CN105869148A - Target detection method and device - Google Patents
Target detection method and device Download PDFInfo
- Publication number
- CN105869148A CN105869148A CN201610171936.XA CN201610171936A CN105869148A CN 105869148 A CN105869148 A CN 105869148A CN 201610171936 A CN201610171936 A CN 201610171936A CN 105869148 A CN105869148 A CN 105869148A
- Authority
- CN
- China
- Prior art keywords
- subimage
- pixel
- alternative
- predetermined
- image
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
Abstract
The invention discloses a target detection method and a target detection device and belongs to the technical field of image processing. The target detection method comprises the following steps of: carrying out edge detection on a target image by employing an edge detection algorithm to obtain an edge bar chart composed of edge lines, wherein the edge lines are lines composed of pixel points whose gray difference values with adjacent pixel points are greater than a predetermined difference value threshold in the target image; traversing the edge bar chart by employing a predetermined sliding window to obtain a plurality of sub-images which are images composed of pixel points in the predetermined sliding window of each frame; determining alternative sub-images from the plurality of sub-images, wherein an accumulated value of pixel values of all pixel points in the alternative sub-images is greater than the predetermined difference value threshold; and carrying out target detection on an area where the alternative sub-images are located. According to the target detection method and the target detection device, the technical problem of low working efficiency of target detection caused by detection of excessive windows and recognition of the windows one by one in correlation technologies is solved and the working efficiency of target detection is improved.
Description
Technical field
It relates to technical field of image processing, particularly to a kind of object detection method and device.
Background technology
Along with the development of image processing techniques, the range of application of target detection technique is the most extensive.
In correlation technique, in order to view picture target image is detected, need to utilize search window to travel through target
Image, each image traversed by search window all individually identifies, additionally as a subimage
Owing to the size needing the target object of detection is unknown, generally also need to be varied multiple times the ratio of search window and
Step-length is to detect the subimage in more search windows.Search window is utilized to travel through target image by above-mentioned
Mode, needs detect a lot of windows and be identified one by one, the inefficiency of this object detection method.
Summary of the invention
The disclosure provides a kind of object detection method and device.Described technical scheme is as follows:
First aspect according to disclosure embodiment, it is provided that a kind of object detection method, described method includes:
Utilize edge detection algorithm that target image is carried out rim detection, obtain the edge being made up of edge lines
Lines figure, described edge lines are that the gray scale difference value in described target image and between neighbor pixel is more than pre-
Determine the lines of the pixel composition of difference threshold;
Use predetermined sliding window that described edge lines figure is traveled through, obtain multiple subimage, described son
Image is the image of the pixel composition in predetermined sliding window described in every frame;
Alternative subimage is determined, each pixel in described alternative subimage from the plurality of subimage
The accumulated value of pixel value is more than predetermined pixel value threshold value;
Described alternative subimage region is carried out target detection.
What the implementation of the first aspect of disclosure embodiment can reach has the beneficial effect that by with predetermined
Edge lines figure is traveled through by sliding window, obtains multiple subimage, and subimage is the predetermined sliding window of every frame
The image of the pixel composition in Kou, determines alternative subimage from multiple subimages that traversal obtains, this
In the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value in the alternative subimage said, right
Alternative subimage region carries out target detection;Owing to utilizing predetermined sliding window traversal edge lines figure true
Determine there may be mesh target area, namely only alternative subimage region has been carried out target detection, reduced
Need to carry out the search window of independent target recognition, solve in correlation technique too much owing to needing to detect
Window is also identified and the ineffective technical problem of target detection that causes one by one, improves target inspection
The work efficiency surveyed.
Optionally, described described alternative subimage region is carried out target detection, including:
Search window is utilized to travel through described alternative subimage region with pre-fixed step size;
Image in described search window is carried out target recognition, extracts the target identified;
Adjust the size of described search window according to the first predetermined ratio, adjust according to the second predetermined ratio described
Pre-fixed step size;If the length of described search window is little less than the width of predetermined length threshold value and described search window
In preset width threshold value, then continue executing with described utilize search window with pre-fixed step size travel through described alternative subgraph
Step as region.
What the optional implementation of first aspect of disclosure embodiment can reach has the beneficial effect that by profit
Travel through alternative subimage region with search window with pre-fixed step size, the image in search window is carried out mesh
Mark is other, to realize the target detection in image.Simultaneously as the size of target in target image is not
Knowing, size and pre-fixed step size to search window are adjusted, in order to enter the target in image exactly
Row identifies.
