CN111767904B - Traffic incident detection method, device, terminal and storage medium - Google Patents
Traffic incident detection method, device, terminal and storage medium Download PDFInfo
- Publication number
- CN111767904B CN111767904B CN202010903369.9A CN202010903369A CN111767904B CN 111767904 B CN111767904 B CN 111767904B CN 202010903369 A CN202010903369 A CN 202010903369A CN 111767904 B CN111767904 B CN 111767904B
- Authority
- CN
- China
- Prior art keywords
- image
- detected
- detection result
- resident object
- dimensional
- 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
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention discloses a traffic incident detection method, a device, a terminal and a storage medium, wherein the traffic incident detection method comprises the following steps: acquiring an image to be detected; identifying non-stationary objects in the image to be detected; performing three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image; performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image; and detecting a traffic incident based on the first image and the second image to obtain a traffic incident detection result. The invention can realize the rapid detection of traffic events.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a traffic incident detection method, a traffic incident detection device, a traffic incident detection terminal and a storage medium.
Background
With the rapid development of the economic level of China, the living standard of residents is continuously improved, and the number of transportation means in China is continuously increased. The rapidly increasing traffic demand and the high-maintenance vehicle operation bring a series of problems such as traffic accidents, traffic jams, etc., and thus perfect traffic safety management is required to solve such traffic event problems. The traffic event refers to an accidental event on a road, and comprises vehicle collision, illegal parking, falling objects, traffic jam and the like.
When a traffic incident occurs, the traffic incident needs to be quickly processed to reduce traffic delay and reduce the probability of occurrence of secondary accidents, so that the requirement of traffic incident detection on timeliness is high, the existing traffic incident detection method is difficult to quickly detect the traffic incident, and quick response is difficult to realize after the traffic incident occurs.
Disclosure of Invention
The invention solves the problem that the existing traffic incident detection method is difficult to quickly detect the traffic incident.
In order to solve the above problems, the present invention provides a traffic incident detection method, including:
acquiring an image to be detected; identifying non-stationary objects in the image to be detected; performing three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image; performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image; and detecting a traffic incident based on the first image and the second image to obtain a traffic incident detection result.
The method has the advantages that the non-stationary object is constructed into the simple and abstract three-dimensional graph and two-dimensional graph through three-dimensional modeling and two-dimensional modeling, the construction efficiency is high, after the construction is completed, the non-stationary object in the image to be detected is replaced into the simple and abstract three-dimensional graph and two-dimensional graph, the requirement on the resolution ratio of the image can be lowered, high-precision event identification can be realized under the conditions of low visibility such as rain and fog weather, the anti-interference performance is high, in addition, the replacement simplifies the complexity of image processing, the detection efficiency of the traffic event can be improved, the response speed of the traffic event is further improved, the traffic delay is reduced, and the occurrence probability of secondary accidents is reduced.
Optionally, the three-dimensional modeling of the non-resident object in the image to be detected to obtain a first image includes: performing cuboid fitting on the basis of each non-resident object to generate a cuboid model corresponding to each non-resident object; and replacing each non-resident object in the image to be detected by using a cuboid model corresponding to each non-resident object to obtain the first image.
The non-stationary object in the image to be detected is replaced by the corresponding cuboid model, so that the complexity of subsequent feature extraction and data processing can be simplified, the calculated amount is reduced, and the traffic incident detection efficiency is improved.
Optionally, the two-dimensional modeling of the non-resident object in the image to be detected, and obtaining the second image includes: constructing a circumscribed rectangular frame of each non-resident object; and replacing each non-resident object in the image to be detected by using the circumscribed rectangular frame of each non-resident object to obtain a second image.
The external rectangular frame is high in construction efficiency, the non-stationary object in the image to be detected is replaced by the corresponding external rectangular frame, the complexity of subsequent feature extraction and data processing can be simplified, the calculated amount is reduced, the rapid identification capability is realized, the reliability of subsequent image processing steps is improved, and the traffic incident detection efficiency is improved.
Optionally, the performing traffic event detection based on the first image and the second image, and obtaining a traffic event detection result includes:
inputting the first image into a preset three-dimensional detection model to obtain a three-dimensional detection result output by the three-dimensional detection model; inputting the second image into a preset two-dimensional detection model to obtain a two-dimensional detection result output by the two-dimensional detection model; determining the traffic event detection result based on the three-dimensional detection result and the two-dimensional detection result; and the three-dimensional detection model and the two-dimensional detection model are both convolution neural network models.
