CN109685062A - A kind of object detection method, device, equipment and storage medium - Google Patents
A kind of object detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN109685062A CN109685062A CN201910001609.3A CN201910001609A CN109685062A CN 109685062 A CN109685062 A CN 109685062A CN 201910001609 A CN201910001609 A CN 201910001609A CN 109685062 A CN109685062 A CN 109685062A
- Authority
- CN
- China
- Prior art keywords
- area
- shooting
- target
- region
- camera
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Closed-Circuit Television Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of object detection method, device, equipment and storage mediums.This method comprises: determining that corresponding shooting area is target area or nontarget area according to the picture of two or more shooting areas, the target area is the region that object to be detected is likely to occur；If shooting area is target area, by the corresponding camera of the shooting area and region-wide camera alternately target detection, the high resolution of the corresponding camera of the shooting area is in the resolution ratio of the entire area camera；If shooting area is nontarget area, target detection is carried out by the region-wide camera.Detection and tracking is carried out through the above technical solutions, realizing and neatly treating test object using polyphaser array.
Description
Technical field
The present embodiments relate to technical field of computer vision more particularly to a kind of object detection method, device, equipment
And storage medium.
Background technique
Multiple target vision-based detection and tracking technique are one of the research hotspots in computer vision field, are answered extensively
For fields such as automatic Pilot, pedestrian detection, recognition of face, safety and protection monitorings.Currently, various camera apparatus is in multiple target vision
There is use in detection and tracking technique, wherein polyphaser array has many advantages, due to there are multiple cameras to participate in image
Acquisition can provide bigger information content and then improve the precision of target detection and tracking, but this also results in operation complexity, determines
Plan slowly, inefficiency a series of problems, such as.
In polyphaser array, the shooting image mosaic between each camera is complicated, when target is from a certain camera unit
It disappears in picture, needs accurately to predict that the target will occur in the picture of which camera unit, and carry out camera unit
Switching.When carrying out target identification and detection using the lower camera of resolution ratio, may be unable to satisfy picture resolution ratio and
The demand of target detection precision, and when camera monitoring higher using resolution ratio, it needs to handle a large amount of shooting picture,
Operation higher cost.It is difficult to efficient collaboration, both between multiple camera units in current object detection method, it can not be neatly right
Object to be detected carries out detection and tracking.
Summary of the invention
The present invention provides a kind of object detection method, device, equipment and storage mediums, utilize polyphaser array to realize
It neatly treats test object and carries out detection and tracking.
In a first aspect, the embodiment of the invention provides a kind of object detection methods, comprising:
Determine that corresponding shooting area is target area or non-mesh according to the picture of two or more shooting areas
Region is marked, the target area is the region that object to be detected is likely to occur；
If shooting area is target area, by the corresponding camera of the shooting area and region-wide camera alternately into
Row target detection, the high resolution of the corresponding camera of the shooting area is in the resolution ratio of the region-wide camera；
If shooting area is nontarget area, target detection is carried out by the region-wide camera.
Further, the picture according to two or more shooting areas determines that corresponding shooting area is target
Region or nontarget area, comprising:
Using center shooting area as current region, using the picture of the center shooting area as the picture of current region
Target Recognition Algorithms are inputted, determine that current region is target area or nontarget area, the center shooting area is described
It is located at the shooting area of center in two or more shooting areas；
If the current region is target area, using the picture of the adjacent shooting area of the current region as newly working as
The picture of forefoot area inputs Target Recognition Algorithms；
If the current region is nontarget area, by the picture of the shooting area on the current region diagonal line
Picture as new current region inputs Target Recognition Algorithms, until determining that whole shooting areas are target area or non-targeted
Region.
It is further, described by the corresponding camera of the shooting area and region-wide camera alternately target detection,
Include:
By the monitoring confidence level difference and Bhattacharyya system of the corresponding camera of the shooting area and region-wide camera
Number fusion, obtains fusion value；
If the fusion value is less than or equal to preset threshold, mesh is carried out to the shooting area by the region-wide camera
Mark detection；
If the fusion value be greater than preset threshold, by the corresponding camera of the shooting area to the shooting area into
Row target detection.
It is further, described by the corresponding camera of the shooting area and region-wide camera alternately target detection,
Further include:
The region-wide camera is obtained to the shooting resolution ratio of the object to be detected；
When the shooting resolution ratio is more than or equal to default resolution ratio, then by the region-wide camera to the shooting area
Domain carries out target detection；
When the shooting resolution ratio is less than default resolution ratio, then by the corresponding camera of the shooting area to the bat
It takes the photograph region and carries out target detection.
Further, if being nontarget area in the shooting area, target is carried out by the region-wide camera
After detection, further includes:
The shooting picture of the region-wide camera is monitored using background difference algorithm；
If detecting object to be detected in the shooting picture, again according to the picture of two or more shooting areas
Face determines that corresponding shooting area is target area or nontarget area.
Further, the method also includes:
When object to be detected enter the shooting area visual angle overlapping region when, by script to the object to be detected into
The corresponding camera of the shooting area of row target detection provides the probability distribution of the object's position to be detected；
According to the probability distribution
By the object to be detected possibly into shooting area be ranked up from high to low；
According to ranking results successively by it is described possibly into shooting area be input to Target Recognition Algorithms, until determine institute
State the shooting area that object to be detected enters.
Further, the method also includes:
It is spaced after predetermined period, corresponding shooting area is determined according to the picture of two or more shooting areas again
Domain is target area or nontarget area.
Second aspect, the embodiment of the invention provides a kind of object detecting devices, comprising:
First object area determination module determines corresponding bat for the picture according to two or more shooting areas
Taking the photograph region is target area or nontarget area, and the target area is the region that object to be detected is likely to occur；
Alternately into detection module, if being target area for shooting area, pass through the corresponding phase of the shooting area
Machine and region-wide camera alternately target detection, the high resolution of the corresponding camera of the shooting area is in the region-wide phase
The resolution ratio of machine；
Region-wide shooting module is carried out if being nontarget area for shooting area by the region-wide camera
Target detection.
The third aspect, the embodiment of the invention provides a kind of equipment, comprising:
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes object detection method as described in relation to the first aspect.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program realizes object detection method as described in relation to the first aspect when the program is executed by processor.
