CN109948436A - The method and device of vehicle on a kind of monitoring road - Google Patents
The method and device of vehicle on a kind of monitoring road Download PDFInfo
- Publication number
- CN109948436A CN109948436A CN201910103337.8A CN201910103337A CN109948436A CN 109948436 A CN109948436 A CN 109948436A CN 201910103337 A CN201910103337 A CN 201910103337A CN 109948436 A CN109948436 A CN 109948436A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- image
- value
- kth
- road
- 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.)
- Granted
Links
Abstract
The invention discloses a kind of method and devices of vehicle on monitoring road, method includes: the image collection for obtaining the monitor video of vehicle on road, for+1 image of k-th of image and kth in image collection, according to deep learning detection model, the coordinate of the vehicle of+1 image of coordinate and kth of the vehicle of k-th of image is determined respectively, and the vehicle distances between the vehicle of+1 image of vehicle and kth of determining k-th of image, according to vehicle distances, determine the quantity of stationary vehicle in k-th of image, the quantity of stationary vehicle is greater than after zero in determining k-th of image, according to forgetting factor and the corresponding aggregate-value of -1 image of kth, determine the corresponding aggregate-value of k-th of image, when the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs abnormal on road.The technical solution improves detection efficiency under the premise of ensuring the detection accuracy to road abnormality detection.
Description
Technical field
The present embodiments relate to the methods and dress of vehicle in technical field of vehicle detection more particularly to a kind of monitoring road
It sets.
Background technique
Vehicle testing techniques based on video are a part important in traffic video monitoring system, can be monitoring system
System provides relevant information of vehicles.By the detection to vehicle location on road in monitor video, judge whether there is vehicle on road
Congestion or parking offense etc. are abnormal, are mainly identified more according to the multiple images intercepted in monitor video in the prior art
The coordinate information of the same vehicle and the same vehicle in a image judges the vehicle according to the coordinate information of the same vehicle
Whether remain static, and then judges whether the condition of road surface is abnormal.It is identified together for multiple images in the prior art
The difficulty of one vehicle is larger, and aforesaid way judges that the efficiency of condition of road surface exception is lower, is unfavorable for the height to road vehicle
Effect control.
Summary of the invention
The embodiment of the present invention provides a kind of method and device for monitoring vehicle on road, to ensure to examine road extremely
Under the premise of the detection accuracy of survey, detection efficiency is improved.
The method of vehicle on a kind of monitoring road provided in an embodiment of the present invention, comprising:
Obtain the image collection of the monitor video of vehicle on road；Image is to be spaced at preset timed intervals in described image set
Intercept what the monitor video obtained；
For+1 image of k-th of image and kth in described image set, according to deep learning detection model, respectively
Determine the coordinate of the vehicle of+1 image of coordinate and the kth of the vehicle of k-th of image；The k is more than or equal to 1；
The deep learning detection model is the seat to the vehicle of the vehicle and marked completion of the marked completion on each picture
What mark determined after being trained；
According to the coordinate of the vehicle of+1 image of the coordinate of the vehicle of k-th of image and the kth, described is determined
Vehicle distances between the vehicle of+1 image of vehicle and the kth of k image；
According to the vehicle distances, the quantity of stationary vehicle in k-th of image is determined；
The quantity of stationary vehicle is greater than after zero in determining k-th of image, is schemed according to forgetting factor and kth -1
As corresponding aggregate-value, the corresponding aggregate-value of k-th of image is determined；
When the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs abnormal on the road.
It is true respectively according to deep learning detection model for+1 image of k-th of image and kth in above-mentioned technical proposal
Make the coordinate of the vehicle of+1 image of coordinate and kth of the vehicle of k-th of image, and determine k-th of image vehicle and
Vehicle distances between the vehicle of+1 image of kth, and determine according to vehicle distances the number of stationary vehicle in k-th of image
Amount determines after the quantity of stationary vehicle in k-th of image is greater than zero according to the aggregate-value of -1 image of forgetting factor and kth
The aggregate-value of k-th of image out determines that vehicle goes out on road when the aggregate-value for determining k-th of image is greater than second threshold
It is now abnormal.The technical solution is not necessarily to be detected for the same vehicle, reduces detection difficulty, opposite to improve detection efficiency,
Using deep learning detection model, the detection accuracy of vehicle is improved.
Optionally, described according to the vehicle distances, determine the quantity of stationary vehicle in k-th of image, comprising:
Judge whether the vehicle distances are less than first threshold, if so, determining that the vehicle of k-th of image is static
Vehicle；
Count the quantity of stationary vehicle described in k-th of image.
In above-mentioned technical proposal, the vehicle of k-th of image can be it is multiple, the vehicle of+1 image of kth may be more
It is a, for any one vehicle in k-th of image, determine the vehicle of all vehicles in any one vehicle and+1 image of kth
As long as distance judges that the vehicle of k-th of image is stationary vehicle have vehicle distances to be less than first threshold.According to k-th
The vehicle-state of multiple vehicles in image counts the quantity of stationary vehicle in k-th of image.
Optionally, on the acquisition road before the image collection of the monitor video of vehicle, further includes:
Initial aggregate-value is reset.