Optionally, the performing traffic event detection based on the first image and the second image, and obtaining a traffic event detection result includes:
performing traffic incident detection based on the first image and the second image respectively to correspondingly obtain a first detection result and a second detection result; and when the first detection result is consistent with the second detection result, taking the first detection result or the second detection result as the traffic event detection result. The error rate of traffic incident identification can be reduced, and the accuracy of the detection result is improved.
Optionally, after the traffic event detection is performed based on the first image and the second image respectively and the first detection result and the second detection result are obtained correspondingly, the traffic event detection method further includes:
and when the first detection result is inconsistent with the second detection result, storing the first detection result and the first image in a correlated manner, and storing the second detection result and the second image in a correlated manner. And the optimization of the traffic incident detection algorithm is convenient to carry out subsequently.
Optionally, the acquiring the image to be detected includes: and acquiring the image to be detected from the road monitoring video based on a preset time interval.
The road monitoring video can be directly obtained on the basis of the existing monitoring equipment, the image to be detected is obtained from the road monitoring video, the old equipment can be fully utilized, the workload of infrastructure construction is reduced, in addition, the image to be detected is extracted in a preset time interval mode, the image extraction operation is simple, the requirements on a processor and a memory are low, the cost is low, and the method is easy to realize, popularize and apply.
The invention also proposes a traffic incident detection device comprising:
the image acquisition module is used for acquiring an image to be detected; the image identification module is used for identifying the non-stationary object in the image to be detected; the modeling module is used for carrying out three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image; performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image; a detection processing module for traffic event detection based on the first image and the second image.
Compared with the prior art, the traffic incident detection device of the invention has the advantages that the beneficial effects are consistent with the traffic incident detection method, and the details are not repeated herein.
The invention also provides a traffic incident detection terminal, which comprises a computer readable storage medium and a processor, wherein the computer readable storage medium stores a computer program, and the computer program is read by the processor and runs to realize the traffic incident detection method.
Compared with the prior art, the traffic incident detection terminal has the beneficial effects consistent with the traffic incident detection method, and is not repeated herein.
The invention also proposes a computer-readable storage medium, in which a computer program is stored, which, when read and executed by a processor, implements a traffic event detection method as defined in any one of the above.
The beneficial effects of the computer readable storage medium of the present invention over the prior art are consistent with the above-mentioned traffic incident detection method, and are not described herein again.
Drawings
FIG. 1 is a schematic diagram of a traffic incident detection method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of another embodiment of the traffic event detection method of the present invention;
FIG. 3 is a schematic diagram of an embodiment of the traffic event detection method of the present invention after step S50 is detailed;
FIG. 4 is a schematic diagram of another embodiment of the traffic event detection method of the present invention after step S50 is refined;
FIG. 5 is a schematic view of a traffic incident detection apparatus according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a traffic incident detection terminal according to an embodiment of the invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
The background in which the present invention is made will first be briefly described. Existing event detection algorithms are largely divided into direct detection methods and indirect detection methods. The direct detection method is a method for detecting abnormal driving behaviors of vehicles and extracting traffic event information by using digital image processing and video recognition technologies. The indirect detection method analyzes the traffic parameters (such as vehicle speed, vehicle flow and road occupancy) acquired by the road detector, and indirectly identifies the traffic events by pattern recognition or statistical analysis. The former needs to perform complex feature extraction on images or needs to perform continuous calculation and identification, the requirements on image definition, resolution and the like are high, the requirement on computing power is also high, and the latter needs to be provided with hardware detectors, so that the cost is high, the continuous calculation and identification are needed, and the requirement on computing power is also high.
The invention provides a traffic incident detection method.
Fig. 1 is a schematic diagram of a traffic incident detection method according to an embodiment of the present invention. As shown in fig. 1, the traffic event detection method includes:
step S10, acquiring an image to be detected;
the image to be detected refers to basic image data for detecting traffic incidents. The image to be detected can be obtained from a preset database and can also be extracted from a road monitoring video. The image to be detected can be one or more frames of images.
Optionally, step S10 includes: and acquiring the image to be detected from the road monitoring video based on a preset time interval.