The embodiment of the invention provides a kind of object detection method, device, equipment and storage mediums.This method comprises: root
According to the picture of two or more shooting areas determine corresponding shooting area be target area or nontarget area, it is described
Target area is the region that object to be detected is likely to occur；If shooting area is target area, pass through the shooting area
Corresponding camera and region-wide camera alternately target detection, the high resolution of the corresponding camera of the shooting area is in described
The resolution ratio of region-wide camera；If shooting area is nontarget area, target detection is carried out by the region-wide camera.
Detection and tracking is carried out through the above technical solutions, realizing and neatly treating test object using polyphaser array.
Detailed description of the invention
Fig. 1 is a kind of flow chart for object detection method that the embodiment of the present invention one provides；
Fig. 2 is the schematic diagram for the polyphaser array that the embodiment of the present invention one provides；
Fig. 3 is the schematic diagram for the polyphaser array shooting area that the embodiment of the present invention one provides；
Fig. 4 is a kind of flow chart of object detection method provided by Embodiment 2 of the present invention；
Fig. 5 A is the schematic diagram of current region provided by Embodiment 2 of the present invention；
Fig. 5 B is the schematic diagram of the adjacent shooting area of current region provided by Embodiment 2 of the present invention；
Fig. 5 C is the schematic diagram of the shooting area on current region diagonal line provided by Embodiment 2 of the present invention；
Fig. 6 is the schematic diagram that object detection method provided by Embodiment 2 of the present invention is realized；
Fig. 7 is a kind of structural schematic diagram for object detecting device that the embodiment of the present invention three provides；
Fig. 8 is a kind of hardware structural diagram for equipment that the embodiment of the present invention four provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is a kind of flow chart for object detection method that the embodiment of the present invention one provides, and the present embodiment is applicable to lead to
Excessive camera array treats the case where test object carries out target detection, specifically, can be applied to military tracking, civilian site chases after
Track/monitoring, automatic Pilot target identification, sports relay etc. are shot by polyphaser array and need to be examined from shooting picture
Survey the various fields of target.The object detection method can be executed by object detecting device, which can pass through
The mode of software and/or hardware is realized, and is integrated in a device.Further, equipment includes but is not limited to: industrial personal computer, platform
Formula computer and integrated intelligent electronic equipment.
Fig. 2 is the schematic diagram for the polyphaser array that the embodiment of the present invention one provides.Polyphaser battle array described in the present embodiment
Column refer to multifocal away from camera array, i.e., to be made of the focal length camera of a wide angle camera and multiple stationary arrangements polyphaser battle array
Column, wherein wide angle camera refers to that focal length is shorter, the biggish camera in visual angle, the shooting area bat that is wider, but taking of covering
Take the photograph that scenery in picture is relatively small, resolution ratio is relatively low；Focal length camera refers to that focal length is longer, the lesser camera in visual angle,
The shooting area of covering is smaller, has higher resolution ratio for the shooting picture of a small range scenery.As shown in Fig. 2, multiphase
It include a wide angle camera 100 in machine array, i.e., region-wide camera, position coordinates are denoted as C (0,0), and are arranged in three rows six
Be distributed in the focal length camera 101 (1 is only indicated in Fig. 2) of 100 two sides of wide angle camera, i.e., the corresponding camera of each shooting area,
The position coordinates of these wide angle cameras are denoted as C (1,1) to C (3,6) respectively.
Fig. 3 is the schematic diagram for the polyphaser array shooting area that the embodiment of the present invention one provides.As shown in figure 3, wide-angle phase
Machine 100 covers shooting area (the overstriking black surround in such as Fig. 3 of whole focal length cameras for region-wide shooting, coverage
It is shown), this coverage is interior to have multiple shooting areas, and in the distribution that three rows six arrange, each shooting area corresponds respectively to one
A focal length camera, and the shooting area of the focal length camera of arranged adjacent has and partially overlaps (such as the part institute that square is overlapped in Fig. 3
Show), the area of intersection can determine in the deployment of polyphaser array and calibration process according to actual needs.
It should be noted that the present embodiment to the determination of target area, to target area alternately target detection and right
The target detection of nontarget area is described in detail, and (more) target recognition and tracking algorithm is not construed as limiting, and
And polyphaser array deployment when have already passed through visual angle, in terms of calibration.
With reference to Fig. 1, which specifically comprises the following steps:
S110, determined according to the picture of two or more shooting areas corresponding shooting area be target area or
Nontarget area, the target area are the regions that object to be detected is likely to occur.
Specifically, it is first determined each shooting area is target area or nontarget area.Polyphaser array is covered
Image in whole region is divided into two classes, and target area refers to region (such as road, square for being possible to occur object to be detected
Deng), nontarget area refers to the region (such as woods, sky) that object to be detected can not occur, usual for target area
It needs to carry out high-precision real-time monitoring, and operation can be used cost is relatively low algorithm for nontarget area (such as background difference is calculated
Method) it is monitored.Object to be detected can be pedestrian, vehicle, traffic mark etc., by each shooting area (three rows as shown in Figure 3
Six column and with intersection each square representated by region) picture sequentially input to Target Recognition Algorithms to carry out image
Detection and classification, determine whether it is the target area for being likely to occur object to be detected.
Whether S120, shooting area is are target area, if so, thening follow the steps S130；If not, thening follow the steps
S140。
Specifically, step S110 has successively determined whether all shooting areas are target area, for different regions, adopt
Target detection is carried out with different camera or algorithm, if it is target area, thens follow the steps S130；If it is not, then executing step
Rapid S140.
S130, pass through the corresponding camera of the shooting area and region-wide camera alternately target detection, the shooting
The high resolution of the corresponding camera in region is in the resolution ratio of the region-wide camera.