Optionally, described from the plurality of subimage, determine alternative subimage, including:
For each subimage, calculate the accumulated value of the pixel value of all pixels in described subimage, will be tired
Value added it is defined as described alternative subimage more than the subimage of described predetermined pixel value threshold value.
What the optional implementation of first aspect of disclosure embodiment can reach has the beneficial effect that by meter
Whether have during in operator image, the accumulated value of the pixel value of all pixels determines this image comprise target can
Energy property, if this image may comprise target, then by this image confirming by alternative subimage, so that follow-up mistake
Alternative subimage region is only utilized search window to be identified by journey, decreases and needs individually to be identified
The quantity of window, improve the efficiency of target detection.
Optionally, described from the plurality of subimage, determine alternative subimage, including:
For each subimage, calculate the quantity of the pixel value pixel more than zero in described subimage, by quantity
It is defined as described alternative subimage more than the subimage of predetermined quantity threshold value.
What the optional implementation of first aspect of disclosure embodiment can reach has the beneficial effect that by meter
Calculate the quantity of the pixel more than zero of pixel value in described subimage, determine whether this image has and comprise target
Probability, if this image may comprise target, then by this image confirming by alternative subimage so that after
Alternative subimage region is only utilized search window to be identified by continuous process, decreases and needs individually to carry out
The quantity of the window identified, improves the efficiency of target detection.
Optionally, described method also includes:
To needing the image carrying out target detection to carry out gray processing and down-sampled, obtain described target image.
What the optional implementation of first aspect of disclosure embodiment can reach has the beneficial effect that by right
The image carrying out target detection is needed to carry out down-sampled and gray processing, due to the feature of the image after down-sampled
Still can represent down-sampled before the feature of image, down-sampled after the pixel that comprised of image far fewer than
Down-sampled before image in pixel, and then ensure down-sampled after image still can represent original image
While, it is also possible to reduce follow-up operand.
Second aspect according to disclosure embodiment, it is provided that a kind of object detecting device, described device includes:
First detection module, is configured to, with edge detection algorithm and target image is carried out rim detection,
To the edge lines figure being made up of edge lines, described edge lines be in described target image with neighbor
Gray scale difference value between point is more than the lines of the pixel composition of predetermined difference value threshold value;
Spider module, is configured to use predetermined sliding window to travel through described edge lines figure, obtains
Multiple subimages, described subimage is the image of the pixel composition in predetermined sliding window described in every frame;
Determine module, be configured to determine alternative subimage, described alternative son from the plurality of subimage
In image, the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value;
Second detection module, is configured to carry out the described alternative subimage region determining that module determines
Target detection.
Optionally, described second detection module, including:
Traversal submodule, is configured to, with search window and travels through described alternative subimage place with pre-fixed step size
Region;
Identify submodule, be configured to the image in described search window is carried out target recognition, extract and identify
The target gone out;
Adjust submodule, be configured to adjust according to the first predetermined ratio the size of described search window, according to
Second predetermined ratio adjusts described pre-fixed step size;The length of described search window less than predetermined length threshold value and
The width of described search window, less than in the case of preset width threshold value, continues with search window with predetermined step
The described alternative subimage region of long traversal.
Optionally, described determine module, be also configured to
For each subimage, calculate the accumulated value of the pixel value of all pixels in described subimage, will be tired
Value added it is defined as described alternative subimage more than the subimage of described predetermined pixel value threshold value;Or,
For each subimage, calculate the quantity of the pixel value pixel more than zero in described subimage, by quantity
It is defined as described alternative subimage more than the subimage of predetermined quantity threshold value.
Optionally, described device also includes:
Down-sampled module, is configured to needing the image carrying out target detection to carry out gray processing and down-sampled,
Obtain described target image.
The third aspect according to disclosure embodiment, it is provided that a kind of object detecting device, described device includes:
Processor;
For storing the memorizer of described processor executable;
Wherein, described processor is configured to:
Utilize edge detection algorithm that target image is carried out rim detection, obtain the edge being made up of edge lines
Lines figure, described edge lines are that the gray scale difference value in described target image and between neighbor pixel is more than pre-
Determine the lines of the pixel composition of difference threshold;
Use predetermined sliding window that described edge lines figure is traveled through, obtain multiple subimage, described son
Image is the image of the pixel composition in predetermined sliding window described in every frame;
Alternative subimage is determined, each pixel in described alternative subimage from the plurality of subimage
The accumulated value of pixel value is more than predetermined pixel value threshold value;
Described alternative subimage region is carried out target detection.
It should be appreciated that it is only exemplary that above general description and details hereinafter describe, can not
Limit the disclosure.