Wherein, the value range of the preset time interval is as follows: greater than 0 and less than or equal to 10 seconds.
The preset time interval may be selected as a fixed time interval, the image to be detected may be one or more frames of images, for example, the preset time interval is 3 seconds, and step S10 is: and extracting 1 frame of image from the road monitoring video every 3 seconds to be used as an image to be detected.
The preset time interval may be selected as a plurality of different time intervals. In an embodiment, a smaller time interval is selected in a traffic flow peak period, a larger time interval is selected in a traffic flow valley period, a preset first time period may be set as the traffic flow peak period, a preset second time period may be set as the traffic flow valley period, for example, 7:00 pm to 9:00 pm and 17:30 pm to 18:30 pm are set as the traffic flow peak period, and 23:00 pm to 6:00 pm are set as the traffic flow valley period, and the traffic flow peak period and the traffic flow valley period may also be distinguished by identifying the traffic flow density, for example, when the traffic flow density is greater than a preset density, it is determined that the traffic flow peak period is currently, and when the traffic flow density is less than the preset density, it is determined that the traffic flow valley period is currently. Because the possibility of traffic events such as collision and the like in the rush hour of the traffic flow is higher, the time interval for extracting the video from the road monitoring video is shorter, and the traffic events such as collision and the like can be found in time, so that the traffic events can be rapidly processed, the traffic delay is reduced, and the probability of secondary traffic accidents is reduced; because the possibility of traffic events such as collision and the like in the valley period of the traffic flow is slightly low, the time interval for extracting the video from the road monitoring video is larger, the resource consumption can be reduced, and meanwhile, the detection efficiency of the traffic events cannot be greatly influenced.
Alternatively, the preset time interval may be set/modified by a person based on actual road conditions to accommodate diverse road traffic conditions.
The road monitoring video can be directly obtained on the basis of the existing monitoring equipment, the image to be detected is obtained from the road monitoring video, the old equipment can be fully utilized, the workload of infrastructure construction is reduced, in addition, the image to be detected is extracted in a preset time interval mode, the image extraction operation is simple, the requirements on a processor and a memory are low, the cost is low, and the method is easy to realize, popularize and apply.
Step S20, identifying a non-stationary object in the image to be detected;
the non-stationary object refers to a mobile object such as a vehicle, a person, a falling object and the like, and the relative resident object comprises: roads, trees, roadside lamp posts, sky and other background objects in fixed positions. For a camera with a fixed position and a fixed shooting angle, the shooting area range is fixed, and a resident object is fixed.
In one embodiment, an initial image is obtained, the initial image and an image to be detected are converted into a gray scale image, a difference value operation is performed on the gray scale image of the initial image and the gray scale image of the image to be detected to obtain a difference value image, and binarization processing is performed on the difference value image: and assigning the pixel with the difference absolute value larger than the preset value as 1, and assigning the pixel with the difference absolute value smaller than or equal to the preset value as 0, wherein the pixel with the value of 1 belongs to the non-resident object, and the pixel with the value of 0 belongs to the resident object.
In another embodiment, since the camera is fixed in position and the range of the shooting area is fixed, the road in the lane and beside the lane is fixed in position in the image shot by the camera, and the non-stationary object usually moves in the lane and beside the road, so that the moving object in the lane and in the road area beside the lane can be detected and can be used as the non-stationary object.
Step S30, performing three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image;
three-dimensional modeling herein refers to the abstraction of individual non-stationary objects into a three-dimensional solid figure that is used to characterize the volume and location of the non-stationary objects. Alternatively, the nonresident object is abstracted as a cuboid or cylinder.
Alternatively, as shown in fig. 2, the step S30 includes:
step S300, performing cuboid fitting on the basis of each non-resident object to generate a cuboid model corresponding to each non-resident object;
for any non-stationary object, if the non-stationary object is identified to be a vehicle, firstly identifying the vehicle type of the non-stationary object based on the image to be detected, then determining the length, width and height data of the non-stationary object based on the corresponding relation between the pre-stored vehicle type and the length, width and height data, finally performing cuboid fitting based on the length, width and height data of the non-stationary object, and taking the well-fitted cuboid as the three-dimensional modeling result of the target vehicle. The database can be preset, and various vehicle types and corresponding length, width and height data can be stored in the database.