Specifically, for target area, when there is object to be detected, need accurately and rapidly detection and tracking it is to be checked
Object is surveyed, the detailed information of shooting picture is captured.This implementation is illustratively set for target area, and default setting is utilization pair
The camera answered carries out target detection, to provide clearer shooting picture；For nontarget area, default setting is to utilize the whole district
Domain camera carries out target detection, to reduce testing cost.With the movement of object to be detected, entered according to object to be detected or from
The direction for opening current goal region, can detect which shooting area object to be detected will be transferred to, thus to target area into
Row prediction and update, guarantee to track object to be detected in time.On this basis, corresponding using shooting area to target area
Camera and region-wide camera alternately target detection.Illustratively, when the shooting area for being originally used for nontarget area becomes
When target area, the corresponding camera in the target area is switched to by region-wide camera and carries out target detection；For target area, such as
Fruit its shoot resolution requirement it is not high when, be switched to region-wide camera and carry out target detection, such as object outline to be detected is clear
And background is more single, and a vehicle is monitored such as on intermountain highway, then for the target area where vehicle, can be switched to complete
Area camera carries out target detection；And when needing to obtain the details of vehicle, such as to check license plate, differentiate vehicle model,
When recognizing car owner, the corresponding camera in target area where being switched to the vehicle carries out target detection, to provide higher resolution
Rate and detection accuracy.
S140, target detection is carried out by the region-wide camera.
Specifically, carry out target detection by region-wide camera if shooting area is nontarget area, with provide compared with
Whole shooting picture in wide coverage avoids the splicing and calibration of the image of multiple shooting areas；Simultaneously using fortune
Whether the lower algorithm of calculation amount (background difference algorithm) monitoring there is object to be detected in nontarget area, if there is to be detected
Object then updates target area, and using the corresponding camera of shooting area and region-wide camera according to actual needs to target area
Domain alternately target detection.
It should be noted that the shooting area has multiple, it is corresponding that each shooting area both corresponds to shooting area
Camera (focal length camera), the high resolution of the corresponding camera of the shooting area is in the resolution ratio of region-wide camera.It is region-wide
The coverage of camera covers the shooting area of the corresponding camera of whole shooting areas, do not need in shooting process calibration or
Stitching image can obtain global information, without carrying out image procossing to a large amount of shooting picture, can reduce operand；And
The corresponding camera of shooting area can provide the details of more shooting pictures if necessary, and it is careful to carry out to more image pixels
Calculation process, to improve the precision of detection and tracking object to be detected.Therefore, the object detection method of the present embodiment passes through more
Camera array is respectively adopted different camera to target area and nontarget area and carries out target detection, can take into account detection accuracy with
Testing cost.
A kind of object detection method that the embodiment of the present invention one provides, according to the picture of two or more shooting areas
Determine that corresponding shooting area is target area or nontarget area；If shooting area is target area, by described
The corresponding camera of shooting area and region-wide camera alternately target detection；If shooting area is nontarget area, lead to
It crosses the region-wide camera and carries out target detection.Through the above technical solutions, using two different camera roots to target area
According to actual demand alternately target detection, operand lower region-wide phase simple using image procossing to nontarget area
Machine and algorithm of target detection are monitored, and comprehensively consider testing cost and detection accuracy, are realized flexible using polyphaser array
It treats test object and carries out detection and tracking in ground.
Embodiment two
Fig. 4 is a kind of flow chart of object detection method provided by Embodiment 2 of the present invention, and the present embodiment is in above-mentioned reality
On the basis of applying example, specifically optimized, determination and update to target area and nontarget area, the corresponding phase of shooting area
The alternately target detection of machine and region-wide camera is illustrated, and the technical detail of detailed description can not join in the present embodiment
See above-mentioned any embodiment.
Specifically, this method specifically comprises the following steps: with reference to Fig. 4
S210, using center shooting area as current region, using the picture of the center shooting area as current region
Picture input Target Recognition Algorithms, the center shooting area is to be located at center in described two or more than two shooting areas
The shooting area of position.
Specifically, the startup stage initial in polyphaser array, if using the picture of each shooting area as target area
Domain, and whole shooting pictures is input to algorithm of target detection, it can cause greatly to calculate to image procossing and target detection
Burden, substantially reduces detection efficiency, and if the shooting picture of region-wide camera is only input to algorithm of target detection, due to dividing
The limitation of resolution, may result in can not clearly monitor object to be detected.Starting of the present embodiment in polyphaser array
Stage successively activates the corresponding camera of each shooting area according to certain rule, the corresponding camera of each shooting area is monitored
Picture is input to algorithm of target detection successively in a certain order to judge whether each shooting area is target area, can be quick
It determines target area, improves detection efficiency.
Specifically, whether the center shooting area of first determining whether is target area, the center shooting area refers to multiple
It is located at the shooting area of center in shooting area.Fig. 5 A is shooting area current region provided by Embodiment 2 of the present invention
Schematic diagram.With reference to Fig. 5 A, the present embodiment illustratively uses 1 region-wide camera and the corresponding camera of 18 shooting areas, entirely
The coverage of area camera covers the shooting area of the corresponding camera of all shooting areas, and 18 shooting areas can be divided into three
Six column of row, each shooting area correspond respectively to the corresponding camera of 1 shooting area.By 18 shooting area geometry equal part (etc.
It is divided into left and right two parts), each equal subregions are in the shooting area of center, i.e. C (2,2) and the corresponding shooting of C (2,5)
Region (in Fig. 5 A shown in shade) is used as current region, and the picture of current region is inputted Target Recognition Algorithms, current to determine
Whether region is target area.
Whether S220, current region are target area, if so, thening follow the steps S230；If not, thening follow the steps
S240。
S230, target identification is inputted using the picture of the adjacent shooting area of the current region as the picture of new current region
Algorithm.
Specifically, if current region is target area, continue using the picture of the adjacent shooting area of current region as
Whether the picture of new current region inputs Target Recognition Algorithms, be target area with the adjacent shooting area of determination.
Fig. 5 B is the schematic diagram of the adjacent shooting area of current region provided by Embodiment 2 of the present invention.It is calculated according to target following
The direction that object to be detected enters and leaves current region (target area) can be obtained in method, thus by that this side up, adjacent
The picture of shooting area (in Fig. 5 B shown in shade) inputs Target Recognition Algorithms.It should be noted that the present embodiment is illustratively
Adjacent shooting area is limited to the shooting area on the horizontal or vertical direction of current region, since current region is target area
Domain, so the probability that object to be detected occurs in shooting area adjacent on horizontal or vertical direction is bigger, be conducive to quickly positioning to
Test object.
S240, mesh is inputted using the picture of the shooting area on the current region diagonal line as the picture of new current region
Mark recognizer.
Specifically, if current region is nontarget area, by the picture of the shooting area on current region diagonal line
Target Recognition Algorithms are inputted, to determine whether the shooting area on diagonal line is target area.