Accompanying drawing explanation
Accompanying drawing herein is merged in description and constitutes the part of this specification, it is shown that meet the disclosure
Embodiment, and in description together for explaining the principle of the disclosure.
Fig. 1 is the flow chart according to a kind of object detection method shown in an exemplary embodiment;
Fig. 2 A is the flow chart according to a kind of object detection method shown in another exemplary embodiment;
Fig. 2 B is according to a kind of edge lines figure shown in an exemplary embodiment;
Fig. 3 is the block diagram according to a kind of object detecting device shown in an exemplary embodiment;
Fig. 4 is the block diagram according to a kind of object detecting device shown in another exemplary embodiment.
Detailed description of the invention
Here will illustrate exemplary embodiment in detail, its example represents in the accompanying drawings.Following retouches
Stating when relating to accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represents same or analogous key element.
Embodiment described in following exemplary embodiment does not represent all embodiment party consistent with the disclosure
Formula.On the contrary, they only with describe in detail in appended claims, the disclosure some in terms of mutually one
The example of the apparatus and method caused.
Fig. 1 is the flow chart according to a kind of object detection method shown in an exemplary embodiment, and this target is examined
Survey method can include following several step.
In a step 101, utilize edge detection algorithm that target image is carried out rim detection, obtain by edge
The edge lines figure of lines composition, these edge lines are the gray scale difference in target image and between neighbor pixel
Value is more than the lines of the pixel composition of predetermined difference value threshold value.
In a step 102, use predetermined sliding window that edge lines figure is traveled through, obtain multiple subgraph
Picture, this subimage is the image of the pixel composition in the predetermined sliding window of every frame.
Utilize predetermined sliding window that edge lines figure carries out traversal to be not intended to edge lines figure is carried out mesh
Mark detection, but in order to determine, target image there may be mesh target area.
In step 103, from multiple subimages, determine alternative subimage, in this alternative subimage each
The accumulated value of the pixel value of pixel is more than predetermined pixel value threshold value.
If the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value in subimage, then it is assumed that should
Subimage may comprise target, this subimage is defined as alternative subimage, so that follow-up to alternative
Subimage region carries out target detection.
At step 104, alternative subimage region is carried out target detection.
In sum, the object detection method provided in disclosure embodiment, by with predetermined sliding window pair
Edge lines figure travels through, and obtains multiple subimage, and subimage is the pixel in the predetermined sliding window of every frame
The image of some composition, determines alternative subimage from multiple subimages that traversal obtains, and that is said here is standby
The accumulated value of the pixel value of each pixel in subimage is selected to be more than predetermined pixel value threshold value, to alternative subimage
Region carries out target detection;Owing to utilizing predetermined sliding window traversal edge lines figure to determine and may deposit
In mesh target area, namely only alternative subimage region is carried out target detection, decrease needs and carry out
The individually search window of target recognition, solves in correlation technique owing to needing to detect too much window one by one
The ineffective technical problem of target detection being identified and cause, improves the work effect of target detection
Rate.
Fig. 2 A is the flow chart according to a kind of object detection method shown in another exemplary embodiment, this target
Detection method can include following several step.
In step 201, utilize edge detection algorithm that target image is carried out rim detection, obtain by edge
The edge lines figure of lines composition, these edge lines are the gray scale difference in target image and between neighbor pixel
Value is more than the lines of the pixel composition of predetermined difference value threshold value.
Here the edge detection algorithm said can be Sobel Sobel operator, it is also possible to be other this area skills
The edge detection algorithm that art personnel commonly use, such as Luo Baici Roberts operator, Canny operator etc., this reality
Execute example edge detection algorithm is not especially limited.In the present embodiment, with edge detection algorithm as Sobel
Operator is illustrated.
For each pixel in target image, edge detection algorithm is utilized to calculate this pixel and its phase
Gray scale difference value between adjacent pixel, is defined as limit by this gray scale difference value more than the pixel of predetermined difference value threshold value
Edge pixel, say, that edge lines are the lines that the pixel that around gray scale difference value is bigger is formed.
Edge lines figure is formed, specifically, by these edge pictures according to the edge pixel point in target image
Vegetarian refreshments shows in edge lines figure, and the pixel value of each edge pixel point is (255,255,255),
Namely white pixel point, the pixel value of other pixels is (0,0,0), namely black pixel point.Such as figure
Shown in 2B, Fig. 2 B is according to a kind of edge lines figure shown in an exemplary embodiment, these edge pixels
Point is shown as white, and composition edge lines show in edge lines figure.
