WO2018035768A1

WO2018035768A1 - Method for acquiring dimension of candidate frame and device

Info

Publication number: WO2018035768A1
Application number: PCT/CN2016/096598
Authority: WO
Inventors: 覃剑; 肖婷; 王美华
Original assignee: 深圳天珑无线科技有限公司
Priority date: 2016-08-24
Filing date: 2016-08-24
Publication date: 2018-03-01

Abstract

A method for acquiring the dimension of a candidate frame, related to the field of target detection in computer vision and pattern recognition, capable of determining, by means of establishing a Gaussian distribution model, dimension information of a target candidate frame in a video to be detected. The method comprises: acquiring an image to be searched (101); acquiring image information of each raw partition in the image to be searched and a Gaussian distribution function corresponding to each raw partition (102); determining dimension information of a search target in each raw partition on the basis of the image information and the Gaussian distribution function of each raw partition (103); and determining dimension information of a candidate frame in each raw partition on the basis of the dimension information of the search target (104). The technical solution provided is applicable in target detection processes such as pedestrian detection and vehicle detection in scenarios such as static surveillance video and vehicle-mounted surveillance video.

Description

Method and device for acquiring candidate frame scale

Technical field

The invention relates to the field of object detection in computer vision and pattern recognition, and in particular to a method and a device for acquiring candidate frame scales.

Background technique

With the rapid development and wide application of computer image processing technology, the demand for target detection technology has gradually increased. Target detection has become a basic problem in the field of computer vision and pattern recognition, and the determination of the candidate frame size of the detection target is an important preliminary work of the target recognition classification. At present, the existing method for generating a target candidate frame is generally a sliding window search mode, and when the target search is performed, a candidate frame of a plurality of scales is set to search through the entire scan window.

In the process of implementing the present invention, the inventors have found that at least the following problems exist in the prior art:

According to the existing target search method, in the process of searching for the target, the candidate frames of many scales are set to search in the entire scan window, the number of candidate frames is large, the target search time is too long, and the detection rate is low.

Summary of the invention

In view of this, the embodiment of the present invention provides a method and a device for acquiring a candidate frame size, which can determine a candidate frame size according to the scale information of the search target in the region.

In one aspect, an embodiment of the present invention provides a method for obtaining a candidate frame size, where the method includes:

Obtain an image to be searched;

Obtaining image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each original partition;

Determining the scale information of the search target in each original partition according to the image information of each original partition and the Gaussian distribution function;

The candidate frame size information in each of the original partitions is determined based on the scale information of the search target.

On the other hand, an embodiment of the present invention provides a device for acquiring a candidate frame size, and the device includes:

a first acquiring unit, configured to acquire an image to be searched;

a second acquiring unit, configured to acquire image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each original partition;

a first determining unit, configured to determine, according to image information of each original partition and a Gaussian distribution function, scale information of a search target in each original partition;

And a second determining unit, configured to determine candidate frame size information in each original partition according to the scale information of the search target.

A method and a device for acquiring a candidate frame scale according to an embodiment of the present invention, by establishing a Gaussian model by partitioning, the scale information of a specific target in the detection area of each block can be obtained, and the specific target candidate frame can be reasonably set by using the method. The size of the candidate box and the real target have a large coverage, and can achieve a higher detection rate when less specific target candidate frames are set.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the present invention. One of ordinary skill in the art can also obtain other drawings based on these drawings without paying for inventive labor.

FIG. 1 is a flowchart of a method for acquiring candidate frame scales according to an embodiment of the present invention;

2 is a flowchart of another method for acquiring candidate frame sizes according to an embodiment of the present invention;

3 is a flowchart of another method for acquiring candidate frame sizes according to an embodiment of the present invention;

4 is a block diagram of a component of a candidate frame size acquiring apparatus according to an embodiment of the present invention;

FIG. 5 is a block diagram showing the composition of another candidate frame size acquiring apparatus according to an embodiment of the present invention; FIG.

FIG. 6 is a block diagram showing the composition of another candidate frame size acquiring apparatus according to an embodiment of the present invention.

detailed description

For a better understanding of the technical solutions of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

It should be understood that the described embodiments are only a part of the embodiments of the invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

The terms used in the embodiments of the present invention are for the purpose of describing particular embodiments only. It is not intended to limit the invention. The singular forms "a", "the" and "the"

It should be understood that the term "and/or" as used herein is merely an association describing the associated object, indicating that there may be three relationships, for example, A and/or B, which may indicate that A exists separately, while A and B, there are three cases of B alone. In addition, the character "/" in this article generally indicates that the contextual object is an "or" relationship.