Fig. 5 C is the schematic diagram of the shooting area on current region diagonal line provided by Embodiment 2 of the present invention.If current
Region is the nontarget area for being not in object to be detected, then by its diagonally adjacent shooting area (shade in Fig. 5 C
It is shown) picture input Target Recognition Algorithms.It should be noted that current region is not in object to be detected, then it is horizontal
Or on vertical direction adjacent shooting area occur object to be detected probability it is relatively small, at this time first to the shooting on diagonal line
Region is judged, search radius can be expanded, conducive to the target area that object to be detected is likely to occur quickly is positioned.
S250, whether have determined that whole shooting areas are target area or nontarget area, if so, thening follow the steps
S260；If not, thening follow the steps S220.
Further, using the shooting area on adjacent shooting area or diagonal line as new current region, judge
After whether new current region is target area, continues to expand search range, continued to select new shooting area with same method
Judged, until all shooting areas are all targeted region or nontarget area.If it is determined that all shootings
Region is target area or nontarget area, thens follow the steps S260, carries out mesh to target area and nontarget area respectively
Mark detection, if having not determined whole shooting areas, repeats step S220-S250, judges new current region.
S260, target area alternately target is examined by the corresponding camera of the shooting area and region-wide camera
It surveys, target detection is carried out to nontarget area by the region-wide camera.
Specifically, then passing through shooting area if it is determined that whole shooting area is target area or nontarget area
Corresponding camera and region-wide camera carry out alternately target detection to target area, by region-wide camera to non-target area
Domain carries out target detection.
Further, by the corresponding camera of the shooting area and region-wide camera to target area alternately target
Detection, comprising:
By the monitoring confidence level difference and Bhattacharyya system of the corresponding camera of the shooting area and region-wide camera
Number fusion, obtains fusion value；
If the fusion value is less than or equal to preset threshold, mesh is carried out to the shooting area by the region-wide camera
Mark detection；
If the fusion value be greater than preset threshold, by the corresponding camera of the shooting area to the shooting area into
Row target detection.
Specifically, two kinds of switchings between camera are related to for the target detection of target area and nontarget area,
Include: when the picture of region-wide camera being input to Target Recognition Algorithms still guaranteeing the precision of target detection and tracking enough,
It is shot being switched to the shooting of target area using region-wide camera, or when target area becomes nontarget area, switching
To be shot using region-wide camera；When shooting purpose changes, needs analysis or magnified picture details, requires higher shooting
When resolution ratio, target detection is carried out by the corresponding camera of the shooting area where is switched to the shooting of nontarget area.
Specifically, needing to assess when to entire area for being switched to the case where carrying out target detection using region-wide camera
When the low resolution picture that camera takes is handled, whether algorithm of target detection is capable of providing enough detection and tracking essences
Degree, if enough detection and tracking precision can be provided, will be transferred to region-wide camera to the shooting of target area, by the whole district
Domain camera is detected and is tracked to the shooting picture input Target Recognition Algorithms of this target area.For example, to intermountain public affairs
Vehicle in road carries out detection and tracking, and the clear-cut and background of object to be detected is more single, utilizes region-wide phase at this time
The low resolution picture that machine takes also guarantees the position of accurate detection and tracking vehicle enough, at this point for vehicle institute
Target area, can be switched to and carry out target detection using region-wide camera, obtained without using the corresponding camera of shooting area
To high-resolution shooting picture and splicing and the picture for calibrating each shooting area.Specific appraisal procedure is combined to two types
The camera view of type carries out the difference of the confidence level of target detection and the heterochromia of Bhattacharyya coefficient description.
Specifically, the histogram of the corresponding camera shooting picture of shooting area is denoted as p, the histogram of region-wide camera shooting picture
Seal is q, it is assumed that two histograms are m grades, then Bhattacharyya coefficient D (p, q) is
Remember and the confidence level of target detection and the shooting picture to the corresponding camera of shooting area are carried out to the shooting picture of region-wide camera
The difference for carrying out the confidence level of target detection is Dconfidence, then appraisal procedure is embodied as:
If meeting: α D (p, q)+β Dconfidence≤Thandover, then region-wide phase will be transferred to the shooting of this target area
Machine；If being unsatisfactory for: α D (p, q)+β Dconfidence≤Thandover, then shooting area pair will be transferred to the shooting of this target area
The camera answered.Wherein, confidence level refers to that Target Recognition Algorithms detect the confidence level of object to be detected from shooting picture, and α and β are
The weight rule of thumb set, α D (p, q)+β DconfidenceValue is merged for confidence level difference and Bhattacharyya coefficient,
ThandoverFor the preset threshold set according to experience, α, β, ThandoverIt can set according to actual needs.
It is further, described by the corresponding camera of the shooting area and region-wide camera alternately target detection,
Further include:
The region-wide camera is obtained to the shooting resolution ratio of the object to be detected；
When the shooting resolution ratio is more than or equal to default resolution ratio, then by the region-wide camera to the shooting area
Domain carries out target detection；
When the shooting resolution ratio is less than default resolution ratio, then by the corresponding camera of the shooting area to the bat
It takes the photograph region and carries out target detection.
Specifically, where being switched to region-wide camera for nontarget area the case where shooting area corresponding camera,
Usually for the shooting higher shooting task of resolution requirement, such as not only need to monitor vehicle also and to detect license plate, not only
It monitors pedestrian also and to carry out Face datection etc. and cut behind the target area where region-wide camera monitors object to be detected
It gives the corresponding camera in the target area and carries out target detection.
Illustratively, it can default and first pass through region-wide camera the object to be detected is shot, obtain region-wide phase
The shooting resolution ratio of machine, or shooting resolution ratio is calculated according to the performance parameter of region-wide camera, and by itself and default resolution ratio
It is compared, default resolution ratio can be set according to the actual demand of different task.For example, region-wide camera is to described to be detected right
The shooting resolution ratio of elephant is 960P, and for vehicle tracking task, presetting resolution ratio is 960P, then region-wide camera can meet need
It asks, vehicle tracking is carried out to target area by region-wide camera；But for car plate detection task, default resolution ratio is
1080P then needs to switch to the corresponding camera of shooting area and carries out target detection to the target area.It should be noted that
Shooting resolution ratio and default resolution ratio can also be expressed as form, such as super clear, high definition of grade etc..