Optionally, described according to forgetting factor and the corresponding aggregate-value of -1 image of kth, determine k-th of image pair
The aggregate-value answered, comprising:
Determine the product of the forgetting factor and the corresponding aggregate-value of -1 image of the kth；
It is determined as the corresponding aggregate-value of k-th of image after the product is added 1.
In above-mentioned technical proposal, when the vehicle for determining k-th of image is stationary vehicle, according to forgetting factor and kth-
The aggregate-value of 1 image determines the aggregate-value of k-th of image, specifically, determining the accumulative of -1 image of forgetting factor and kth
The product of value, and it is determined as the aggregate-value of k-th of image after the product is added 1, it can filter out accidentally to go out using forgetting factor
Existing false-alarm testing result, to improve the accuracy of detection.
Optionally, further includes:
The quantity of stationary vehicle is equal to after zero in determining k-th of image, determines the forgetting factor and described
The product of the corresponding aggregate-value of -1 image of kth；
The product is determined as the corresponding aggregate-value of k-th of image.
In above-mentioned technical proposal, when the vehicle for determining k-th of image is stationary vehicle, according to forgetting factor and kth-
The aggregate-value of 1 image determines the aggregate-value of k-th of image, specifically, determining the accumulative of -1 image of forgetting factor and kth
The product of value, and the product is determined as to the aggregate-value of k-th of image, the void accidentally occurred can be filtered out using forgetting factor
Alert testing result, to improve the accuracy of detection.
Optionally, there is exception in vehicle on the determination road, comprising:
When the sum of stationary vehicle in the vehicle of k-th of image is less than third threshold value, determine to go out on the road
Existing parking offense；
When the sum of stationary vehicle in the vehicle of k-th of image is not less than third threshold value, determine on the road
There is vehicle congestion.
It, can be according to k-th different of images when vehicle on determining road occurs abnormal in above-mentioned technical proposal
The quantity of stationary vehicle determines different abnormal conditions, when the quantity of the stationary vehicle of k-th of image is less than third threshold value,
Determine on the road parking offense occur；When the quantity of the stationary vehicle of k-th of image is not less than third threshold value, the road is determined
There is vehicle congestion in road, for different road abnormal conditions, different control strategies can be executed, to effectively solve different
Chang Wenti.
Optionally, the coordinate of the vehicle of the vehicle and marked completion to the marked completion on each picture into
The deep learning detection model is determined after row training, comprising:
Obtain initial model and training sample；Including marked on multiple pictures and each picture in the training sample
The coordinate of the vehicle of completion and the vehicle of the marked completion；
The multiple picture is input to the initial model and is trained study, obtains the corresponding calculating knot of each picture
Fruit；
According to the vehicle of the marked completion in the corresponding calculated result of each picture and each picture and described mark
The coordinate for remembering the vehicle completed, adjusts the initial model, until determining the deep learning detection model.
In above-mentioned technical proposal, carrying out vehicle detection by deep learning detection model can be improved detection performance, more
Vehicle is detected under kind light and a variety of visual angles, improves the detection accuracy of vehicle.
Correspondingly, the embodiment of the invention also provides a kind of devices of vehicle on monitoring road, comprising:
Acquiring unit, for obtaining the image collection of the monitor video of vehicle on road；Image is in described image set
Interval intercepts what the monitor video obtained at preset timed intervals；
Processing unit ,+1 image of k-th of image and kth for being directed in described image set, according to deep learning
Detection model determines the coordinate of the vehicle of+1 image of coordinate and the kth of the vehicle of k-th of image respectively；Institute
K is stated more than or equal to 1；The deep learning detection model is to the vehicle of the marked completion on each picture and described marked complete
At vehicle coordinate be trained after determine；According to+1 image of the coordinate of the vehicle of k-th of image and the kth
Vehicle coordinate, determine the vehicle distances between the vehicle of+1 image of vehicle and the kth of k-th of image；Root
According to the vehicle distances, the quantity of stationary vehicle in k-th of image is determined；The static vehicle in determining k-th of image
Quantity be greater than zero after, according to forgetting factor and the corresponding aggregate-value of -1 image of kth, determine k-th of image pair
The aggregate-value answered；When the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs different on the road
Often.
Optionally, the processing unit is also used to:
Judge whether the vehicle distances are less than first threshold, if so, determining that the vehicle of k-th of image is static
Vehicle；
Count the quantity of stationary vehicle described in k-th of image.
Optionally, the processing unit is also used to:
It obtains controlling the acquiring unit on road before the image collection of the monitor video of vehicle, by initial aggregate-value
It resets.
Optionally, the processing unit is specifically used for:
Determine the product of the forgetting factor and the corresponding aggregate-value of -1 image of the kth；
It is determined as the corresponding aggregate-value of k-th of image after the product is added 1.
Optionally, the processing unit is also used to:
The quantity of stationary vehicle is equal to after zero in determining k-th of image, determines the forgetting factor and described
The product of the corresponding aggregate-value of -1 image of kth；
The product is determined as the corresponding aggregate-value of k-th of image.