Depending on the context, the word "if" as used herein may be interpreted as "when" or "when" or "in response to determining" or "in response to detecting." Similarly, depending on the context, the phrase "if determined" or "if detected (conditions or events stated)" may be interpreted as "when determined" or "in response to determination" or "when detected (stated condition or event) "Time" or "in response to a test (condition or event stated)".

The embodiment of the invention provides a method for acquiring a candidate frame length, which can be applied to a target detection process such as pedestrian detection and vehicle detection in a scene including static monitoring video and vehicle monitoring video, as shown in FIG. 1 . include:

101. Acquire an image to be searched.

The image to be searched refers to all the images to be detected in the target detection process.

102. Acquire image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each original partition.

It should be noted that, in the embodiment of the present invention, for scene monitoring such as static monitoring video and vehicle monitoring video, the size of the target in a certain area of the image approximates a Gaussian normal distribution.

Wherein, each of the original partitions refers to an area that blocks the detection area.

Wherein, the image information of each of the original partitions refers to the scale information of the search target in each block area.

The scale information of the search target in each block refers to the size of the search target, and is also the size of the search target in each frame of the image.

The search target refers to an object to be detected, such as a person, a vehicle, an object, and the like in the target detection process.

Wherein, the Gaussian distribution function is based on a mathematical Gaussian model and is suitable for description The size of the area, etc., conforms to the function of the positive distribution.

103. Determine, according to image information of each original partition and a Gaussian distribution function, scale information of a search target in each original partition.

104. Determine candidate frame size information in each original partition according to the scale information of the search target.

The size information of the candidate frame refers to the size of the candidate frame.

The determining the candidate frame size information in each of the original partitions is a process of adjusting the size of the candidate frame area according to the target scale information based on the sliding window search mode.

A method for acquiring a candidate frame size according to an embodiment of the present invention, by establishing a Gaussian model by partitioning, can obtain the scale information of a specific target in the detection area of each block, and thereby appropriately setting the area of the specific target candidate frame. The size ensures the coverage of the candidate frame to the real target, and can achieve a higher detection rate when less specific target candidate frames are set.

Further, in conjunction with the foregoing method flow, the embodiment of the present invention provides another possible implementation manner. As shown in FIG. 2, before the acquiring the image to be searched, the method further includes:

201. Obtain a original partition in the original image.

Wherein, the original image refers to an n-frame image in the detection area.

Wherein n is an integer greater than zero.

The original partition is divided according to the size and characteristics of the detection area.

The more the number of the original partitions, the more accurately the function can reflect the size of the target within the partition, and the statistical distribution inside each original partition can be considered to be basically the same.

The more the number of the original partitions is, the more complicated the calculation process is. According to the complexity and accuracy of the calculation method, the number of the original partitions is determined according to the size and characteristics of the search area.

202. Collect scale information of the search target in the original partition.

The collecting the scale information of the search target in the original partition refers to collecting the scale size of the search target in each original partition of the n-frame image.

203. Determine a Gaussian distribution function of each original partition according to the scale information of the search target.

The Gaussian distribution function obtains an initial value by performing parameter estimation by collecting scale information of the n-frame image search target.

The Gaussian distribution function is trained in the process of detecting the search target, and dynamically acquires the scale information of the search target in each region during the learning process.

Further, in combination with the foregoing method flow, in another possible implementation manner of the embodiment of the present invention, for step 103, determining, based on the scale information of the search target, determining that the candidate frame size information in each original partition is implemented The following specific process, as shown in Figure 3, includes:

301. Obtain a candidate frame adjustment ratio.

The candidate frame adjustment ratio ranges from (0.6 to 2.6).

302. Determine candidate frame size information in each original partition according to the candidate frame adjustment ratio and the scale information of the search target.

The area size of the candidate frame is determined according to the adjusted scale range on the basis of the target reference scale, and has three candidate frame areas: a first candidate frame area, a second candidate frame area, and a third candidate. Frame area.

Wherein the target reference scale is a scale size determined to be closest to an actual target scale size according to the Gaussian distribution function.

The first candidate frame area refers to the product of the area of the target reference scale and the candidate frame adjustment ratio of 0.6.