Further, if being nontarget area in the shooting area, target is carried out by the region-wide camera
After detection, further includes:
The shooting picture of the region-wide camera is monitored using background difference algorithm；
If detecting object to be detected in the shooting picture, again according to the picture of two or more shooting areas
Face determines that corresponding shooting area is target area or nontarget area.
Specifically, carrying out target detection using region-wide camera, and lower-cost using calculating for nontarget area
Background difference algorithm is monitored the picture of region-wide shooting line machine, wherein background difference algorithm is to introduce normal distribution
Background difference algorithm, is embodied as:
Wherein, μt=ρ It+(1-ρ)μt-1,D=| (It-μt) |, k is empirical value, can basis
Actual demand setting, the present embodiment set k=2.5, and initial value when t=0 isμ0、σ0It is set according to experience
Value, ItFor the intensity of the pixel of t moment, μt、σtFor the parameter of normal distribution.It can determine whether the t moment pixel according to above-mentioned formula
Point is prospect (foreground) or background (background), with detect in nontarget area whether occur it is to be detected right
As.
Further, after determining that whole shooting areas are target area or nontarget area, to target area and
Constantly update the accuracy to guarantee algorithm of target detection in nontarget area.For the picture that a certain moment takes, if intensity
Meet for the pixel p (x, y) of I (x, y):
Then judge that pixel p (x, y) belongs to target area, if not satisfied, then pixel p (x, y) belong to it is non-targeted
Region, wherein rneighborFor empirical value, expression be may determine that as the neighborhood distance of same object to be detected, with object to be detected
Size be positively correlated, for example, object to be detected be vehicle, rneighborBe 20, indicate exist can be used as differentiate radius, less than 20
R, to differentiate the pixel whether be object to be detected target area；Object to be detected is behaved, rneighborIt is 10, indicates
Differentiate radius, r less than 10 in the presence of can be used as, to judge the pixel whether be object to be detected target area,
PtargetIndicate the set of the pixel of the target area of object to be detected.
Further, when object to be detected enter the shooting area visual angle overlapping region when, by script to it is described to
The corresponding camera of shooting area that test object carries out target detection provides the probability distribution of the object's position to be detected；
According to the probability distribution by the object to be detected possibly into shooting area be ranked up from high to low；
According to ranking results successively by it is described possibly into shooting area be input to Target Recognition Algorithms, until determine institute
State the shooting area that object to be detected enters.
Specifically, having intersection between adjacent shooting area, when object to be detected enters intersection, by originally
The corresponding camera of the shooting area being monitored to it or region-wide camera provide the probability distribution of its position, to be detected to predict
Object will enter which shooting area, and according to the sequence of probability distribution from high to low to possibly into shooting area carry out
Sequence, is successively input to Target Recognition Algorithms for shooting area according to ranking results, until determining which object to be detected enters
Shooting area, to realize the real-time tracing for treating test object.Optionally, object to be detected can also be predicted according to Gaussian Profile
The target area that will enter, after determining the shooting area that object to be detected enters, using the corresponding camera in the region or
Region-wide camera treats test object and carries out alternately detection.
After S270, interval predetermined period, determined again according to the picture of two or more shooting areas corresponding
Shooting area is target area or nontarget area.
Specifically, whenever detecting object to be detected in nontarget area using background difference algorithm, or at interval of pre-
If after cycle T, redefining all shooting areas is target area or nontarget area, to avoid due to object to be detected
The movement of position, the error of nontarget area background difference algorithm lead to not promptly and accurately detect object to be detected, again
Determining method is identical as step S210-S250.Wherein, predetermined period T can be set according to actual needs, can be for definite value (such as
10 minutes), or the value of variation.The present embodiment preferably sets predetermined period T according to iteration the reason of leading to end cycle
Variation or resetting.Specifically, predetermined period T is reset to initial value T if occurring object to be monitored in nontarget area0
(such as T0It is 10 minutes), otherwise predetermined period T is target area or nontarget area with all shooting areas are redefined
The increase of number and increase, i.e. the first minor tick T0After redefine target area and nontarget area, later be spaced (1+a) T0
After redefine target area and nontarget area, followed by be spaced (1+2a) T0After redefine target area and non-target area
Domain, and so on.Predetermined period T is embodied as:
Wherein, a is the growth rate of predetermined period, T0It can rule of thumb with a or actual demand setting, n indicate the last
Reset to initial value T0Later, nontarget area does not occur the quantity of the continuous cycles of object to be detected.
It should be noted that the present embodiment dynamically shoots all shooting areas using polyphaser array in real time
And target detection, fully consider the position movement and the update of target area, nontarget area of object to be detected, and drop
The data volume of the low shooting picture for being input to algorithm of target detection, also reduced while guaranteeing detection accuracy testing cost,
Improve detection efficiency.
On the basis of the above embodiments, Fig. 6 is the signal that object detection method provided by Embodiment 2 of the present invention is realized
Figure.As shown in fig. 6, the object detection method of the present embodiment is broadly divided into determining target area, updates target area, utilizes background
Difference algorithm monitoring nontarget area changes to target area alternately target detection, predetermined period using different cameras
Five parts of generation and resetting, this five part circulations are executed to realize real-time, continuous shooting and detection.Wherein, target area
Update include being updated according to probability distribution, in nontarget area detecting that object to be detected then updates or be spaced predetermined period
It is updated, any time being likely to occur in object detection method, the present embodiment is suitable with the execution of other parts to it
Sequence is without limitation.
A kind of object detection method provided by Embodiment 2 of the present invention, optimizes on the basis of the above embodiments, leads to
It crosses and successively activates the corresponding camera of each shooting area according to certain rule, the picture that the corresponding camera of each shooting area is monitored
It is sequentially inputted to algorithm of target detection, quickly position object to be detected and determines target area；By respectively to target area and
Nontarget area selects suitable camera to carry out target detection, has taken into account detection accuracy and testing cost；Pass through assessment target inspection
The detection and tracking precision for the low resolution picture that method of determining and calculating takes region-wide camera, by the shooting area for target area
The corresponding camera in domain is switched to region-wide camera, and the calculation amount of image procossing is reduced while guaranteeing detection accuracy, reduces
Testing cost；The shooting of region-wide camera is switched to corresponding camera according to whether shooting resolution ratio is higher than default resolution ratio,
Conducive to detection accuracy is improved, the efficient collaboration, both of polyphaser array is realized；By updating target area, non-according to probability distribution
After target area detects object to be detected or interval predetermined period, again according to the picture of two or more shooting areas
Face determines that corresponding shooting area is target area or nontarget area, realizes the dynamic of target area and nontarget area
It updates.