For any non-stationary object, if the non-stationary object is identified to be a human or other non-vehicle object, the contour boundary of the non-stationary object can be obtained, the length, the width and the height are determined based on the contour boundary of the non-stationary object, cuboid fitting is performed, the height and the width of the non-stationary object in an image to be detected can be specifically obtained to serve as the length and the width of a cuboid, and a fixed preset value can be obtained to serve as the height. A cuboid with fixed length, width and height can also be used as a cuboid model corresponding to all people.
Step S301, replacing each non-resident object in the image to be detected with a rectangular parallelepiped model corresponding to each non-resident object, to obtain the first image.
And replacing the non-resident object, namely reserving the cuboid model corresponding to the non-resident object and removing the non-resident object.
The non-stationary object in the image to be detected is replaced by the corresponding cuboid model, so that the complexity of subsequent feature extraction and data processing can be simplified, the calculated amount is reduced, and the traffic incident detection efficiency is improved.
Optionally, in step S30, after or before the three-dimensional modeling, the method further includes: and eliminating the resident object, wherein the first image is the image with the eliminated resident object. Specifically, the resident object may be removed before step S300, or after step S300 and before step S301, specifically, after the resident object is recognized, all the pixel gradations corresponding to the resident object are set to 0, and the resident object may be removed.
In one embodiment, before the permanent object is removed, a road edge line is identified, wherein the road edge line refers to a boundary line between a lane and a roadside pedestrian road or between the lane and a roadside guardrail, the road edge line is strengthened, the permanent object except the road edge line is removed, and the situation that a vehicle drives to the pedestrian road or the vehicle collides with the guardrail is identified in the subsequent detection.
Step S40, performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image;
two-dimensional modeling herein refers to abstracting each nonresident object into a quadrilateral to represent the location of the nonresident object. Alternatively, the nonresident object is abstracted as a rectangle or square.
Alternatively, as shown in fig. 2, the step S40 includes:
step S400, constructing an external rectangular frame of each non-resident object;
determining the height and width of the non-stationary object in the image to be detected based on the pixel distribution of the non-stationary object in the image to be detected, and taking the height and width of the non-stationary object in the image to be detected as the length and width of the circumscribed rectangular frame.
Step S401, replacing each non-resident object in the image to be detected with an external rectangular frame of each non-resident object, to obtain a second image.
And replacing the non-resident object, namely reserving a circumscribed rectangular frame corresponding to the non-resident object and removing the non-resident object.
The external rectangular frame is high in construction efficiency, the non-stationary object in the image to be detected is replaced by the corresponding external rectangular frame, the complexity of subsequent feature extraction and data processing can be simplified, the calculated amount is reduced, the rapid identification capability is realized, the reliability of subsequent image processing steps is improved, and the traffic incident detection efficiency is improved.
Optionally, in step S40, after the two-dimensional modeling, or before the two-dimensional modeling, the method further includes: and eliminating the resident object, wherein the second image is the image with the eliminated resident object. Specifically, the resident objects may be removed before step S400, or after step S400 and before step S401. In one embodiment, before the permanent object is removed, a road edge line is identified, wherein the road edge line refers to a boundary line between a lane and a roadside pedestrian road or between the lane and a roadside guardrail, the road edge line is strengthened, the permanent object except the road edge line is removed, and the situation that a vehicle drives to the pedestrian road or the vehicle collides with the guardrail is identified in the subsequent detection.
Step S30 and step S40 may be executed simultaneously or sequentially, and the execution order of both is not limited.
And step S50, detecting traffic incidents based on the first image and the second image, and obtaining a traffic incident detection result.
In one embodiment, the traffic event is identified based on the first image and the second image, and if the traffic events identified based on the first image and the second image are consistent, the traffic events identified by the first image and the second image are the final traffic event detection result, for example, if a collision event is identified based on both the first image and the second image, the collision event is the traffic event detection result; and if no traffic event is identified on the basis of the first image and the second image, the traffic event which does not occur is the traffic event detection result.
Optionally, step S50 is followed by: when the traffic incident detection result is a specific traffic incident, outputting the traffic incident, and returning to execute the steps S10-S50; and when the traffic event detection result is that the traffic event is not detected, no output response is made, and the step S10-step S50 are executed.