The second candidate frame area refers to the product of the area of the target reference scale and the candidate frame adjustment ratio 1.

The third candidate frame area refers to the product of the area of the target reference scale and the candidate frame adjustment ratio of 2.6.

A method for acquiring a candidate frame size according to an embodiment of the present invention, by setting a Gaussian model by partitioning, can obtain a scale information of a specific target in a detection area of each block, and thereby appropriately setting a size of a specific target candidate frame. To ensure the coverage of the candidate frame to the real target, it is possible to achieve a higher detection rate when less specific target candidate frames are set.

An embodiment of the present invention provides a device for acquiring a candidate frame size, which can be used to implement the foregoing method flows. The composition thereof is as shown in FIG. 4, and the device includes:

The first obtaining unit 41 is configured to acquire an image to be searched.

The second obtaining unit 42 is configured to acquire image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each of the original partitions.

a first determining unit 43 configured to use image information and a Gaussian distribution function of each original partition, Determine the scale information of the search target in each raw partition.

The second determining unit 44 is configured to determine candidate frame size information in each original partition according to the scale information of the search target.

Optionally, as shown in FIG. 5, the device further includes:

The third obtaining unit 45 is configured to acquire a raw partition within the original image.

The collecting unit 46 is configured to collect the scale information of the search target in the original partition.

The third determining unit 47 is configured to determine a Gaussian distribution function of each original partition according to the scale information of the search target.

Optionally, as shown in FIG. 6, the second determining unit 44 includes:

The obtaining module 441 is configured to obtain a candidate frame adjustment ratio.

The determining module 442 is configured to determine candidate frame size information in each original partition according to the candidate frame adjustment ratio and the scale information of the search target.

A candidate frame size acquiring apparatus provided by an embodiment of the present invention can obtain a scale information of a specific target in a detection area of each block by establishing a Gaussian model by partitioning, and thereby appropriately setting a size of a specific target candidate frame. To ensure the coverage of the candidate frame to the real target, it is possible to achieve a higher detection rate when less specific target candidate frames are set.

A person skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present invention, it should be understood that the disclosed system, apparatus, and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner. For example, multiple units or components may be combined. Or it can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

The above-described integrated unit implemented in the form of a software functional unit can be stored in a computer readable storage medium. The above software functional unit is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to perform the methods of the various embodiments of the present invention. Part of the steps. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like, which can store program codes. .

The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalents, improvements, etc., which are made within the spirit and principles of the present invention, should be included in the present invention. Within the scope of protection.

Claims

A method for obtaining a candidate frame scale, the method comprising:

Obtain an image to be searched;

Obtaining image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each original partition;

Determining the scale information of the search target in each original partition according to the image information of each original partition and the Gaussian distribution function;

The candidate frame size information in each of the original partitions is determined based on the scale information of the search target.
The method according to claim 1, wherein before the acquiring the image to be searched, the method further comprises:

Obtaining the original partition within the original image;

Collecting scale information of the search target in the original partition;

A Gaussian distribution function of each original partition is determined according to the scale information of the search target.
The method according to claim 2, wherein the determining the candidate frame size information in each of the original partitions according to the scale information of the search target comprises:

Get the candidate frame adjustment ratio;

The candidate frame size information in each of the original partitions is determined according to the candidate frame adjustment ratio and the scale information of the search target.
A candidate frame size acquiring device, wherein the device comprises:

a first acquiring unit, configured to acquire an image to be searched;

a second acquiring unit, configured to acquire image information of each original partition in the image to be searched and a Gaussian distribution function corresponding to each original partition;

a first determining unit, configured to determine, according to image information of each original partition and a Gaussian distribution function, scale information of a search target in each original partition;

And a second determining unit, configured to determine candidate frame size information in each original partition according to the scale information of the search target.
The device according to claim 4, wherein the device further comprises:

a third acquiring unit, configured to acquire a original partition in the original image;

An acquisition unit, configured to collect the scale information of the search target in the original partition;

a third determining unit, configured to determine each original score according to the scale information of the search target The Gaussian distribution function of the region.
The apparatus according to claim 5, wherein the second determining unit comprises:

An obtaining module, configured to obtain a candidate frame adjustment ratio;

And a determining module, configured to determine candidate frame size information in each original partition according to the candidate frame adjustment ratio and the scale information of the search target.