Embodiment three
Fig. 7 is a kind of structure chart for object detecting device that the embodiment of the present invention three provides.Target provided in this embodiment
Detection device includes:
First object area determination module 310 is corresponded to for being determined according to the picture of two or more shooting areas
Shooting area be target area or nontarget area, the target area is the region that object to be detected is likely to occur；
Alternately detection module 320 passes through the corresponding phase of the shooting area if being target area for shooting area
Machine and region-wide camera alternately target detection, the high resolution of the corresponding camera of the shooting area is in the region-wide phase
The resolution ratio of machine；
Region-wide detection module 330, if being nontarget area for shooting area, by the region-wide camera into
Row target detection.
A kind of object detecting device that the embodiment of the present invention three provides, by first object area determination module according to two
Or the picture of more than two shooting areas determines that corresponding shooting area is target area or nontarget area；If shooting area
Domain is target area, then utilizes the corresponding camera of the shooting area and region-wide camera alternately by alternately detection module
Target detection；If shooting area is nontarget area, target is carried out using the region-wide camera by region-wide module
Detection.Testing cost and detection accuracy are comprehensively considered through the above technical solutions, realizing, and are carried out using polyphaser array comprehensive
Target detection.
On the basis of the above embodiments, the target area determining module 310, comprising:
Current region determination unit is used for using center shooting area as current region, by the center shooting area
Picture as current region picture input Target Recognition Algorithms, determine current region be target area or nontarget area,
The center shooting area is the shooting area for being located at center in described two or more than two shooting areas；
Adjacent shooting area activates unit, if being target area for the current region, by the current region
The picture of adjacent shooting area inputs Target Recognition Algorithms as the picture of new current region；
Diagonal line shooting area activates unit, will be described current if being nontarget area for the current region
The picture of shooting area on the diagonal line of region inputs Target Recognition Algorithms as the picture of new current region, until determining all
Shooting area is target area or nontarget area.
On the basis of the above embodiments, the alternately detection module 320 further include:
Integrated unit, for by the monitoring confidence level difference of the corresponding camera of the shooting area and region-wide camera with
The fusion of Bhattacharyya coefficient, obtains fusion value；
Region-wide camera activation unit passes through the entire area if being less than or equal to preset threshold for the fusion value
Camera carries out target detection to the shooting area；
First area camera activation unit passes through the shooting area if being greater than preset threshold for the fusion value
Corresponding camera carries out target detection to the shooting area.
Further, the alternately detection module 320 further include:
Resolution ratio acquiring unit is shot, the shooting of the object to be detected is differentiated for obtaining the region-wide camera
Rate；
Second area camera activation unit, when the shooting resolution ratio is more than or equal to default resolution ratio, then by described
Region-wide camera carries out target detection to the shooting area；
Third area camera activates unit, when the shooting resolution ratio is less than default resolution ratio, then passes through the shooting
The corresponding camera in region carries out target detection to the shooting area.
Further, described device further include:
Background difference monitoring modular, for monitoring the shooting picture of the region-wide camera using background difference algorithm；
Second target area determining module, if for detecting object to be detected in the shooting picture, basis again
The picture of two or more shooting areas determines that corresponding shooting area is target area or nontarget area.
Further, described device further include:
Probability distribution computing module, when for entering the visual angle overlapping region of the shooting area when object to be detected, by
Originally the corresponding camera of shooting area for carrying out target detection to the object to be detected provides the object's position to be detected
Probability distribution；
Sorting module, for according to the probability distribution by the object to be detected possibly into shooting area by height to
It is low to be ranked up；
Shooting area determining module, for according to ranking results successively by it is described possibly into shooting area be input to mesh
Recognizer is marked, until determining the shooting area that the object to be detected enters.
Further, the second target area determining module is also used to:
It is spaced after predetermined period, corresponding shooting area is determined according to the picture of two or more shooting areas again
Domain is target area or nontarget area.
The interface arrangement that the embodiment of the present invention three provides can be used for executing the target detection that above-mentioned any embodiment provides
Method has corresponding function and beneficial effect.
Example IV
Fig. 8 is a kind of hardware structural diagram for equipment that the embodiment of the present invention four provides.As shown in figure 8, the present embodiment
A kind of equipment provided, comprising: processor 410 and storage device 420.Processor in the equipment can be one or more,
In Fig. 8 by taking a processor 410 as an example, processor 410 and storage device 420 in the equipment can by bus or other
Mode connects, in Fig. 8 for being connected by bus.
One or more of programs are executed by one or more of processors 410, so that one or more of places
Reason device realizes any object detection method in above-described embodiment.
Storage device 420 in the equipment is used as a kind of computer readable storage medium, can be used for storing one or more
Program, described program can be software program, computer executable program and module, such as target detection in the embodiment of the present invention
Corresponding program instruction/the module of method is (for example, the module in attached object detecting device shown in Fig. 7, comprising: first object area
Domain determining module 310, alternately detection module 320 and region-wide detection module 330).Processor 410 is stored in storage by operation
Software program, instruction and module in device 420, thereby executing the various function application and data processing of equipment, i.e., in fact
Object detection method in existing above method embodiment.
Storage device 420 mainly includes storing program area and storage data area, wherein storing program area can store operation system
Application program needed for system, at least one function；Storage data area, which can be stored, uses created data etc. (such as according to equipment
Picture, algorithm of target detection correlative code of shooting area in above-described embodiment etc.).In addition, storage device 420 may include
High-speed random access memory can also include nonvolatile memory, for example, at least disk memory, a flash memories
Part or other non-volatile solid state memory parts.In some instances, storage device 420 can further comprise relative to processing
The remotely located memory of device 410, these remote memories can pass through network connection to equipment.The example of above-mentioned network includes
But be not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
Also, when one or more program included in above equipment is held by one or more of processors 410
When row, program is proceeded as follows:
Determine that corresponding shooting area is target area or non-mesh according to the picture of two or more shooting areas
Region is marked, the target area is the region that object to be detected is likely to occur；
If shooting area is target area, by the corresponding camera of the shooting area and region-wide camera alternately into
Row target detection, the high resolution of the corresponding camera of the shooting area is in the resolution ratio of the region-wide camera；
If shooting area is nontarget area, target detection is carried out by the region-wide camera.
