WO2011079691A1

WO2011079691A1 - Method and device for detecting nighttime vehicles

Info

Publication number: WO2011079691A1
Application number: PCT/CN2010/079525
Authority: WO
Inventors: 胡健; 周子庭; 魏俊华
Original assignee: 北京世纪高通科技有限公司
Priority date: 2009-12-29
Filing date: 2010-12-07
Publication date: 2011-07-07
Also published as: CN101770571B; CN101770571A

Abstract

The present invention relates to the field of intelligent traffic systems, and particularly discloses a method and device for detecting nighttime vehicles so as to resolve the problem that the accuracy of nighttime vehicles detection is reduced due to the impact of illumination and halo of vehicle lamps. The method for detecting nighttime vehicles comprises the following steps of: acquiring adjacent video frames (101); extracting an interested area on the basis of a difference between two adjacent video frames (102); according to the interested area, detecting nighttime vehicles (103). The technical scheme, provided by examples of the present invention, is suitable for intelligent traffic systems.

Description

The present invention claims the priority of the Chinese application filed on December 29, 2009, the application number is 200910244106. The citations are incorporated herein by reference.

Technical field

The present invention relates to the field of intelligent transportation systems, and more particularly to a method and apparatus for nighttime vehicle detection.

Background technique

Vehicle detection is an important part of the intelligent transportation system. Low night illumination and large changes in light intensity are the biggest challenges for all-weather testing of vehicles.

In the prior art, the region of interest is mostly extracted by the background difference method, and the vehicle is detected according to the extracted region of interest. The background difference method obtains the motion foreground by the difference between the current frame image and the background frame image, and the background extraction and real-time updating are the key steps of the background difference method.

In the process of implementing the present invention, the inventors have found that due to the large change of light intensity at night, especially the influence of illumination and halo of the lamp, the background update is quite difficult, and the region of interest extracted by the background subtraction method contains a large number of redundant lights. Information reduces the accuracy of nighttime vehicle detection.

Summary of the invention

Embodiments of the present invention provide a method and apparatus for nighttime vehicle detection, which solves

In order to achieve the above object, the embodiment of the present invention adopts the following technical solutions:

A method for nighttime vehicle detection, comprising: acquiring adjacent video frames; extracting a region of interest according to a difference between two adjacent video frames; detecting a night vehicle according to the region of interest.

A device for detecting a vehicle at night, comprising:

An acquiring unit, configured to acquire an adjacent video frame;

An extracting unit, configured to extract a region of interest according to a difference between two adjacent video frames acquired by the acquiring unit;

a first detecting unit, configured to enter a nighttime vehicle according to the region of interest extracted by the extracting unit Line detection.

The method and apparatus for nighttime vehicle detection provided by the embodiments of the present invention utilizes a difference between two adjacent video frames such that a target of motion is retained in the difference image, thereby obtaining a region of interest, the region of interest having A certain degree of gradient effect, so that some of the slowly changing shadows and halos are subtracted when performing two adjacent video difference processing, thereby achieving the effect of eliminating partial shadow and halo effects, and solving the prior art due to shadows and halos. The impact of the problem is that the accuracy of vehicle detection at night is reduced.

DRAWINGS

1 is a flowchart of a method for detecting a night vehicle according to an embodiment of the present invention;

2 is a flowchart of a method for detecting a night vehicle according to still another embodiment of the present invention;

3 is a flow chart of step 205 in the flowchart of the night vehicle detection method provided by the embodiment of the invention shown in FIG. 2;

4 is a schematic structural diagram 1 of a device for detecting a nighttime vehicle according to an embodiment of the present invention; FIG. 5 is a schematic structural diagram of a first detecting unit in a device for detecting a nighttime vehicle according to an embodiment of the invention shown in FIG.

FIG. 6 is a schematic structural diagram of a device for detecting a nighttime vehicle according to an embodiment of the present invention; FIG. 7 is a schematic structural diagram of a judging unit in a device for detecting a nighttime vehicle provided by the embodiment of the invention shown in FIG.

detailed description

In order to solve the problem of reduced night vehicle detection accuracy due to shadows and halos, embodiments of the present invention provide a method and apparatus for nighttime vehicle detection.