Optionally, the processing unit is specifically used for:
When the sum of stationary vehicle in the vehicle of k-th of image is less than third threshold value, determine to go out on the road
Existing parking offense；
When the sum of stationary vehicle in the vehicle of k-th of image is not less than third threshold value, determine on the road
There is vehicle congestion.
Optionally, the processing unit is specifically used for:
Obtain initial model and training sample；Including marked on multiple pictures and each picture in the training sample
The coordinate of the vehicle of completion and the vehicle of the marked completion；
The multiple picture is input to the initial model and is trained study, obtains the corresponding calculating knot of each picture
Fruit；
According to the vehicle of the marked completion in the corresponding calculated result of each picture and each picture and described mark
The coordinate for remembering the vehicle completed, adjusts the initial model, until determining the deep learning detection model.
Correspondingly, the embodiment of the invention also provides a kind of calculating equipment, comprising:
Memory, for storing program instruction；
Processor executes above-mentioned monitoring according to the program of acquisition for calling the program instruction stored in the memory
The method of vehicle on road.
Correspondingly, the embodiment of the invention also provides a kind of computer-readable non-volatile memory medium, including computer
Readable instruction, when computer is read and executes the computer-readable instruction, so that computer executes on above-mentioned monitoring road
The method of vehicle.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill in field, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of schematic diagram of system architecture provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of the method for vehicle on a kind of monitoring road provided in an embodiment of the present invention；
Fig. 3 is the schematic diagram that a kind of deep learning detection model provided in an embodiment of the present invention detects vehicle；
Fig. 4 is the flow diagram of the method for vehicle on another monitoring road provided in an embodiment of the present invention；
Fig. 5 is the structural schematic diagram of the device of vehicle on a kind of monitoring road provided in an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention make into
It is described in detail to one step, it is clear that described embodiments are only a part of the embodiments of the present invention, rather than whole implementation
Example.Based on the embodiments of the present invention, obtained by those of ordinary skill in the art without making creative efforts
All other embodiment, shall fall within the protection scope of the present invention.
Fig. 1 illustratively shows the system tray that the method that the embodiment of the present invention provides vehicle on monitoring road is applicable in
Structure, the system architecture may include analysis module 101, image processing module 102, event recognition module 103, three kinds of modules
It is sequentially connected and connects, wherein analysis module 101 is used to obtain the monitor video of vehicle on road, and presses from monitor video
It is sent to image processing module 102 according to prefixed time interval interception video image, image processing module 102 handles the image, knows
Not Chu information of vehicles in image, and information of vehicles is sent to event recognition module 103, so that event recognition module 103
According to information of vehicles judge vehicle whether remain static and road on whether have vehicle violation stop or congestion phenomenon.
Based on foregoing description, Fig. 2 illustratively shows vehicle on a kind of monitoring road provided in an embodiment of the present invention
The process of method, the process can be executed by the device of vehicle on monitoring road.As shown in Fig. 2, the process specifically includes:
Step 201, the image collection of the monitor video of vehicle on road is obtained.
Wherein, in image collection image be spaced at preset timed intervals interception monitor video obtain.It is interpreted as, obtains road
The monitor video of upper vehicle, and video image is intercepted according to prefixed time interval, form image collection, illustratively, adjacent two
The interception time of a image can be separated by 2s, can also rule of thumb set the prefixed time interval.
Step 202, for+1 image of k-th of image and kth in described image set, mould is detected according to deep learning
Type determines the coordinate of the vehicle of+1 image of coordinate and the kth of the vehicle of k-th of image respectively.
Herein, k is more than or equal to 1, and deep learning detection model is the vehicle to the marked completion on each picture and has marked
What the coordinate for the vehicle that note is completed determined after being trained.
Specifically, each car in image can be identified using deep learning detection model, and export each
The coordinate of vehicle as shown in figure 3, the deep learning detection model detects 7 vehicles, and can export the coordinate of each car, for example,
Result as shown in Table 1 can be exported.
Table 1
For support vehicles accuracy in detection, deep learning detection model can be determined in the following manner:
Obtain initial model and training sample；Wherein, including the mark on multiple pictures and each picture in training sample
Remember the coordinate of the vehicle of the vehicle and marked completion completed；Multiple pictures are input to initial model and are trained study, are obtained
To the corresponding calculated result of each picture；According to the vehicle of the marked completion on the corresponding calculated result of each picture and each picture
With the coordinate of the vehicle of marked completion, initial model is adjusted, until determining deep learning detection model.
Deep learning detection model can be the YOLO model or SSD mould in the target detection model based on end-to-end study
Type.
Step 203, according to the coordinate of the vehicle of+1 image of the coordinate of the vehicle of k-th of image and the kth, really
Vehicle distances between the vehicle of+1 image of vehicle and the kth of fixed k-th of image.