The object detection method that the equipment and above-described embodiment that the present embodiment proposes propose belongs to same inventive concept, does not exist
The technical detail of detailed description can be found in above-mentioned any embodiment in the present embodiment, and the present embodiment has and performance objective inspection
The identical beneficial effect of survey method.
On the basis of the above embodiments, the present embodiment also provides a kind of computer readable storage medium, is stored thereon with
Computer program realizes the target detection side in the above-mentioned any embodiment of the present invention when program is executed by object detecting device
Method, this method comprises:
Determine that corresponding shooting area is target area or non-mesh according to the picture of two or more shooting areas
Region is marked, the target area is the region that object to be detected is likely to occur；
If shooting area is target area, by the corresponding camera of the shooting area and region-wide camera alternately into
Row target detection, the high resolution of the corresponding camera of the shooting area is in the resolution ratio of the region-wide camera；
If shooting area is nontarget area, target detection is carried out by the region-wide camera.
Certainly, a kind of storage medium comprising computer executable instructions, computer provided by the embodiment of the present invention
The object detection method operation that executable instruction is not limited to the described above, can also be performed provided by any embodiment of the invention
Relevant operation in object detection method, and have corresponding function and beneficial effect.
By the description above with respect to embodiment, it is apparent to those skilled in the art that, the present invention
It can be realized by software and required common hardware, naturally it is also possible to which by hardware realization, but in many cases, the former is more
Good embodiment.Based on this understanding, technical solution of the present invention substantially in other words contributes to the prior art
Part can be embodied in the form of software products, which can store in computer readable storage medium
In, floppy disk, read-only memory (Read-Only Memory, ROM), random access memory (Random such as computer
Access Memory, RAM), flash memory (FLASH), hard disk or CD etc., including some instructions are with so that a computer is set
Standby (can be personal computer, server or the network equipment etc.) executes target detection described in each embodiment of the present invention
Method.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of object detection method characterized by comprising
Determine that corresponding shooting area is target area or non-target area according to the picture of two or more shooting areas
Domain, the target area are the regions that object to be detected is likely to occur；
If shooting area is target area, pass through the corresponding camera of the shooting area and region-wide camera alternately mesh
Mark detection, the high resolution of the corresponding camera of the shooting area is in the resolution ratio of the region-wide camera；
If shooting area is nontarget area, target detection is carried out by the region-wide camera.
2. the method according to claim 1, wherein the picture according to two or more shooting areas
Determine that corresponding shooting area is target area or nontarget area, comprising:
Using center shooting area as current region, inputted the picture of the center shooting area as the picture of current region
Target Recognition Algorithms determine that current region is target area or nontarget area, and the center shooting area is described two
Or it is located at the shooting area of center in more than two shooting areas；
If the current region is target area, work as proparea using the picture of the adjacent shooting area of the current region as new
The picture in domain inputs Target Recognition Algorithms；
If the current region be nontarget area, using the picture of the shooting area on the current region diagonal line as
The picture of new current region inputs Target Recognition Algorithms, until determining that whole shooting areas are target area or non-target area
Domain.
3. the method according to claim 1, wherein described pass through the corresponding camera of the shooting area and the whole district
Domain camera alternately target detection, comprising:
The monitoring confidence level difference of the corresponding camera of the shooting area and region-wide camera is melted with Bhattacharyya coefficient
It closes, obtains fusion value；
If the fusion value is less than or equal to preset threshold, target inspection is carried out to the shooting area by the region-wide camera
It surveys；
If the fusion value is greater than preset threshold, mesh is carried out to the shooting area by the corresponding camera of the shooting area
Mark detection.
4. according to the method described in claim 3, it is characterized in that, described pass through the corresponding camera of the shooting area and the whole district
Domain camera alternately target detection, further includes:
The region-wide camera is obtained to the shooting resolution ratio of the object to be detected；
When the shooting resolution ratio is more than or equal to default resolution ratio, then by the region-wide camera to the shooting area into
Row target detection；
When the shooting resolution ratio is less than default resolution ratio, then by the corresponding camera of the shooting area to the shooting area
Domain carries out target detection.
If 5. the method according to claim 1, wherein the shooting area be nontarget area, lead to
It crosses after the region-wide camera progress target detection, further includes:
The shooting picture of the region-wide camera is monitored using background difference algorithm；
It is again true according to the picture of two or more shooting areas if detecting object to be detected in the shooting picture
Fixed corresponding shooting area is target area or nontarget area.
6. the method according to claim 1, wherein further include:
When object to be detected enters the visual angle overlapping region of the shooting area, mesh is carried out to the object to be detected by script
The corresponding camera of shooting area of mark detection provides the probability distribution of the object's position to be detected；
According to the probability distribution by the object to be detected possibly into shooting area be ranked up from high to low；
According to ranking results successively by it is described possibly into shooting area be input to Target Recognition Algorithms, until determine it is described to
The shooting area that test object enters.
7. the method according to claim 1, wherein further include:
It is spaced after predetermined period, determines that corresponding shooting area is according to the picture of two or more shooting areas again
Target area or nontarget area.
8. a kind of object detecting device characterized by comprising
First object area determination module determines corresponding shooting area for the picture according to two or more shooting areas
Domain is target area or nontarget area, and the target area is the region that object to be detected is likely to occur；
Alternating detection module, if being target area for shooting area, by the corresponding camera of the shooting area and entirely
Area camera alternately target detection, the high resolution of the corresponding camera of the shooting area is in point of the region-wide camera
Resolution；
Region-wide shooting module carries out target by the region-wide camera if being nontarget area for shooting area
Detection.