The explanation will be given taking the traffic event detection of the first image as an example. In one embodiment, the traffic event includes a collision and an abnormal stay, and first, whether the collision occurs is detected based on the first image, if the collision is detected, the collision is used as a detection result of the first image, if the collision does not occur, whether the abnormal stay occurs is judged based on the first images adjacent to the multiple frames, if the abnormal stay occurs, the abnormal stay is used as a detection result of the first image, and if the abnormal stay does not occur, the traffic event which does not occur is used as a detection result of the first image. The explanation of the second image is similar to the first image and is not repeated here.
In the embodiment of the invention, the non-stationary object is constructed into the simple and abstract three-dimensional graph and two-dimensional graph through three-dimensional modeling and two-dimensional modeling, the construction efficiency is higher, and after the construction is finished, the non-stationary object in the image to be detected is replaced into the simple and abstract three-dimensional graph and two-dimensional graph, so that the requirement on the resolution ratio of the image can be reduced, and therefore, high-precision event identification can be realized under the condition of low visibility such as rain and fog weather, and the like, the anti-interference performance is strong, in addition, the replacement simplifies the complexity of image processing, the detection efficiency of the traffic event can be improved, the response speed of the traffic event is further improved, the traffic delay is reduced, and the occurrence probability of.
Alternatively, as shown in fig. 3, the step S50 includes:
step S500, traffic incident detection is carried out respectively on the basis of the first image and the second image, and a first detection result and a second detection result are correspondingly obtained;
the first detection result is a detection result obtained based on the first image, and the second detection result is a detection result obtained based on the second image.
Judging whether the first detection result is consistent with the second detection result;
step S501, when the first detection result is consistent with the second detection result, the first detection result or the second detection result is used as the traffic incident detection result.
If the traffic events identified based on the first image and the second image are consistent, the traffic events identified by the first image and the second image are the final traffic event detection result, for example, if a collision event is identified based on both the first image and the second image, the collision event is the traffic event detection result; and if no traffic event is identified on the basis of the first image and the second image, the traffic event which does not occur is the traffic event detection result.
Consistency check is carried out on a first detection result obtained based on a first image of three-dimensional modeling and a second detection result obtained based on a second image of two-dimensional modeling to determine a final traffic incident detection result, and when the first detection result is consistent with the second detection result, the first detection result and the second detection result are used as the traffic incident detection result, so that the error rate of traffic incident identification can be reduced, and the accuracy of the detection result is improved.
When the first detection result and the second detection result are not consistent, the traffic event detection result is empty, and the steps S10 to S50 can be directly returned to. Optionally, while returning to perform steps S10-S50, the first detection result and the second detection result are respectively output to remind people to check, so as to avoid omission of the traffic incident and facilitate timely finding of the traffic incident.
Optionally, when the first detection result and the second detection result are inconsistent, the first detection result and the first image, and the second detection result and the second image are stored in a correlated manner, and the stored first detection result and the first image are used as basic data for optimizing a traffic event detection algorithm in a subsequent process. Specifically, the method comprises the steps of manually rechecking, determining an error detection result from a first detection result and a second detection result, marking a traffic event (traffic event type or no traffic event) on an image (a first image/a second image) corresponding to the error detection result, and retraining the image marked by the traffic event as a training sample of a corresponding traffic event detection model to improve the accuracy of the traffic event detection model, wherein the traffic event modeling model is a convolutional neural network model.
Alternatively, as shown in fig. 4, the step S50 includes:
step S510, inputting the first image into a preset three-dimensional detection model to obtain a three-dimensional detection result output by the three-dimensional detection model;
the three-dimensional detection result is a detection result obtained based on the first image.
Step S511, inputting the second image into a preset two-dimensional detection model to obtain a two-dimensional detection result output by the two-dimensional detection model;
the two-dimensional detection result is a detection result obtained based on the second image.
Step S512, determining the traffic incident detection result based on the three-dimensional detection result and the two-dimensional detection result; and the three-dimensional detection model and the two-dimensional detection model are both convolution neural network models.
The training process of the three-dimensional detection model and the two-dimensional detection model is also included before step S10, and the following explanation takes the three-dimensional detection model as an example, and the training process of the two-dimensional detection model is similar to that, and is not repeated here. Specifically, three-dimensional detection model parameters are initialized; obtaining a training sample, training the three-dimensional detection model based on the training sample to obtain an optimal model parameter, and taking the three-dimensional detection model with the optimal model parameter as the three-dimensional detection model preset in step S510, wherein the training sample is an image set labeled with various traffic events or labeled without traffic events, the image used for training the three-dimensional detection model in the training sample is subjected to three-dimensional modeling (model replacement) of a non-stationary object, and the image used for training the two-dimensional detection model is subjected to two-dimensional modeling (model replacement) of the non-stationary object.