According to the vehicle coordinate of the vehicle of+1 image of the vehicle coordinate of the vehicle of k-th of image and kth, kth is determined
Vehicle distances between the vehicle of+1 image of vehicle and kth of a image.Herein, the vehicle of k-th of image can be multiple,
The vehicle of+1 image of kth may be it is multiple, for any one vehicle in k-th of image, determine any one vehicle with
The vehicle distances of all vehicles in+1 image of kth, for example, in step 202, detecting that the vehicle of k-th of image is respectively
Car11, car12, car13 totally 3 vehicles, detect+1 image of kth vehicle be respectively car21, car22, car23,
Car24, car25 totally 5 vehicles, calculate separately car11 to car21, car22, car23, car24, car25 vehicle distances,
Car12 to car21, the vehicle distances of car22, car23, car24, car25 and car13 to car21, car22, car23,
The vehicle distances of car24, car25.
Step 204, according to the vehicle distances, the quantity of stationary vehicle in k-th of image is determined.
Judge whether vehicle distances are less than first threshold, if so, the vehicle for determining k-th of image is stationary vehicle,
That is calculating multiple vehicle distances in step 203, as long as there are a vehicle distances to be less than first threshold, exist in other words
Minimum value in step 203 in calculated multiple vehicle distances can be determined this k-th as long as being less than first threshold
The vehicle of image is stationary vehicle.For example, the example in step 202, however, it is determined that the vehicle distances for going out car11 to car21 are less than
First threshold can then determine that car11 is stationary vehicle.Wherein, first threshold can rule of thumb be set, exemplary
, first threshold can be determined as 1m.
The quantity of stationary vehicle in k-th of image is counted, as in above-mentioned example, however, it is determined that go out three vehicles of k-th of image
It there are car11, car12 is stationary vehicle in, it is determined that the quantity of stationary vehicle is 2 in k-th of image.
Step 205, the quantity of stationary vehicle is greater than after zero in determining k-th of image, according to forgetting factor and
The corresponding aggregate-value of -1 image of kth determines the corresponding aggregate-value of k-th of image.
In the embodiment of the present invention, forgetting factor is introduced, for filtering out the false-alarm testing result accidentally occurred, specifically,
The quantity of stationary vehicle is greater than after zero in determining k-th of image, determines that forgetting factor and -1 image of kth are corresponding tired
The product of evaluation, and it is determined as the corresponding aggregate-value of k-th of image after product is added 1.For example, forgetting factor is 0.8, kth -1
The corresponding aggregate-value of a image is 1.2, then can determine that the corresponding aggregate-value of k-th of image is 0.8 × 1.2+1=1.96.
It certainly, further include determining that the quantity of stationary vehicle in k-th of image is equal to zero in above-described embodiment, further root
When determining the corresponding aggregate-value of k-th of image according to forgetting factor and the corresponding aggregate-value of -1 image of kth, it can first determine to lose
Forget the product of the factor and the corresponding aggregate-value of -1 image of kth, it is corresponding accumulative that the product is then determined as k-th of image
Value.For example, forgetting factor is 0.8, the corresponding aggregate-value of -1 image of kth is 1.2, then can determine that k-th of image is corresponding
Aggregate-value be 0.8 × 1.2=0.96.
Step 206, when the corresponding aggregate-value of k-th of image is greater than second threshold, vehicle on the road is determined
Occur abnormal.
Specifically, the sum and the size of third threshold value of stationary vehicle in the vehicle of k-th of image can be compared, specifically
, the sum of stationary vehicle in the vehicle of k-th of image is first determined, for example, the vehicle of k-th of image is 10, for this
Any one vehicle in 10 vehicles determines the vehicle distances of all vehicles in any one vehicle and+1 image of kth, if this
Minimum value in one vehicle and+1 image of kth in the vehicle distances of all vehicles is less than first threshold, and determines k-th of figure
When being greater than second threshold as corresponding aggregate-value, the sum of stationary vehicle in the vehicle of k-th of image is counted, and by the kth
In the vehicle of a image stationary vehicle sum make comparisons with third threshold value, when in the vehicle of k-th of image stationary vehicle it is total
When number is less than third threshold value, determine occur parking offense on road；When the sum of stationary vehicle in the vehicle of k-th of image is not small
When third threshold value, determine occur vehicle congestion on road.
In addition, the accuracy in order to ensure aggregate-value, on obtaining road before the image collection of the monitor video of vehicle,
It also needs to reset initial aggregate-value.That is, needing first before a process cycle starts by the tired of a upper process cycle
Evaluation is reset.
In order to preferably explain the present invention, the stream of vehicle on the monitoring road will be described under specific implement scene below
Journey is arranged in a process cycle and gets K image in total, it may also be said to, vehicle on the road got in step 201
Monitor video image collection in include K image, as shown in figure 4, detailed process is as follows:
Step 401 ,+1 image of k-th of image and kth is obtained.
Step 402, the vehicle of k-th of image and the distance between the vehicle of+1 image of kth are determined.
Step 403, the quantity of stationary vehicle in k-th of image is determined.
Step 404, judge whether the quantity of stationary vehicle in k-th of image is greater than zero, if so, step 405 is turned to, it is no
Then turn to step 406.
Step 405, the product of forgetting factor and the corresponding aggregate-value of -1 image of kth is determined, and after the product is added 1 really
It is set to the corresponding aggregate-value of k-th of image.
Step 406, it determines the product of forgetting factor and the corresponding aggregate-value of -1 image of kth, and the product is determined as
The corresponding aggregate-value of k-th of image.