9. a kind of equipment characterized by comprising
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now object detection method as described in any in claim 1-7.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The object detection method as described in any in claim 1-7 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910001609.3A CN109685062B (en) | 2019-01-02 | 2019-01-02 | Target detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910001609.3A CN109685062B (en) | 2019-01-02 | 2019-01-02 | Target detection method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109685062A true CN109685062A (en) | 2019-04-26 |
| CN109685062B CN109685062B (en) | 2023-07-25 |
Family
ID=66191539
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910001609.3A Active CN109685062B (en) | 2019-01-02 | 2019-01-02 | Target detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109685062B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110177256A (en) * | 2019-06-17 | 2019-08-27 | 北京影谱科技股份有限公司 | A kind of tracking video data acquisition methods and device |
| CN110458198A (en) * | 2019-07-10 | 2019-11-15 | 哈尔滨工业大学（深圳） | Multiresolution target identification method and device |
| CN111770266A (en) * | 2020-06-15 | 2020-10-13 | 北京世纪瑞尔技术股份有限公司 | Intelligent visual perception system |
| CN112766210A (en) * | 2021-01-29 | 2021-05-07 | 苏州思萃融合基建技术研究所有限公司 | Safety monitoring method and device for building construction and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100834465B1 (en) * | 2006-12-21 | 2008-06-05 | 주식회사 다누시스 | System and method for security using motion detection |
| US20090262195A1 (en) * | 2005-06-07 | 2009-10-22 | Atsushi Yoshida | Monitoring system, monitoring method and camera terminal |
| US20100166260A1 (en) * | 2008-12-25 | 2010-07-01 | Ching-Chun Huang | Method for automatic detection and tracking of multiple targets with multiple cameras and system therefor |
| US20120154579A1 (en) * | 2010-12-20 | 2012-06-21 | International Business Machines Corporation | Detection and Tracking of Moving Objects |
| CN106249267A (en) * | 2016-09-30 | 2016-12-21 | 南方科技大学 | A kind of target location tracking method and device |
| US20180191950A1 (en) * | 2015-06-30 | 2018-07-05 | Yutou Technology (Hangzhou) Co., Ltd. | Target tracking device and target tracking method |
-
2019
- 2019-01-02 CN CN201910001609.3A patent/CN109685062B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090262195A1 (en) * | 2005-06-07 | 2009-10-22 | Atsushi Yoshida | Monitoring system, monitoring method and camera terminal |
| KR100834465B1 (en) * | 2006-12-21 | 2008-06-05 | 주식회사 다누시스 | System and method for security using motion detection |
| US20100166260A1 (en) * | 2008-12-25 | 2010-07-01 | Ching-Chun Huang | Method for automatic detection and tracking of multiple targets with multiple cameras and system therefor |
| US20120154579A1 (en) * | 2010-12-20 | 2012-06-21 | International Business Machines Corporation | Detection and Tracking of Moving Objects |
| US20180191950A1 (en) * | 2015-06-30 | 2018-07-05 | Yutou Technology (Hangzhou) Co., Ltd. | Target tracking device and target tracking method |
| CN106249267A (en) * | 2016-09-30 | 2016-12-21 | 南方科技大学 | A kind of target location tracking method and device |
Non-Patent Citations (2)
| Title |
|---|
| 胡亚洲: "基于背景建模和帧间差分法的高点监控行人检测", 实验室研究与探索 * |
| 苑泊舟: "基于差分法和概率估计方法的运动目标检测", 天津理工大学学报 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110177256A (en) * | 2019-06-17 | 2019-08-27 | 北京影谱科技股份有限公司 | A kind of tracking video data acquisition methods and device |
| CN110458198A (en) * | 2019-07-10 | 2019-11-15 | 哈尔滨工业大学（深圳） | Multiresolution target identification method and device |
| CN110458198B (en) * | 2019-07-10 | 2022-03-29 | 哈尔滨工业大学（深圳） | Multi-resolution target identification method and device |
| CN111770266A (en) * | 2020-06-15 | 2020-10-13 | 北京世纪瑞尔技术股份有限公司 | Intelligent visual perception system |
| CN111770266B (en) * | 2020-06-15 | 2021-04-06 | 北京世纪瑞尔技术股份有限公司 | Intelligent visual perception system |
| WO2021253961A1 (en) * | 2020-06-15 | 2021-12-23 | 北京世纪瑞尔技术股份有限公司 | Intelligent visual perception system |
| CN112766210A (en) * | 2021-01-29 | 2021-05-07 | 苏州思萃融合基建技术研究所有限公司 | Safety monitoring method and device for building construction and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109685062B (en) | 2023-07-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109685062A (en) | A kind of object detection method, device, equipment and storage medium | |
| US10645299B2 (en) | Method for tracking and shooting moving target and tracking device | |
| CN110443210B (en) | Pedestrian tracking method and device and terminal | |
| US20170248971A1 (en) | Method for detecting target object, detection apparatus and robot | |
| CN108985162A (en) | Object real-time tracking method, apparatus, computer equipment and storage medium | |
| CN110610150B (en) | Tracking method, device, computing equipment and medium of target moving object | |
| CN106027931A (en) | Video recording method and server | |
| CN108038415B (en) | Unmanned aerial vehicle automatic detection and tracking method based on machine vision | |
| CN108986152B (en) | Foreign matter detection method and device based on difference image | |
| CN109886951A (en) | Method for processing video frequency, device and electronic equipment | |
| CN110570454A (en) | Method and device for detecting foreign matter invasion | |
| KR102199094B1 (en) | Method and Apparatus for Learning Region of Interest for Detecting Object of Interest | |
| CN109255802B (en) | Pedestrian tracking method, device, computer equipment and storage medium | |
| CN112465855B (en) | Passenger flow statistical method, device, storage medium and equipment | |
| CN113490965A (en) | Image tracking processing method and device, computer equipment and storage medium | |
| CN111931720A (en) | Method, apparatus, computer device and storage medium for tracking image feature points | |
| CN112347817B (en) | Video target detection and tracking method and device | |
| CN111524086B (en) | Moving object detection device, moving object detection method, and storage medium | |
| CN113256683B (en) | Target tracking method and related equipment | |
| CN110930437B (en) | Target tracking method and device | |
| CN112347818B (en) | Method and device for screening difficult sample images of video target detection model | |
| CN110516559B (en) | Target tracking method and device suitable for accurate monitoring and computer equipment | |
| CN112070035A (en) | Target tracking method and device based on video stream and storage medium | |
| CN109238286B (en) | Intelligent navigation method, intelligent navigation device, computer equipment and storage medium | |
| CN111652907A (en) | Multi-target tracking method and device based on data association and electronic equipment |
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 |