, and the serial numbers of other cloud storage nodes are set according to the hop count of each other cloud storage node relative to the basic cloud storage node. Illustratively, when they are located in the same rack, the number of hops between them is 2; when the machine room is positioned on adjacent racks in the same machine room, through 2-stage exchange, the hop number between the adjacent racks is 4; and it is located in different machine rooms, and through 3-stage switching, the hop count between them is 6.
S10542, for each growth interval in the grid growth record, storing effective position data falling into the growth interval in a cloud storage node pointed by a cloud storage serial number corresponding to the growth interval, and storing other data related to the effective position data in the cloud storage node.
The data can be stored in the cloud storage node pointed by the cloud storage serial number, and if the cloud storage node has abnormal conditions such as insufficient space, the cloud storage node is skipped over, and the next cloud storage node is used as the cloud storage node of the data, and so on.
In the embodiment of the present invention, other data associated with the valid position data is not specifically limited, and for example, if the valid position data is pedestrian position data, the other data may further include a current state of a pedestrian and a surrounding environment. If the valid location data is user location data, the other data may also be user identification, user application program state, and other information.
In a possible embodiment, the visual expression of the population distribution can be performed according to the management table, as shown in fig. 5, including:
s201, acquiring a time period to be rendered, and extracting each time segment of the time period to be rendered.
S202, for each time slice, inquiring the management table to obtain a corresponding grid growth record.
And S203, for each grid growth record, acquiring a cloud storage serial number corresponding to each growth interval of the grid growth record.
S204, accessing the cloud storage nodes pointed by the cloud storage serial numbers, counting the number of people corresponding to the growth interval according to the access result, and calculating the number of people corresponding to the growth interval.
S205, rendering the crowd distribution map of each time slice in the time period to be rendered, wherein the crowd distribution map comprises a growth interval and the number of crowds in the growth interval.
According to the intelligent traffic data processing method based on the neural network, intelligent traffic data can be obtained through an artificial intelligence means and an infinite positioning means, and the intelligent traffic data can be uniformly stored in the cloud storage nodes, so that data inclination is avoided, the utilization rate of a data storage space is high, and the storage scheme is more reasonable.
The embodiment of the present disclosure further provides an intelligent traffic data processing apparatus based on a neural network, as shown in fig. 6, including:
the intelligent traffic data acquisition module is used for responding to an intelligent traffic data acquisition instruction and triggering a camera to acquire image data, and a shooting area of the image data forms a target area;
the pedestrian recognition module is used for carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result;
the pedestrian position data acquisition module is used for acquiring corresponding pedestrian position data according to each pedestrian in the pedestrian identification result;
the indoor positioning data acquisition module is used for acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment and acquiring each indoor positioning data in the indoor positioning results;
and the storage module is used for uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space.
In one embodiment, further comprising:
the system comprises an image sample set acquisition module, a pedestrian detection module and a pedestrian detection module, wherein the image sample set acquisition module is used for acquiring an image sample set, each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing a region with a pedestrian from a region without the pedestrian, and the annotation result of the image sample represents the pedestrian in the image sample;
the first training module is used for inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result, obtaining an image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting the parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
the second training module is used for inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into the target identification network to obtain a pedestrian identification result; obtaining image identification loss according to the pedestrian identification result and the labeling result, and adjusting parameters of the feature extraction network and the target identification network according to the image identification loss until the preset convergence condition is reached
In one embodiment, the storage module includes:
a valid position data determination unit, configured to acquire valid position data of the target area, where the valid position data includes the pedestrian position data and the indoor positioning data that fall within the target area;
the numbering unit is used for carrying out grid division on the target area and numbering the divided grids according to a preset sequence;
the growth unit is used for carrying out grid growth according to the grid according to the relation between the effective position data in the target area and the grid to obtain a grid growth result;
and the storage unit is used for storing the effective position data on cloud nodes of the cloud storage space according to the grid growth result.