Step 407, judge whether the corresponding aggregate-value of k-th of image is greater than second threshold, if so, step 408 is turned to,
Otherwise step 411 is turned to.
Step 408, judge whether the quantity of stationary vehicle in k-th of image is less than third threshold value, if so, turning to step
409, otherwise turn to step 410.
Step 409, determine occur parking offense on road.
Step 410, determine occur vehicle congestion on road.
Step 411, judge whether the present treatment period judges to finish；If so, terminating, step 401 is otherwise turned to.
Assuming that K=20, forgetting factor=0.8 after resetting first to accumulated value, carry out the 1st image and the 2nd image
Analysis, judges after having stationary vehicle on the 1st image according to the 1st image and the 2nd image, it is determined that the 1st image
Aggregate-value is 1；2nd image and the 3rd image are analyzed, judge the 2nd according to the 2nd image and the 3rd image
After having stationary vehicle on image, it is determined that the aggregate-value of the 2nd image is 1 × 0.8+1=1.8；To the 3rd image and the 4th
Image is analyzed, after being judged on the 3rd image according to the 3rd image and the 4th image without stationary vehicle, it is determined that the 3rd
The aggregate-value of a image is 1.8 × 0.8=1.44, and so on, until present treatment end cycle or output road are abnormal.This
Place, when the corresponding aggregate-value of k-th of image is not more than second threshold, then judge the present treatment period whether it is determined that finish,
If it is not, then obtaining+2 images of+1 image of kth and kth again, the corresponding aggregate-value of+1 image of kth is further determined that.
It is true respectively according to deep learning detection model for+1 image of k-th of image and kth in above-mentioned technical proposal
The vehicle of+1 image of vehicle and kth of k-th of image is made, and determine+1 image of vehicle and kth of k-th of image
Vehicle distances between vehicle, herein, the vehicle of k-th of image can be it is multiple, the vehicle of+1 image of kth may be more
It is a, for any one vehicle in k-th of image, determine the vehicle of all vehicles in any one vehicle and+1 image of kth
As long as distance is judged that the vehicle of k-th of image is stationary vehicle, is being determined have vehicle distances to be less than first threshold
The quantity of stationary vehicle is greater than after zero in k-th of image, determines the according to the aggregate-value of -1 image of forgetting factor and kth
The aggregate-value of k image, and when the aggregate-value for determining k-th of image is greater than second threshold, determine that vehicle occurs on road
It is abnormal.The technical solution uses the multiple images joint-detection changed for vehicle relative position to improve parking offense or vehicle
Whether the testing result of congestion is same vehicle without tracking and identifying in picture, it is only necessary in each vehicle detected
Heart point coordinate participates in operation you can get it conclusion, under the premise of ensuring the detection accuracy to road abnormality detection, improves detection effect
Rate.
Based on the same inventive concept, Fig. 5 illustratively shows a kind of monitoring road provided in an embodiment of the present invention and gets on the bus
Device structure, the device can execute monitoring road on vehicle method process.
Acquiring unit 501, for obtaining the image collection of the monitor video of vehicle on road；Image in described image set
Interval intercepts the monitor video and obtains at preset timed intervals；
Processing unit 502 ,+1 image of k-th of image and kth for being directed in described image set, according to depth
Detection model is practised, determines the coordinate of the vehicle of+1 image of coordinate and the kth of the vehicle of k-th of image respectively；
The k is more than or equal to 1；The deep learning detection model is to the vehicle of the marked completion on each picture and described marked
What the coordinate of the vehicle of completion determined after being trained；Schemed according to the coordinate of the vehicle of k-th of image and the kth+1
The coordinate of the vehicle of picture determines the vehicle distances between the vehicle of+1 image of vehicle and the kth of k-th of image；
According to the vehicle distances, the quantity of stationary vehicle in k-th of image is determined；It is static in determining k-th of image
The quantity of vehicle is greater than after zero, according to forgetting factor and the corresponding aggregate-value of -1 image of kth, determines k-th of image
Corresponding aggregate-value；When the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs on the road
It is abnormal.
Optionally, the processing unit 502 is also used to:
Judge whether the vehicle distances are less than first threshold, if so, determining that the vehicle of k-th of image is static
Vehicle；
Count the quantity of stationary vehicle described in k-th of image.
Optionally, the processing unit 502 is also used to:
On controlling the acquisition of acquiring unit 501 road before the image collection of the monitor video of vehicle, will initially it tire out
Evaluation is reset.
Optionally, the processing unit 502 is specifically used for:
Determine the product of the forgetting factor and the corresponding aggregate-value of -1 image of the kth；
It is determined as the corresponding aggregate-value of k-th of image after the product is added 1.
Optionally, the processing unit 502 is also used to:
The quantity of stationary vehicle is equal to after zero in determining k-th of image, determines the forgetting factor and described
The product of the corresponding aggregate-value of -1 image of kth；
The product is determined as the corresponding aggregate-value of k-th of image.