In step 202., use predetermined sliding window that edge lines figure is traveled through, obtain multiple subgraph
Picture, this subimage is the image of the pixel composition in the predetermined sliding window of every frame, from multiple subimages really
Making alternative subimage, in this alternative subimage, the accumulated value of the pixel value of each pixel is more than intended pixel
Value threshold value.
Predetermined sliding window is utilized with particular step size, edge lines figure to be traveled through, by predetermined sliding window
Every frame pixel composition image be defined as subimage.
When determining alternative subimage from multiple subimages, can be at least to include the following two kinds mode:
In first kind of way, for each subimage, calculate the pixel value of all pixels in this subimage
Accumulated value, is defined as alternative subimage by accumulated value more than the subimage of predetermined pixel value threshold value.
Owing to the pixel in edge lines figure includes the white pixel being positioned at edge line and black not
The pixel being positioned on edge line, the pixel value of white pixel point is higher than the pixel value of black pixel point, group
In image, the accumulated value of the pixel value of all pixels is more than predetermined pixel value threshold value, shows limit in this subimage
Edge pixel is more, now it has been generally acknowledged that and there may be target to be detected in this subimage, by true for this subimage
It is set to alternative subimage, so that follow-up carries out target detection to alternative subimage region.
In the second way, for each subimage, calculate the number of the pixel value pixel more than zero in subimage
Amount, is defined as alternative subimage by quantity more than the subimage of predetermined quantity threshold value.
When image after target image is gray processing, due to the gray value of each pixel on target image
It is 0 (black) or 1 (white), therefore, the most only needs to calculate pixel value in subimage and be more than zero
The quantity (namely pixel of white) of pixel, is defined as standby by quantity more than the subimage of predetermined quantity threshold value
Select subimage.
It addition, the present embodiment is not to the size of predetermined sliding window, the numerical value of particular step size and intended pixel
The numerical value of value threshold value carries out concrete restriction, can determine according to practical situation.
In step 203, search window is utilized to travel through alternative subimage region with pre-fixed step size, to searching
Image in rope window carries out target recognition, extracts the target identified.
It is the more maturation that those skilled in the art have grasped that image in search window carries out target recognition
Technology, the present embodiment no longer repeats.Additionally, the present embodiment is not to the size of search window and predetermined
The numerical value of step-length specifically limits, and can determine according to practical situation.
In step 204, the size of search window is adjusted according to the first predetermined ratio, according to the second pre-definite proportion
Example adjusts pre-fixed step size, if the length of search window is less than less than the width of predetermined length threshold value and search window
Preset width threshold value, then continue executing with and utilize search window to travel through alternative subimage region with pre-fixed step size
Step.
Owing to the size of the target object in target image is unknown, and only when target is completely in search window,
Can be successfully identified.Accordingly, it would be desirable to target image is traveled through by the size repeatedly adjusting search window,
Image in search window is individually identified.Specifically, after having carried out once traversal, according to first
Predetermined ratio adjusts the size of search window, correspondingly adjusts pre-fixed step size according to the second predetermined ratio, according to
After adjustment search window and pre-fixed step size again travel through, the first predetermined ratio here is usually big
In the numerical value of 1.
When the length of search window is less than preset width threshold value less than the width of predetermined length threshold value and search window
Time, then continue executing with in step 203 and utilize search window to travel through alternative subimage region with pre-fixed step size
Step.
When the length of search window is more than preset width threshold more than the width of predetermined length threshold value or search window
During value, then terminate this target detection.
It should be noted is that, not to the first predetermined ratio and the numerical value of the second predetermined ratio in the present embodiment
Specifically limit, can determine according to practical situation.
In sum, the object detection method provided in disclosure embodiment, by with predetermined sliding window pair
Edge lines figure travels through, and obtains multiple subimage, and subimage is the pixel in the predetermined sliding window of every frame
The image of some composition, determines alternative subimage from multiple subimages that traversal obtains, and that is said here is standby
The accumulated value of the pixel value of each pixel in subimage is selected to be more than predetermined pixel value threshold value, to alternative subimage
Region carries out target detection;Owing to utilizing predetermined sliding window traversal edge lines figure to determine and may deposit
In mesh target area, namely only alternative subimage region is carried out target detection, decrease needs and carry out
The individually search window of target recognition, solves in correlation technique owing to needing to detect too much window one by one
The ineffective technical problem of target detection being identified and cause, improves the work effect of target detection
Rate.
In actual application scenarios, right with this image to needing the rim detection carrying out the image of target detection
The color of elephant there is no much relations, therefore to reduce operand, the most only needs to consider the gray scale in image
It is worth, namely can obtain first to needing the image carrying out target detection to carry out gray processing and down-sampled
Target image.