The traffic incident detection is carried out through the convolutional neural network model, the accuracy of traffic incident identification can be improved, and the workload of manual rechecking can be reduced to a greater extent.
The invention further provides a traffic incident detection device. Fig. 5 is a schematic view of an embodiment of the traffic incident detection apparatus according to the present invention, as shown in fig. 5, the traffic incident detection apparatus includes:
the image acquisition module is used for acquiring an image to be detected;
the image identification module is used for identifying the non-stationary object in the image to be detected;
the modeling module is used for carrying out three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image; performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image;
a detection processing module for traffic event detection based on the first image and the second image.
Optionally, the modeling module is further configured to perform cuboid fitting based on each of the non-resident objects, and generate a cuboid model corresponding to each of the non-resident objects; and replacing each non-resident object in the image to be detected by using a cuboid model corresponding to each non-resident object to obtain the first image.
Optionally, the modeling module is further configured to construct a bounding rectangular border of each of the non-resident objects; and replacing each non-resident object in the image to be detected by using the circumscribed rectangular frame of each non-resident object to obtain a second image.
Optionally, the detection processing module is further configured to input the first image into a preset three-dimensional detection model, and obtain a three-dimensional detection result output by the three-dimensional detection model; inputting the second image into a preset two-dimensional detection model to obtain a two-dimensional detection result output by the two-dimensional detection model; determining the traffic event detection result based on the three-dimensional detection result and the two-dimensional detection result; and the three-dimensional detection model and the two-dimensional detection model are both convolution neural network models.
Optionally, the detection processing module is further configured to perform traffic event detection based on the first image and the second image, and correspondingly obtain a first detection result and a second detection result; and when the first detection result is consistent with the second detection result, taking the first detection result or the second detection result as the traffic event detection result.
Optionally, the traffic event detection device further includes a storage module, configured to store the first detection result in association with the first image and the second detection result in association with the second image when the first detection result and the second detection result are inconsistent.
Optionally, the obtaining module is further configured to obtain the image to be detected from the road monitoring video based on a preset time interval.
Compared with the prior art, the traffic incident detection device of the invention has the advantages that the beneficial effects are consistent with the traffic incident detection method, and the details are not repeated herein.
The present invention further provides a traffic event detecting terminal, as shown in fig. 6, where the traffic event detecting terminal includes a computer readable storage medium storing a computer program and a processor, and the computer program is read by the processor and executed by the processor to implement the traffic event detecting method according to any of the above.
Compared with the prior art, the traffic incident detection terminal has the beneficial effects consistent with the traffic incident detection method, and is not repeated herein.
The invention also provides a computer-readable storage medium, which is characterized by storing a computer program, wherein the computer program is read and executed by a processor to implement the traffic event detection method according to any one of the above.
The beneficial effects of the computer readable storage medium of the present invention over the prior art are consistent with the above-mentioned traffic incident detection method, and are not described herein again.
Although the present disclosure has been described above, the scope of the present disclosure is not limited thereto. Various changes and modifications may be effected therein by one of ordinary skill in the pertinent art without departing from the spirit and scope of the present disclosure, and these changes and modifications are intended to be within the scope of the present disclosure.