The embodiment of the intelligent traffic data processing device and method based on the neural network disclosed by the embodiment of the invention is based on the same inventive concept, and is not repeated herein.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the computer program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (2)
1. An intelligent traffic data processing method based on a neural network is characterized in that,
the method comprises the following steps:
responding to an intelligent traffic data acquisition instruction, triggering a camera to acquire image data, wherein a shooting area of the image data forms a target area;
carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result;
obtaining corresponding pedestrian position data according to each pedestrian in the pedestrian identification result;
acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment, and acquiring each indoor positioning data in the indoor positioning results;
uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space;
shooting areas of different cameras do not overlap, and a union set of the shooting areas of the cameras is the target area;
the neural network is a preset pedestrian recognition model, and the training method of the pedestrian recognition model comprises the following steps:
acquiring an image sample set, wherein each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing a region with a pedestrian from a region without the pedestrian, and the labeling result of the image sample represents the pedestrian in the image sample;
inputting the image sample into a feature extraction network to obtain a feature extraction result; inputting the feature extraction result into an image segmentation network to obtain an image segmentation result, obtaining image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into a target identification network to obtain a pedestrian identification result; obtaining image identification loss according to the pedestrian identification result and the labeling result, and adjusting parameters of the feature extraction network and the target identification network according to the image identification loss until the preset convergence condition is reached;
the method further comprises the step of determining a time interval formed by the first N minutes and the last N minutes of the intelligent traffic data acquisition instruction acquisition time as the time segment;
the evenly storing the pedestrian position data and the indoor positioning data in a cloud storage space comprises:
obtaining effective position data of the target area, wherein the effective position data comprises the pedestrian position data and the indoor positioning data which fall into the target area;
carrying out grid division on the target area, and numbering the divided grids according to a preset sequence;
according to the relation between the effective position data in the target area and the grids, carrying out grid growth according to the grids to obtain grid growth results;
storing the effective position data on cloud nodes of a cloud storage space according to a grid growth result;
the grid growing according to the grid to obtain a grid growing result comprises the following steps: randomly distributing a preset number of growing seeds in the target area, and growing the growing seeds in the target area, wherein the growing process of the growing seeds is a process of covering grids in a growing interval corresponding to the growing seeds, until the total number difference of the effective position data in each growing interval is smaller than a preset difference threshold after the growing is finished.
2. An intelligent traffic data processing device based on a neural network is characterized in that,
the method comprises the following steps:
the intelligent traffic data acquisition module is used for responding to an intelligent traffic data acquisition instruction and triggering a camera to acquire image data, and a shooting area of the image data forms a target area;
the pedestrian recognition module is used for carrying out pedestrian recognition on the image data based on a neural network to obtain a pedestrian recognition result;
the pedestrian position data acquisition module is used for acquiring corresponding pedestrian position data according to each pedestrian in the pedestrian identification result;
the indoor positioning data acquisition module is used for acquiring a time segment corresponding to the intelligent traffic data acquisition instruction, extracting indoor positioning results in the time segment and acquiring each indoor positioning data in the indoor positioning results;
the storage module is used for uniformly storing the pedestrian position data and the indoor positioning data in a cloud storage space;
further comprising: the system comprises an image sample set acquisition module, a pedestrian detection module and a pedestrian detection module, wherein the image sample set acquisition module is used for acquiring an image sample set, each image sample in the image sample set corresponds to a segmentation mask, the segmentation masks are used for distinguishing a region with a pedestrian from a region without the pedestrian, and the annotation result of the image sample represents the pedestrian in the image sample;
the first training module is used for inputting the image sample into a feature extraction network to obtain a feature extraction result; inputting the feature extraction result into an image segmentation network to obtain an image segmentation result, obtaining image segmentation loss according to the difference between the image segmentation result and the segmentation mask, and adjusting parameters of the feature extraction network and the image segmentation network according to the image segmentation loss feedback until a preset convergence condition is reached; fixing parameters of the image segmentation network;
the second training module is used for inputting the image sample into the feature extraction network to obtain a feature extraction result; inputting the feature extraction result into the image segmentation network to obtain an image segmentation result; inputting the image segmentation result into a target identification network to obtain a pedestrian identification result; obtaining image identification loss according to the pedestrian identification result and the labeling result, and adjusting parameters of the feature extraction network and the target identification network according to the image identification loss until the preset convergence condition is reached;
the memory module includes:
a valid position data determination unit, configured to acquire valid position data of the target area, where the valid position data includes the pedestrian position data and the indoor positioning data that fall within the target area;
the numbering unit is used for carrying out grid division on the target area and numbering the divided grids according to a preset sequence;
the growth unit is used for carrying out grid growth according to the grid according to the relation between the effective position data in the target area and the grid to obtain a grid growth result;
the storage unit is used for storing the effective position data on cloud nodes of a cloud storage space according to a grid growth result;
the grid growing according to the grid to obtain a grid growing result comprises the following steps: randomly distributing a preset number of growing seeds in the target area, and growing the growing seeds in the target area, wherein the growing process of the growing seeds is a process of covering grids in a growing interval corresponding to the growing seeds, until the total number difference of the effective position data in each growing interval is smaller than a preset difference threshold after the growing is finished.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110227091.2A CN112949486B (en) | 2021-03-01 | 2021-03-01 | Intelligent traffic data processing method and device based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110227091.2A CN112949486B (en) | 2021-03-01 | 2021-03-01 | Intelligent traffic data processing method and device based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112949486A CN112949486A (en) | 2021-06-11 |
| CN112949486B true CN112949486B (en) | 2022-05-17 |
Family
ID=76247042
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110227091.2A Active CN112949486B (en) | 2021-03-01 | 2021-03-01 | Intelligent traffic data processing method and device based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112949486B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111428653A (en) * | 2020-03-27 | 2020-07-17 | 湘潭大学 | Pedestrian congestion state determination method, device, server and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108399361A (en) * | 2018-01-23 | 2018-08-14 | 南京邮电大学 | A kind of pedestrian detection method based on convolutional neural networks CNN and semantic segmentation |
| CN108830205B (en) * | 2018-06-04 | 2019-06-14 | 江南大学 | Based on the multiple dimensioned perception pedestrian detection method for improving full convolutional network |
| US10410352B1 (en) * | 2019-01-25 | 2019-09-10 | StradVision, Inc. | Learning method and learning device for improving segmentation performance to be used for detecting events including pedestrian event, vehicle event, falling event and fallen event using edge loss and test method and test device using the same |
| CN109961009B (en) * | 2019-02-15 | 2023-10-31 | 平安科技（深圳）有限公司 | Pedestrian detection method, system, device and storage medium based on deep learning |
| CN109886245A (en) * | 2019-03-02 | 2019-06-14 | 山东大学 | A kind of pedestrian detection recognition methods based on deep learning cascade neural network |
| CN110188760B (en) * | 2019-04-01 | 2021-10-22 | 上海卫莎网络科技有限公司 | Image processing model training method, image processing method and electronic equipment |
| CN111582032A (en) * | 2020-04-03 | 2020-08-25 | 普联技术有限公司 | Pedestrian detection method and device, terminal equipment and storage medium |
| CN112396658B (en) * | 2020-11-30 | 2024-03-19 | 同济人工智能研究院(苏州)有限公司 | Indoor personnel positioning method and system based on video |
-
2021
- 2021-03-01 CN CN202110227091.2A patent/CN112949486B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111428653A (en) * | 2020-03-27 | 2020-07-17 | 湘潭大学 | Pedestrian congestion state determination method, device, server and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112949486A (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107547633B (en) | User constant standing point processing method and device and storage medium | |
| CN106790468B (en) | Distributed implementation method for analyzing WiFi (Wireless Fidelity) activity track rule of user | |
| CN111756848B (en) | QoS optimization method based on federal learning and mobile perception under mobile edge environment | |
| US9305110B2 (en) | Method and arrangement for supporting analysis of social networks in a communication network | |
| CN109691193A (en) | It is analyzed for mobile device cellular marker and mobile device network identifier to be carried out matched telecommunication network signal | |
| CN106982413B (en) | Regional passenger flow trend prediction method and device | |
| CN106055630A (en) | Log storage method and device | |
| CN109741227B (en) | Processing method and system for predicting human-room consistency based on nearest neighbor algorithm | |
| CN108934016A (en) | Division methods, device, computer equipment and the storage medium of cell scenario classification | |
| CN109104731A (en) | Construction method, device and the computer equipment of cell scenario category classification model | |
| CN110287219A (en) | A kind of data processing method and system | |
| CN110326323B (en) | Method and device for acquiring emission probability, transition probability and sequence positioning | |
| CN110662212B (en) | Mobile phone terminal identification method based on big data | |
| CN111163482B (en) | Data processing method, device and storage medium | |
| CN109241193A (en) | The treating method and apparatus and server cluster of distributed data base | |
| CN112949486B (en) | Intelligent traffic data processing method and device based on neural network | |
| CN104182348B (en) | Software test method and device | |
| CN111177481A (en) | User identifier mapping method and device | |
| Botella et al. | Overview of GeoLifeCLEF 2018: location-based species recommendation | |
| CN112949485B (en) | Data processing method and device based on artificial intelligence | |
| CN111368862A (en) | Method for distinguishing indoor and outdoor marks, training method and device of classifier and medium | |
| CN112613530A (en) | Cell resident identification method and system based on adaptive density clustering algorithm | |
| CN114745275A (en) | Node updating method and device in cloud service environment and computer equipment | |
| CN114078269A (en) | Face image clustering method, device, server and storage medium | |
| CN105684492B (en) | For assigning the method and apparatus of cell ID identifier value and the method and apparatus of the assignment for managing cell ID identifier value in communication network |
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 |