Owing to the gray value of the image after down-sampled all represents the average gray value of down-sampled front corresponding region,
The feature of the image after the most down-sampled still can represent the feature of down-sampled front image.
Then step 201 is performed according to the target image obtained.
Following for disclosure device embodiment, may be used for performing method of disclosure embodiment.For the disclosure
The details not disclosed in device embodiment, refer to method of disclosure embodiment.
Fig. 3 is the block diagram according to a kind of object detecting device shown in an exemplary embodiment, this target detection
Device may include that first detection module 310, spider module 320, determines module 330 and the second detection mould
Block 340.
First detection module 310, is configured to, with edge detection algorithm and target image is carried out rim detection,
Obtaining the edge lines figure being made up of edge lines, edge lines are in target image and between neighbor pixel
Gray scale difference value more than predetermined difference value threshold value pixel composition lines.
Spider module 320, is configured to use predetermined sliding window to obtain first detection module 310 detection
Edge lines figure travels through, and obtains multiple subimage, in this subimage is this predetermined sliding window of every frame
The image of pixel composition.
Determine module 330, be configured to obtain multiple subimage is determined alternative from spider module 320 traversal
Subimage, in this alternative subimage, the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value.
Second detection module 340, is configured to determining that the alternative subimage region that module 330 determines is entered
Row target detection.
In sum, the object detecting device provided in disclosure embodiment, by with predetermined sliding window pair
Edge lines figure travels through, and obtains multiple subimage, and subimage is the pixel in the predetermined sliding window of every frame
The image of some composition, determines alternative subimage from multiple subimages that traversal obtains, and that is said here is standby
The accumulated value of the pixel value of each pixel in subimage is selected to be more than predetermined pixel value threshold value, to alternative subimage
Region carries out target detection;Owing to utilizing predetermined sliding window traversal edge lines figure to determine and may deposit
In mesh target area, namely only alternative subimage region is carried out target detection, decrease needs and carry out
The individually search window of target recognition, solves in correlation technique owing to needing to detect too much window one by one
The ineffective technical problem of target detection being identified and cause, improves the work effect of target detection
Rate.
Fig. 4 is the block diagram according to a kind of object detecting device shown in another exemplary embodiment, and this target is examined
Survey device to may include that first detection module 410, spider module 420, determine module 430 and the second detection
Module 440.
First detection module 410, is configured to, with edge detection algorithm and target image is carried out rim detection,
Obtaining the edge lines figure being made up of edge lines, edge lines are in target image and between neighbor pixel
Gray scale difference value more than predetermined difference value threshold value pixel composition lines.
Spider module 420, is configured to use predetermined sliding window to obtain first detection module 410 detection
Edge lines figure travels through, and obtains multiple subimage, in this subimage is this predetermined sliding window of every frame
The image of pixel composition.
Determine module 430, be configured to determine from spider module 420 multiple subimages of obtaining of traversal standby
Selecting subimage, in this alternative subimage, the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value.
Second detection module 440, is configured to determining that the alternative subimage region that module 430 determines is entered
Row target detection.
Optionally, the second detection module 440, including: traversal submodule 440a, identify submodule 440b and
Adjust submodule 440c.
Traversal submodule 440a, is configured to, with search window and travels through alternative subimage place with pre-fixed step size
Region.
Identify submodule 440b, be configured to the image in search window is carried out target recognition, extract and identify
The target gone out.
Adjust submodule 440c, be configured to adjust according to the first predetermined ratio the size of search window, according to
Second predetermined ratio adjusts pre-fixed step size;Length at search window is less than predetermined length threshold value and search window
Width less than in the case of preset width threshold value, continue to be performed to utilize search window by traversal submodule 440a
Described alternative subimage region is traveled through with pre-fixed step size.
Optionally, determine module 430, be also configured to
For each subimage, calculate the accumulated value of the pixel value of all pixels in subimage, by accumulated value
It is defined as alternative subimage more than the subimage of predetermined pixel value threshold value;
Or, for each subimage, calculate the quantity of the pixel value pixel more than zero in subimage, by number
Amount is defined as alternative subimage more than the subimage of predetermined quantity threshold value.
Optionally, this object detecting device, also include:
Down-sampled module 450, is configured to needing the image carrying out target detection to carry out gray processing and fall is adopted
Sample, obtains target image.