Claims (6)
1. A traffic event detection method, comprising:
acquiring an image to be detected, wherein the acquiring of the image to be detected comprises: acquiring the image to be detected from a road monitoring video based on a preset time interval;
identifying non-stationary objects in the image to be detected, wherein the non-stationary objects comprise vehicles and people;
three-dimensional modeling is carried out on the non-resident object in the image to be detected to obtain a first image, and the method specifically comprises the following steps: abstracting each non-resident object into a three-dimensional stereo graph, and replacing each non-resident object in the image to be detected by using the three-dimensional stereo graph corresponding to each non-resident object to obtain the first image; abstracting each non-resident object into a three-dimensional stereo graph, and replacing each non-resident object in the image to be detected with the three-dimensional stereo graph corresponding to each non-resident object to obtain the first image, wherein the three-dimensional stereo graph comprises: performing cuboid fitting on the basis of each non-resident object to generate a cuboid model corresponding to each non-resident object; replacing each non-resident object in the image to be detected by using a cuboid model corresponding to each non-resident object to obtain a first image; if the non-stationary object is a person, acquiring the height and the width of the non-stationary object in the image to be detected as the length and the width of the cuboid, acquiring a fixed preset value as the height, and performing cuboid fitting;
performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image, and specifically comprising the following steps: abstracting each non-resident object into a quadrangle, and replacing each non-resident object in the image to be detected by using the quadrangle corresponding to each non-resident object to obtain a second image; abstracting each non-resident object into a quadrangle, replacing each non-resident object in the image to be detected with the quadrangle corresponding to each non-resident object, and obtaining the second image comprises: constructing a circumscribed rectangular frame of each non-resident object; replacing each non-resident object in the image to be detected by using the circumscribed rectangular frame of each non-resident object to obtain a second image; the constructing of the circumscribed rectangular border of each of the non-resident objects comprises: determining the height and width of each non-resident object in the image to be detected based on the pixel distribution of each non-resident object in the image to be detected, and taking the height and width of each non-resident object in the image to be detected as the length and width of the circumscribed rectangular frame;
performing traffic incident detection based on the first image and the second image to obtain a traffic incident detection result, specifically including: performing traffic incident detection based on the first image and the second image respectively to correspondingly obtain a first detection result and a second detection result; and when the first detection result is consistent with the second detection result, taking the first detection result or the second detection result as the traffic event detection result.
2. The traffic event detection method of claim 1, wherein performing traffic event detection based on the first image and the second image, and obtaining a traffic event detection result comprises:
inputting the first image into a preset three-dimensional detection model to obtain a three-dimensional detection result output by the three-dimensional detection model;
inputting the second image into a preset two-dimensional detection model to obtain a two-dimensional detection result output by the two-dimensional detection model;
determining the traffic event detection result based on the three-dimensional detection result and the two-dimensional detection result; and the three-dimensional detection model and the two-dimensional detection model are both convolution neural network models.
3. The traffic event detection method of claim 1, wherein the traffic event detection is performed based on the first image and the second image, respectively, and after obtaining the first detection result and the second detection result, the traffic event detection method further comprises:
and when the first detection result is inconsistent with the second detection result, storing the first detection result and the first image in a correlated manner, and storing the second detection result and the second image in a correlated manner.
4. A traffic event detection device, comprising:
the system comprises an image acquisition module, a data processing module and a data processing module, wherein the image acquisition module is used for acquiring an image to be detected, and is specifically used for acquiring the image to be detected from a road monitoring video based on a preset time interval;
an image identification module for identifying non-stationary objects in the image to be detected, wherein the non-stationary objects comprise vehicles and people;
the modeling module is used for performing three-dimensional modeling on the non-resident object in the image to be detected to obtain a first image, and specifically comprises: abstracting each non-resident object into a three-dimensional stereo graph, and replacing each non-resident object in the image to be detected with the three-dimensional stereo graph corresponding to each non-resident object to obtain the first image, which specifically comprises: performing cuboid fitting on the basis of each non-resident object to generate a cuboid model corresponding to each non-resident object; replacing each non-resident object in the image to be detected by using a cuboid model corresponding to each non-resident object to obtain a first image; if the non-stationary object is a person, acquiring the height and the width of the non-stationary object in the image to be detected as the length and the width of the cuboid, acquiring a fixed preset value as the height, and performing cuboid fitting;
performing two-dimensional modeling on the non-resident object in the image to be detected to obtain a second image, and specifically comprising the following steps: abstracting each non-resident object into a quadrangle, replacing each non-resident object in the image to be detected with the quadrangle corresponding to each non-resident object, and obtaining the second image, wherein the method specifically comprises the following steps: constructing a circumscribed rectangular frame of each non-resident object; replacing each non-resident object in the image to be detected by using the circumscribed rectangular frame of each non-resident object to obtain a second image; the constructing of the circumscribed rectangular border of each of the non-resident objects comprises: determining the height and width of each non-resident object in the image to be detected based on the pixel distribution of each non-resident object in the image to be detected, and taking the height and width of each non-resident object in the image to be detected as the length and width of the circumscribed rectangular frame;
the detection processing module is configured to perform traffic event detection based on the first image and the second image, and specifically includes: performing traffic incident detection based on the first image and the second image respectively to correspondingly obtain a first detection result and a second detection result; and when the first detection result is consistent with the second detection result, taking the first detection result or the second detection result as the traffic event detection result.