Optionally, the processing unit 502 is specifically used for:
When the sum of stationary vehicle in the vehicle of k-th of image is less than third threshold value, determine to go out on the road
Existing parking offense；
When the sum of stationary vehicle in the vehicle of k-th of image is not less than third threshold value, determine on the road
There is vehicle congestion.
Optionally, the processing unit 502 is specifically used for:
Obtain initial model and training sample；Including marked on multiple pictures and each picture in the training sample
The coordinate of the vehicle of completion and the vehicle of the marked completion；
The multiple picture is input to the initial model and is trained study, obtains the corresponding calculating knot of each picture
Fruit；
According to the vehicle of the marked completion in the corresponding calculated result of each picture and each picture and described mark
The coordinate for remembering the vehicle completed, adjusts the initial model, until determining the deep learning detection model.
Based on the same inventive concept, the embodiment of the invention also provides a kind of calculating equipment, comprising:
Memory, for storing program instruction；
Processor executes above-mentioned monitoring according to the program of acquisition for calling the program instruction stored in the memory
The method of vehicle on road.
Based on the same inventive concept, the embodiment of the invention also provides a kind of computer-readable non-volatile memory medium,
Including computer-readable instruction, when computer is read and executes the computer-readable instruction, so that computer execution is above-mentioned
The method for monitoring vehicle on road.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (16)
1. a kind of method of vehicle on monitoring road characterized by comprising
Obtain the image collection of the monitor video of vehicle on road；Image is interval interception at preset timed intervals in described image set
What the monitor video obtained；
It is determined respectively for+1 image of k-th of image and kth in described image set according to deep learning detection model
The coordinate of the vehicle of+1 image of the coordinate of the vehicle of k-th of image and the kth out；The k is more than or equal to 1；It is described
Deep learning detection model be to the coordinate of the vehicle of the vehicle and marked completion of the marked completion on each picture into
It is determined after row training；
According to the coordinate of the vehicle of+1 image of the coordinate of the vehicle of k-th of image and the kth, determine described k-th
Vehicle distances between the vehicle of+1 image of the vehicle of image and the kth；
According to the vehicle distances, the quantity of stationary vehicle in k-th of image is determined；
The quantity of stationary vehicle is greater than after zero in determining k-th of image, according to -1 image pair of forgetting factor and kth
The aggregate-value answered determines the corresponding aggregate-value of k-th of image；
When the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs abnormal on the road.
2. the method as described in claim 1, which is characterized in that it is described according to the vehicle distances, determine k-th of image
The quantity of middle stationary vehicle, comprising:
Judge whether the vehicle distances are less than first threshold, if so, the vehicle for determining k-th of image is static vehicle
?；
Count the quantity of stationary vehicle described in k-th of image.
3. the method as described in claim 1, which is characterized in that the image set of the monitor video of vehicle on the acquisition road
Before conjunction, further includes:
Initial aggregate-value is reset.
4. the method as described in claim 1, which is characterized in that described corresponding tired according to forgetting factor and -1 image of kth
Evaluation determines the corresponding aggregate-value of k-th of image, comprising:
Determine the product of the forgetting factor and the corresponding aggregate-value of -1 image of the kth；
It is determined as the corresponding aggregate-value of k-th of image after the product is added 1.
5. the method as described in claim 1, which is characterized in that further include:
The quantity of stationary vehicle is equal to after zero in determining k-th of image, determines the forgetting factor and the kth -1
The product of the corresponding aggregate-value of a image；
The product is determined as the corresponding aggregate-value of k-th of image.
6. the method as described in claim 1, which is characterized in that vehicle occurs abnormal on the determination road, comprising:
When the sum of stationary vehicle in the vehicle of k-th of image is less than third threshold value, determines and disobeyed on the road
Chapter parking；
When the sum of stationary vehicle in the vehicle of k-th of image is not less than third threshold value, determines and occur on the road
Vehicle congestion.
7. such as method as claimed in any one of claims 1 to 6, which is characterized in that the marked completion on each picture
The coordinate of vehicle and the vehicle of the marked completion determines the deep learning detection model after being trained, comprising:
Obtain initial model and training sample；It include the marked completion on multiple pictures and each picture in the training sample
Vehicle and the marked completion vehicle coordinate；
The multiple picture is input to the initial model and is trained study, obtains the corresponding calculated result of each picture；
According to the vehicle of the marked completion in the corresponding calculated result of each picture and each picture and described marked complete
At vehicle coordinate, the initial model is adjusted, until determining the deep learning detection model.
8. the device of vehicle on a kind of monitoring road characterized by comprising
Acquiring unit, for obtaining the image collection of the monitor video of vehicle on road；Image is by pre- in described image set
If time interval intercepts what the monitor video obtained；
Processing unit, for being detected according to deep learning for+1 image of k-th of image and kth in described image set
Model determines the coordinate of the vehicle of+1 image of coordinate and the kth of the vehicle of k-th of image respectively；The k is big
In equal to 1；The deep learning detection model is the vehicle and the marked completion to the marked completion on each picture
What the coordinate of vehicle determined after being trained；According to the vehicle of+1 image of the coordinate of the vehicle of k-th of image and the kth
Coordinate, determine the vehicle distances between the vehicle of+1 image of vehicle and the kth of k-th of image；According to institute
Vehicle distances are stated, determine the quantity of stationary vehicle in k-th of image；The stationary vehicle in determining k-th of image
Quantity is greater than after zero, according to forgetting factor and the corresponding aggregate-value of -1 image of kth, determines that k-th of image is corresponding
Aggregate-value；When the corresponding aggregate-value of k-th of image is greater than second threshold, determine that vehicle occurs abnormal on the road.