In sum, the object detecting device provided in disclosure embodiment, by with predetermined sliding window pair
Edge lines figure travels through, and obtains multiple subimage, and subimage is the pixel in the predetermined sliding window of every frame
The image of some composition, determines alternative subimage from multiple subimages that traversal obtains, and that is said here is standby
The accumulated value of the pixel value of each pixel in subimage is selected to be more than predetermined pixel value threshold value, to alternative subimage
Region carries out target detection;Owing to utilizing predetermined sliding window traversal edge lines figure to determine and may deposit
In mesh target area, namely only alternative subimage region is carried out target detection, decrease needs and carry out
The individually search window of target recognition, solves in correlation technique owing to needing to detect too much window one by one
The ineffective technical problem of target detection being identified and cause, improves the work effect of target detection
Rate.
The disclosure one exemplary embodiment provides a kind of object detecting device, it is possible to realize what the disclosure provided
Object detection method, this object detecting device includes: processor, for storing processor executable
Memorizer;
Wherein, processor is configured to:
Utilize edge detection algorithm that target image is carried out rim detection, obtain the edge being made up of edge lines
Lines figure, these edge lines are that the gray scale difference value in target image and between neighbor pixel is more than predetermined difference value
The lines of the pixel composition of threshold value;
Use predetermined sliding window that this edge lines figure is traveled through, obtain multiple subimage, this subimage
The image formed for the pixel in the predetermined sliding window of every frame;
Alternative subimage is determined, the pixel value of each pixel in this alternative subimage from multiple subimages
Accumulated value more than predetermined pixel value threshold value;
Alternative subimage region is carried out target detection.
Those skilled in the art, after considering description and putting into practice invention disclosed herein, will readily occur to these public affairs
Other embodiment opened.The application is intended to any modification, purposes or the adaptations of the disclosure,
These modification, purposes or adaptations are followed the general principle of the disclosure and include that the disclosure is not disclosed
Common knowledge in the art or conventional techniques means.Description and embodiments is considered only as exemplary
, the true scope of the disclosure and spirit are pointed out by claim below.
It should be appreciated that the disclosure is not limited to accurate knot described above and illustrated in the accompanying drawings
Structure, and various modifications and changes can carried out without departing from the scope.The scope of the present disclosure is only by appended
Claim limits.
Claims (10)
1. an object detection method, it is characterised in that described method includes:
Utilize edge detection algorithm that target image is carried out rim detection, obtain the edge being made up of edge lines
Lines figure, described edge lines are that the gray scale difference value in described target image and between neighbor pixel is more than pre-
Determine the lines of the pixel composition of difference threshold;
Use predetermined sliding window that described edge lines figure is traveled through, obtain multiple subimage, described son
Image is the image of the pixel composition in predetermined sliding window described in every frame;
Alternative subimage is determined, each pixel in described alternative subimage from the plurality of subimage
The accumulated value of pixel value is more than predetermined pixel value threshold value;
Described alternative subimage region is carried out target detection.
Method the most according to claim 1, it is characterised in that described to described alternative subimage place
Region carries out target detection, including:
Search window is utilized to travel through described alternative subimage region with pre-fixed step size;
Image in described search window is carried out target recognition, extracts the target identified;
Adjust the size of described search window according to the first predetermined ratio, adjust according to the second predetermined ratio described
Pre-fixed step size;If the length of described search window is little less than the width of predetermined length threshold value and described search window
In preset width threshold value, then continue executing with described utilize search window with pre-fixed step size travel through described alternative subgraph
Step as region.
Method the most according to claim 1, it is characterised in that described true from the plurality of subimage
Make alternative subimage, including:
For each subimage, calculate the accumulated value of the pixel value of all pixels in described subimage, will be tired
Value added it is defined as described alternative subimage more than the subimage of described predetermined pixel value threshold value.
Method the most according to claim 1, it is characterised in that determine from the plurality of subimage
Alternative subimage, including:
For each subimage, calculate the quantity of the pixel value pixel more than zero in described subimage, by quantity
It is defined as described alternative subimage more than the subimage of predetermined quantity threshold value.
5. according to described method arbitrary in Claims 1-4, it is characterised in that described method also includes:
To needing the image carrying out target detection to carry out gray processing and down-sampled, obtain described target image.
6. an object detecting device, it is characterised in that described device includes:
First detection module, is configured to, with edge detection algorithm and target image is carried out rim detection,
To the edge lines figure being made up of edge lines, described edge lines be in described target image with neighbor
Gray scale difference value between point is more than the lines of the pixel composition of predetermined difference value threshold value;
Spider module, is configured to use predetermined sliding window to travel through described edge lines figure, obtains
Multiple subimages, described subimage is the image of the pixel composition in predetermined sliding window described in every frame;
Determine module, be configured to determine alternative subimage, described alternative son from the plurality of subimage
In image, the accumulated value of the pixel value of each pixel is more than predetermined pixel value threshold value;
Second detection module, is configured to carry out the described alternative subimage region determining that module determines
Target detection.