5. A traffic event detection terminal comprising a computer readable storage medium having a computer program stored thereon and a processor, the computer program when read and executed by the processor implementing the traffic event detection method according to any one of claims 1-3.
6. A computer-readable storage medium, characterized in that it stores a computer program which, when read and executed by a processor, implements the traffic event detection method according to any one of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010903369.9A CN111767904B (en) | 2020-09-01 | 2020-09-01 | Traffic incident detection method, device, terminal and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010903369.9A CN111767904B (en) | 2020-09-01 | 2020-09-01 | Traffic incident detection method, device, terminal and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111767904A CN111767904A (en) | 2020-10-13 |
| CN111767904B true CN111767904B (en) | 2021-02-02 |
Family
ID=72729615
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010903369.9A Active CN111767904B (en) | 2020-09-01 | 2020-09-01 | Traffic incident detection method, device, terminal and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111767904B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113343905B (en) * | 2021-06-28 | 2022-06-14 | 山东理工大学 | Method and system for training road abnormity intelligent recognition model and recognizing road abnormity |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104658036B (en) * | 2015-01-30 | 2018-06-15 | 郑州大学 | A kind of quick method for establishing intelligent transportation three dimensional field |
| CN105812733B (en) * | 2016-03-15 | 2017-12-15 | 中国民用航空总局第二研究所 | A kind of scene monitoring guiding system of air traffic control |
| CN105869109B (en) * | 2016-03-28 | 2018-12-07 | 长安大学 | Parking based on different height inverse projection face and leave object differentiating method |
-
2020
- 2020-09-01 CN CN202010903369.9A patent/CN111767904B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111767904A (en) | 2020-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210319561A1 (en) | Image segmentation method and system for pavement disease based on deep learning | |
| CN110188807B (en) | Tunnel pedestrian target detection method based on cascading super-resolution network and improved Faster R-CNN | |
| CN111368687B (en) | Sidewalk vehicle illegal parking detection method based on target detection and semantic segmentation | |
| US11455805B2 (en) | Method and apparatus for detecting parking space usage condition, electronic device, and storage medium | |
| KR102197946B1 (en) | object recognition and counting method using deep learning artificial intelligence technology | |
| KR102015947B1 (en) | Method for extracting image of learning object for autonomous driving and apparatus thereof | |
| CN109935080B (en) | Monitoring system and method for real-time calculation of traffic flow on traffic line | |
| WO2021139049A1 (en) | Detection method, detection apparatus, monitoring device, and computer readable storage medium | |
| KR102122859B1 (en) | Method for tracking multi target in traffic image-monitoring-system | |
| US20230222844A1 (en) | Parking lot management and control method based on object activity prediction, and electronic device | |
| CN112766069A (en) | Vehicle illegal parking detection method and device based on deep learning and electronic equipment | |
| CN113822285A (en) | Vehicle illegal parking identification method for complex application scene | |
| CN111813997A (en) | Intrusion analysis method, device, equipment and storage medium | |
| CN114332776B (en) | Non-motor vehicle occupant pedestrian lane detection method, system, device and storage medium | |
| CN110705370A (en) | Deep learning-based road condition identification method, device, equipment and storage medium | |
| CN111259736B (en) | Real-time pedestrian detection method based on deep learning in complex environment | |
| JP2019192209A (en) | Learning target image packaging device and method for artificial intelligence of video movie | |
| CN114049572A (en) | Detection method for identifying small target | |
| CN111767904B (en) | Traffic incident detection method, device, terminal and storage medium | |
| CN114926791A (en) | Method and device for detecting abnormal lane change of vehicles at intersection, storage medium and electronic equipment | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN113486885A (en) | License plate recognition method and device, electronic equipment and storage medium | |
| CN112906471A (en) | Traffic signal lamp identification method and device | |
| CN113505638A (en) | Traffic flow monitoring method, traffic flow monitoring device and computer-readable storage medium | |
| JP6681965B2 (en) | Apparatus and method for extracting learning target image for autonomous driving |
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 |