9. device as claimed in claim 8, which is characterized in that the processing unit is also used to:
Judge whether the vehicle distances are less than first threshold, if so, the vehicle for determining k-th of image is static vehicle
?；
Count the quantity of stationary vehicle described in k-th of image.
10. device as claimed in claim 8, which is characterized in that the processing unit is also used to:
It is obtained on road before the image collection of the monitor video of vehicle controlling the acquiring unit, initial aggregate-value is clear
Zero.
11. device as claimed in claim 8, which is characterized in that the processing unit is specifically used for:
Determine the product of the forgetting factor and the corresponding aggregate-value of -1 image of the kth；
It is determined as the corresponding aggregate-value of k-th of image after the product is added 1.
12. device as claimed in claim 8, which is characterized in that the processing unit is also used to:
The quantity of stationary vehicle is equal to after zero in determining k-th of image, determines the forgetting factor and the kth -1
The product of the corresponding aggregate-value of a image；
The product is determined as the corresponding aggregate-value of k-th of image.
13. device as claimed in claim 8, which is characterized in that the processing unit is specifically used for:
When the sum of stationary vehicle in the vehicle of k-th of image is less than third threshold value, determines and disobeyed on the road
Chapter parking；
When the sum of stationary vehicle in the vehicle of k-th of image is not less than third threshold value, determines and occur on the road
Vehicle congestion.
14. such as the described in any item devices of claim 8 to 13, which is characterized in that the processing unit is specifically used for:
Obtain initial model and training sample；It include the marked completion on multiple pictures and each picture in the training sample
Vehicle and the marked completion vehicle coordinate；
The multiple picture is input to the initial model and is trained study, obtains the corresponding calculated result of each picture；
According to the vehicle of the marked completion in the corresponding calculated result of each picture and each picture and described marked complete
At vehicle coordinate, the initial model is adjusted, until determining the deep learning detection model.
15. a kind of calculating equipment characterized by comprising
Memory, for storing program instruction；
Processor requires 1 to 7 according to the program execution benefit of acquisition for calling the program instruction stored in the memory
Described in any item methods.
16. a kind of computer-readable non-volatile memory medium, which is characterized in that including computer-readable instruction, work as computer
When reading and executing the computer-readable instruction, so that computer executes method as described in any one of claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910103337.8A CN109948436B (en) | 2019-02-01 | 2019-02-01 | Method and device for monitoring vehicles on road |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910103337.8A CN109948436B (en) | 2019-02-01 | 2019-02-01 | Method and device for monitoring vehicles on road |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109948436A true CN109948436A (en) | 2019-06-28 |
| CN109948436B CN109948436B (en) | 2020-12-08 |
Family
ID=67007395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910103337.8A Active CN109948436B (en) | 2019-02-01 | 2019-02-01 | Method and device for monitoring vehicles on road |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109948436B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113096394A (en) * | 2021-03-29 | 2021-07-09 | 浙江大华技术股份有限公司 | Method and device for determining object of target type, storage medium and electronic device |
| CN113408320A (en) * | 2020-03-16 | 2021-09-17 | 上海博泰悦臻网络技术服务有限公司 | Method, electronic device, and computer storage medium for vehicle collision avoidance |
| CN113689587A (en) * | 2021-08-25 | 2021-11-23 | 山东新一代信息产业技术研究院有限公司 | Method and system for inspecting trunk occupation in park |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1868212A (en) * | 2003-09-24 | 2006-11-22 | 杨伯翰大学 | Automated estimation of average stopped delay at signalized intersections |
| CN103778785A (en) * | 2012-10-23 | 2014-05-07 | 南开大学 | Vehicle tracking and detecting method based on parking lot environment video monitoring |
| CN105321342A (en) * | 2015-12-07 | 2016-02-10 | 北京航空航天大学 | Intersection vehicle queuing length detection method based on aerial video |
| CN105809981A (en) * | 2016-05-26 | 2016-07-27 | 陕西理工学院 | Single-intersection signal light control method based on instantaneous stop delay |
| CN106650913A (en) * | 2016-12-31 | 2017-05-10 | 中国科学技术大学 | Deep convolution neural network-based traffic flow density estimation method |
| CN106652445A (en) * | 2016-11-15 | 2017-05-10 | 成都通甲优博科技有限责任公司 | Road traffic accident judging method and device |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
| CN107330373A (en) * | 2017-06-02 | 2017-11-07 | 重庆大学 | A kind of parking offense monitoring system based on video |
| CN107730903A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Parking offense and the car vision detection system that casts anchor based on depth convolutional neural networks |