As shown in FIG. 1 , a method for detecting a night vehicle according to an embodiment of the present invention includes: Step 101: Acquire an adjacent video frame;

Step 102: Extract a region of interest according to a difference between two adjacent video frames.

Step 103: Perform a night vehicle detection according to the region of interest.

The nighttime vehicle detection method provided by the embodiment of the present invention utilizes a difference between two adjacent video frames such that the target of the motion is retained in the difference image, thereby obtaining a region of interest, the region of interest having a certain degree Gradient effect, which causes some slow-changing shadows and halos to advance The two adjacent video difference processing is subtracted, thereby achieving the effect of eliminating partial shadow and halo effects, and solving the problem of reduced night vehicle detection accuracy due to the influence of shadows and halos in the prior art.

In order to enable a person skilled in the art to more clearly understand the technical solutions provided by the embodiments of the present invention, the method for detecting nighttime vehicles provided by the embodiments of the present invention will be described in detail below with reference to specific embodiments.

As shown in FIG. 2, a method for nighttime vehicle detection according to another embodiment of the present invention includes: Step 201: Acquire an adjacent video frame;

Step 202: Extract a region of interest according to a difference between two adjacent video frames.

In this embodiment, the difference between the two adjacent videos is obtained by A f =f ₂ (x, y) - fx, y), where (^) is the video frame of the moment, f ₂ (x, y) is the video frame at time t ₂ . That is, the difference between the two adjacent videos is realized by the frame difference method. The frame difference method subtracts the two frames before and after, and the target of the motion is retained in the difference image to become the region of interest.

Step 203: Detect a night vehicle according to the region of interest.

It is worth noting that in the night scene, the increase of large trucks makes the road surface prone to resonance, causing camera shake and sudden flashing of the lights to bring a lot of noise. At the same time, the brightness of the lamp part or the highly reflective part is very different from other areas, resulting in a large difference. The intensity distribution of the foreground of the motion is extremely uneven, and the corresponding extracted regions of interest will have obvious faults. Random noise and fault phenomena will lead to false detection of vehicles at night. In order to prevent random noise and the occurrence of a fault phenomenon, before the step 203, the method further includes:

Step 204: Detect a ratio of the region of interest to a preset virtual coil, where the virtual coil is a quadrilateral in a corresponding region of the video image;

Step 205: Determine, according to the ratio, whether a vehicle passes through the virtual coil.

In the present embodiment, on the basis of analyzing the random noise and the cause of the fault phenomenon, a two-point hypothesis is proposed, assuming a straight line distance between a plurality of regions of interest that are mis-segmented due to the fault. To be less than the linear distance between two vehicles, the time domain is expressed as: the number of frame intervals of multiple regions of interest that are mis-segmented by the fault through the preset virtual coil is less than the frame interval between the two vehicles. Number. Set the minimum number of frame intervals between the two cars. Hypothesis 2, the number of frames occupied by the region of interest due to random noise is less than the number of frames occupied by the actual vehicle passing through the virtual coil, and the minimum number of possessions of the vehicle is set. On the basis of the above two assumptions, the step 205, as shown in FIG. 3, includes: Step 2051, determining the region of interest of the vehicle entering the virtual coil for the first time and the region of interest of the vehicle entering the virtual coil Whether the frame interval between them is less than the preset minimum number of frame intervals between the two cars;

Step 2052, if the frame interval between the region of interest of the first time entering the virtual coil and the region of interest of the vehicle entering the virtual coil is not less than a preset minimum number of frame intervals between the two cars, Whether the number of frames occupied by the region of interest through the virtual coil is less than a preset number of frames;

It is worth noting that when the frame interval between the region of interest of the first time entering the virtual coil and the region of interest of the vehicle entering the virtual coil is less than the preset minimum number of frame intervals between the two cars, Explain that the region of interest is a mis-segmentation caused by the fault phenomenon.

Step 2053: When the number of frames is greater than a preset number of frames, outputting a result of passing the vehicle in the virtual coil, otherwise outputting a result of not passing the vehicle in the virtual coil.

It should be noted that when the number of frames is less than or equal to the preset number of frames, it indicates that the region of interest is a false segmentation due to random noise.