Device the most according to claim 6, it is characterised in that described second detection module, including:
Traversal submodule, is configured to, with search window and travels through described alternative subimage place with pre-fixed step size
Region;
Identify submodule, be configured to the image in described search window is carried out target recognition, extract and identify
The target gone out;
Adjust submodule, be configured to adjust according to the first predetermined ratio the size of described search window, according to
Second predetermined ratio adjusts described pre-fixed step size;The length of described search window less than predetermined length threshold value and
The width of described search window, less than in the case of preset width threshold value, continues with search window with predetermined step
The described alternative subimage region of long traversal.
Device the most according to claim 7, it is characterised in that described determine module, is also configured to
For each subimage, calculate the accumulated value of the pixel value of all pixels in described subimage, will be tired
Value added it is defined as described alternative subimage more than the subimage of described predetermined pixel value threshold value;Or,
For each subimage, calculate the quantity of the pixel value pixel more than zero in described subimage, by quantity
It is defined as described alternative subimage more than the subimage of predetermined quantity threshold value.
9. according to described device arbitrary in claim 6 to 8, it is characterised in that described device also includes:
Down-sampled module, is configured to needing the image carrying out target detection to carry out gray processing and down-sampled,
Obtain described target image.
10. an object detecting device, it is characterised in that including:
Processor;
For storing the memorizer of the executable instruction of described processor;
Wherein, described processor is configured to:
Utilize edge detection algorithm that target image is carried out rim detection, obtain the edge being made up of edge lines
Lines figure, described edge lines are that the gray scale difference value in described target image and between neighbor pixel is more than pre-
Determine the lines of the pixel composition of difference threshold;
Use predetermined sliding window that described edge lines figure is traveled through, obtain multiple subimage, described son
Image is the image of the pixel composition in predetermined sliding window described in every frame;
Alternative subimage is determined, each pixel in described alternative subimage from the plurality of subimage
The accumulated value of pixel value is more than predetermined pixel value threshold value;
Described alternative subimage region is carried out target detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610171936.XA CN105869148A (en) | 2016-03-24 | 2016-03-24 | Target detection method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610171936.XA CN105869148A (en) | 2016-03-24 | 2016-03-24 | Target detection method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105869148A true CN105869148A (en) | 2016-08-17 |
Family
ID=56625674
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610171936.XA Pending CN105869148A (en) | 2016-03-24 | 2016-03-24 | Target detection method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105869148A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874882A (en) * | 2017-02-24 | 2017-06-20 | 广东诚泰交通科技发展有限公司 | Pavement disease edge detection method and device |
CN108682019A (en) * | 2018-04-25 | 2018-10-19 | 高雪亚 | Height-adjustable is hurdled system |
CN108694161A (en) * | 2017-04-05 | 2018-10-23 | 北大方正集团有限公司 | Type-setting document processing method and processing device |
CN110020647A (en) * | 2018-01-09 | 2019-07-16 | 杭州海康威视数字技术股份有限公司 | A kind of contraband object detection method, device and computer equipment |
WO2019148894A1 (en) * | 2018-02-01 | 2019-08-08 | 汕头大学 | Method and device for tracking measurement offset by using image patches, and storage medium |
CN113785327A (en) * | 2021-06-18 | 2021-12-10 | 商汤国际私人有限公司 | Image processing method, image processing device, edge calculation device and computer storage medium |
WO2022096954A1 (en) * | 2021-06-18 | 2022-05-12 | Sensetime International Pte. Ltd. | Image processing method and device, edge computing device, and computer storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102779267A (en) * | 2011-05-12 | 2012-11-14 | 株式会社理光 | Method and device for detection of specific object region in image |
CN103839062A (en) * | 2014-03-11 | 2014-06-04 | 东方网力科技股份有限公司 | Image character positioning method and device |
CN104166841A (en) * | 2014-07-24 | 2014-11-26 | 浙江大学 | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network |