| CN107766821A (en) * | 2017-10-23 | 2018-03-06 | 江苏鸿信系统集成有限公司 | All the period of time vehicle detecting and tracking method and system in video based on Kalman filtering and deep learning |
| CN107945523A (en) * | 2017-11-27 | 2018-04-20 | 北京华道兴科技有限公司 | A kind of road vehicle detection method, DETECTION OF TRAFFIC PARAMETERS method and device |
| CN109147341A (en) * | 2018-09-14 | 2019-01-04 | 杭州数梦工场科技有限公司 | Violation vehicle detection method and device |
| CN109241938A (en) * | 2018-09-27 | 2019-01-18 | 深圳金三立视频科技股份有限公司 | Congestion in road detection method and terminal |
-
2019
- 2019-02-01 CN CN201910103337.8A patent/CN109948436B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1868212A (en) * | 2003-09-24 | 2006-11-22 | 杨伯翰大学 | Automated estimation of average stopped delay at signalized intersections |
| CN103778785A (en) * | 2012-10-23 | 2014-05-07 | 南开大学 | Vehicle tracking and detecting method based on parking lot environment video monitoring |
| CN105321342A (en) * | 2015-12-07 | 2016-02-10 | 北京航空航天大学 | Intersection vehicle queuing length detection method based on aerial video |
| CN105809981A (en) * | 2016-05-26 | 2016-07-27 | 陕西理工学院 | Single-intersection signal light control method based on instantaneous stop delay |
| CN106652445A (en) * | 2016-11-15 | 2017-05-10 | 成都通甲优博科技有限责任公司 | Road traffic accident judging method and device |
| CN106650913A (en) * | 2016-12-31 | 2017-05-10 | 中国科学技术大学 | Deep convolution neural network-based traffic flow density estimation method |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
| CN107330373A (en) * | 2017-06-02 | 2017-11-07 | 重庆大学 | A kind of parking offense monitoring system based on video |
| CN107730903A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Parking offense and the car vision detection system that casts anchor based on depth convolutional neural networks |
| CN107766821A (en) * | 2017-10-23 | 2018-03-06 | 江苏鸿信系统集成有限公司 | All the period of time vehicle detecting and tracking method and system in video based on Kalman filtering and deep learning |
| CN107945523A (en) * | 2017-11-27 | 2018-04-20 | 北京华道兴科技有限公司 | A kind of road vehicle detection method, DETECTION OF TRAFFIC PARAMETERS method and device |
| CN109147341A (en) * | 2018-09-14 | 2019-01-04 | 杭州数梦工场科技有限公司 | Violation vehicle detection method and device |
| CN109241938A (en) * | 2018-09-27 | 2019-01-18 | 深圳金三立视频科技股份有限公司 | Congestion in road detection method and terminal |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113408320A (en) * | 2020-03-16 | 2021-09-17 | 上海博泰悦臻网络技术服务有限公司 | Method, electronic device, and computer storage medium for vehicle collision avoidance |
| CN113096394A (en) * | 2021-03-29 | 2021-07-09 | 浙江大华技术股份有限公司 | Method and device for determining object of target type, storage medium and electronic device |
| CN113689587A (en) * | 2021-08-25 | 2021-11-23 | 山东新一代信息产业技术研究院有限公司 | Method and system for inspecting trunk occupation in park |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109948436B (en) | 2020-12-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109948436A (en) | The method and device of vehicle on a kind of monitoring road | |
| CN109086734B (en) | Method and device for positioning pupil image in human eye image | |
| CN104077594B (en) | A kind of image-recognizing method and device | |
| WO2016086792A1 (en) | Driving behavior analysis method and device | |
| CN105095829A (en) | Face recognition method and system | |
| CN112798811B (en) | Speed measurement method, device and equipment | |
| CN110232379A (en) | A kind of vehicle attitude detection method and system | |
| CN106662970B (en) | Method and device for setting interruption threshold of fingerprint identifier and terminal equipment | |
| CN105225523A (en) | A kind of parking space state detection method and device | |
| CN106570478A (en) | Object loss determine method and device in visual tracking | |
| CN109190639A (en) | A kind of vehicle color identification method, apparatus and system | |
| CN111598827A (en) | Appearance flaw detection method, electronic device and storage medium | |
| CN105139384B (en) | The method and apparatus of defect capsule detection | |
| CN110245544A (en) | A kind of method and device of determining dead ship condition | |
| CN106991418A (en) | Winged insect detection method, device and terminal | |
| CN109101922A (en) | Operating personnel device, assay, device and electronic equipment | |
| CN109102026A (en) | A kind of vehicle image detection method, apparatus and system | |
| CN112597995B (en) | License plate detection model training method, device, equipment and medium | |
| CN114155241A (en) | Foreign matter detection method and device and electronic equipment | |
| CN106097708A (en) | One determines method and device with car | |
| CN108460333B (en) | Ground detection method and device based on depth map | |
| CN106355247B (en) | Data processing method and device, chip and electronic equipment | |
| CN103955929B (en) | Image local edge pattern and non-edge mode judging method and judgment means | |
| CN111080701A (en) | Intelligent cabinet object detection method and device, server and storage medium | |
| CN105979153A (en) | Day and night mode identification and switching method and system |
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 |