Step 206: When the virtual coil is converted from vehicle to vehicleless, the night vehicle is allowed to be detected.

Further, in order to conveniently and easily determine the vehicle type of the vehicle when performing vehicle detection, the detecting the night vehicle according to the region of interest includes:

The vehicle type of the night vehicle is obtained according to the distribution dispersion of the region of interest within the preset virtual coil.

In this embodiment, the virtual coil of the current lane is divided into three parts of the left middle and the right, and the distribution of the information points when the motion foreground passes through the virtual coil is statistically analyzed and the dispersion of the distribution is analyzed. The dispersion is measured by the standard deviation. The degree of dispersion of information points distribution of different vehicle types is different. Large vehicles occupy almost the entire lane, and the corresponding distribution is basically the same in the three pre-divided areas. The degree is low; while the small car mainly occupies 2/3 of the lane, and the distribution dispersion is high. The vehicle model is judged by the height of the distribution dispersion.

The nighttime vehicle detection method provided by the embodiment of the present invention utilizes a difference between two adjacent video frames such that the target of the motion is retained in the difference image, thereby obtaining a region of interest, the region of interest having a certain degree The gradient effect, so that some of the slowly changing shadows and halos are subtracted during the processing of the two adjacent video differences, thereby eliminating the effects of partial shadows and halos, and solving the effects of shadows and halos in the prior art. A problem that causes the accuracy of vehicle detection at night to decrease.

As shown in FIG. 4, the nighttime vehicle detecting apparatus provided by the embodiment of the present invention includes:

The obtaining unit 301 is configured to obtain the adjacent video frame. For the specific implementation, refer to step 201 shown in Figure 2, and details are not described herein.

The extracting unit 302 is configured to extract the region of interest according to the difference between the two adjacent video frames acquired by the acquiring unit. For the specific implementation method, refer to step 202 shown in FIG. Narration.

The first detecting unit 303 is configured to detect the night vehicle according to the region of interest extracted by the extracting unit. In this embodiment, the first detecting unit, as shown in FIG. 5, includes a vehicle type acquiring sub-unit 3031, and is configured to discretely distribute the region of interest extracted by the extracting unit in the preset virtual coil. Get the model of the night vehicle. For the specific implementation, refer to step 203 shown in Figure 2, and details are not described herein.

In order to prevent the fault detection caused by the fault phenomenon and the random noise, as shown in FIG. 6, the device further includes:

The second detecting unit 304 is configured to detect a ratio of the region of interest extracted by the extracting unit to a preset virtual coil, where the virtual coil is a quadrangle in a corresponding region of the video image; The description of step 204 is not repeated here.

The determining unit 305 is configured to determine whether a vehicle passes through the virtual coil according to the ratio obtained by the detecting unit. For a specific implementation method, refer to step 205 shown in FIG. 2, and details are not described herein. In this embodiment, the determining unit, as shown in FIG. 7, includes:

The first determining subunit 3051 is configured to determine whether a frame interval between the region of interest of the first time entering the virtual coil and the region of interest of the vehicle entering the virtual coil is less than a preset minimum frame interval between the two vehicles For the specific implementation method, refer to step 2051 shown in FIG. 2, and details are not described herein again.

a second judging subunit 3052, configured to determine, in the first judging subunit, that a frame interval between a region of interest that first enters the virtual coil and a region of interest of a vehicle that enters the virtual coil is not less than a pre-predetermined When the number of the minimum number of frames between the two vehicles is set, it is determined whether the number of frames occupied by the region of interest through the virtual coil is smaller than a preset number of frames. For the specific implementation method, refer to step 2052 shown in FIG. 2 . The description is not repeated here.

The output subunit 3053 is configured to output, when the second judging subunit determines that the number of frames is greater than a preset number of frames, output a result of passing the vehicle in the virtual coil, otherwise outputting the virtual coil without passing through the vehicle the result of. For the specific implementation method, refer to step 2053 shown in Figure 2, and the details are not described here.

The permitting unit 306 is configured to allow the first detecting unit to detect the night vehicle when the determining unit determines that the vehicle passes through the vehicle to transition to no vehicle passing. For a specific implementation method, refer to step 206 shown in Figure 2, and details are not described herein.

The apparatus for nighttime vehicle detection provided by the embodiment of the present invention utilizes a difference between two adjacent video frames such that a target of motion is retained in the difference image, thereby obtaining a region of interest, the region of interest having a certain degree The gradient effect, so that some of the slowly changing shadows and halos are subtracted during the processing of the two adjacent video differences, thereby eliminating the effects of partial shadows and halos, and solving the effects of shadows and halos in the prior art. A problem that causes the accuracy of vehicle detection at night to decrease.

The method and apparatus for nighttime vehicle detection provided by the embodiments of the present invention are applicable to an intelligent transportation system. A person skilled in the art can understand that all or part of the steps of implementing the above embodiments can be completed by a program to instruct related hardware, and the program can be stored in a computer readable storage medium, such as ROM/RAM, magnetic. Disc or CD. The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present invention. It should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the appended claims.

Claims

Claim

A method for detecting a night vehicle, characterized in that it comprises:

Obtain adjacent video frames;

Extracting a region of interest according to a difference between two adjacent video frames;

The night vehicle is detected based on the region of interest.

2. The method according to claim 1, wherein before the detecting the night vehicle according to the region of interest, the method further comprises:

Detecting a ratio of the region of interest to a preset virtual coil, wherein the virtual coil is a quadrangle in a corresponding region of the video image;

Determining, according to the ratio, whether a vehicle passes through the virtual coil;

Night vehicle detection is allowed when the virtual coil is converted from vehicle to vehicleless.

The method according to claim 2, wherein the determining whether the vehicle passes through the virtual coil according to the ratio comprises:

Determining whether a frame interval between the region of interest entering the virtual coil for the first time and the region of interest of the vehicle entering the virtual coil is less than a preset minimum number of frame intervals between the two vehicles;

Determining the interest if the frame interval between the region of interest of the first time entering the virtual coil and the region of interest of the vehicle entering the virtual coil is not less than a preset minimum number of frame intervals between the two vehicles Whether the number of frames occupied by the area through the virtual coil is less than a preset number of frames;

When the number of frames is greater than a preset number of frames, the result of passing the vehicle in the virtual coil is output, otherwise the result of not passing the vehicle in the virtual coil is output.

The method according to claim 1, wherein the detecting the night vehicle according to the region of interest comprises:

5. Method according to claim 4, characterized in that the dispersion is measured by standard deviation.

6. A device for detecting a vehicle at night, comprising:

An acquiring unit, configured to acquire an adjacent video frame;

The first detecting unit is configured to detect the night vehicle according to the region of interest extracted by the extracting unit.

7. The device according to claim 6, further comprising:

a second detecting unit, configured to detect a ratio of a region of interest extracted by the extracting unit to a preset virtual coil, wherein the virtual coil is a quadrangle in a corresponding region of the video image;

a determining unit, configured to determine, according to a ratio obtained by the detecting unit, whether a vehicle passes through the virtual coil;

And an enabling unit, configured to allow the first detecting unit to detect the night vehicle when the determining unit determines that the virtual coil is converted from no vehicle to no vehicle.

The device according to claim 7, wherein the determining unit comprises: a first determining subunit, configured to determine a vehicle that enters the region of interest of the virtual coil for the first time and the vehicle that enters the virtual coil Whether the frame interval between the regions of interest is less than the preset minimum number of frame intervals between the two cars;

a second determining subunit, configured to determine, at the first determining subunit, that a frame interval between a region of interest that first enters the virtual coil and a region of interest of a vehicle that enters the virtual coil is not less than a preset Whether the number of frames occupied by the region of interest passing through the virtual coil is less than a preset number of frames when a minimum number of frame intervals is between the two vehicles;

An output subunit, configured to output a result of passing the vehicle in the virtual coil when the second judging subunit determines that the number of frames is greater than a preset number of frames, otherwise outputting the virtual coil without passing through the vehicle result.

The device according to claim 6, wherein the first detecting unit comprises: a vehicle type acquiring subunit, wherein the region of interest extracted according to the extracting unit is within the preset virtual coil The dispersion of the distribution gets the model of the night vehicle.