CN104463097A (en) * | 2014-10-31 | 2015-03-25 | 武汉工程大学 | High-voltage wire image detection method based on local self-adaptation threshold value partitioning algorithm |
US20150235092A1 (en) * | 2014-02-14 | 2015-08-20 | Toyota Motor Engineering & Manufacturing North America, Inc. | Parts based object tracking method and apparatus |
-
2016
- 2016-03-24 CN CN201610171936.XA patent/CN105869148A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102779267A (en) * | 2011-05-12 | 2012-11-14 | 株式会社理光 | Method and device for detection of specific object region in image |
US20150235092A1 (en) * | 2014-02-14 | 2015-08-20 | Toyota Motor Engineering & Manufacturing North America, Inc. | Parts based object tracking method and apparatus |
CN103839062A (en) * | 2014-03-11 | 2014-06-04 | 东方网力科技股份有限公司 | Image character positioning method and device |
CN104166841A (en) * | 2014-07-24 | 2014-11-26 | 浙江大学 | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network |
CN104463097A (en) * | 2014-10-31 | 2015-03-25 | 武汉工程大学 | High-voltage wire image detection method based on local self-adaptation threshold value partitioning algorithm |
Non-Patent Citations (2)
Title |
---|
何小海,滕奇志: "《图像通信》", 29 May 2005 * |
李俊山,李旭辉: "《数字图像处理 第2版》", 30 November 2013, 清华大学出版社 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874882A (en) * | 2017-02-24 | 2017-06-20 | 广东诚泰交通科技发展有限公司 | Pavement disease edge detection method and device |
CN106874882B (en) * | 2017-02-24 | 2020-02-18 | 广东诚泰交通科技发展有限公司 | Pavement disease edge detection method and device |
CN108694161A (en) * | 2017-04-05 | 2018-10-23 | 北大方正集团有限公司 | Type-setting document processing method and processing device |
CN110020647A (en) * | 2018-01-09 | 2019-07-16 | 杭州海康威视数字技术股份有限公司 | A kind of contraband object detection method, device and computer equipment |
CN110020647B (en) * | 2018-01-09 | 2021-03-02 | 杭州海康威视数字技术股份有限公司 | Contraband target detection method and device and computer equipment |
WO2019148894A1 (en) * | 2018-02-01 | 2019-08-08 | 汕头大学 | Method and device for tracking measurement offset by using image patches, and storage medium |
CN108682019A (en) * | 2018-04-25 | 2018-10-19 | 高雪亚 | Height-adjustable is hurdled system |
CN108682019B (en) * | 2018-04-25 | 2019-03-22 | 六安荣耀创新智能科技有限公司 | Height-adjustable is hurdled system |
CN113785327A (en) * | 2021-06-18 | 2021-12-10 | 商汤国际私人有限公司 | Image processing method, image processing device, edge calculation device and computer storage medium |
WO2022096954A1 (en) * | 2021-06-18 | 2022-05-12 | Sensetime International Pte. Ltd. | Image processing method and device, edge computing device, and computer storage medium |
AU2021204572A1 (en) * | 2021-06-18 | 2023-01-19 | Sensetime International Pte. Ltd. | Image processing method and device, edge computing device, and computer storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105869148A (en) | Target detection method and device | |
CN105913093B (en) | A kind of template matching method for Text region processing | |
US20200026930A1 (en) | Lane line detection method and apparatus | |
CN104685513B (en) | According to the high-resolution estimation of the feature based of the low-resolution image caught using array source | |
US9524558B2 (en) | Method, system and software module for foreground extraction | |
CN106530271B (en) | A kind of infrared image conspicuousness detection method | |
CN108921089A (en) | Method for detecting lane lines, device and system and storage medium | |
CN106663322B (en) | The method and apparatus of feature in image for identification | |
CN106920245B (en) | Boundary detection method and device | |
US20140050387A1 (en) | System and Method for Machine Vision Inspection | |
CN105279372A (en) | Building height computing method and apparatus | |
CN110570435B (en) | Method and device for carrying out damage segmentation on vehicle damage image | |
JP2007148809A (en) | Image analysis program, recording medium recorded with the program, image analyzer, and image analysis method | |
US9767387B2 (en) | Predicting accuracy of object recognition in a stitched image | |
CN111340749B (en) | Image quality detection method, device, equipment and storage medium | |
CN106033535A (en) | Electronic paper marking method | |
CN103927749A (en) | Image processing method and device and automatic optical detector | |
CN104867128A (en) | Image blurring detection method and device | |
CN103093458A (en) | Detecting method and detecting device for key frame | |
CN103514460B (en) | Video monitoring multi-view-angle vehicle detecting method and device | |
CN108009556A (en) | A kind of floater in river detection method based on fixed point graphical analysis | |
CN103985106B (en) | Apparatus and method for carrying out multiframe fusion to very noisy image | |
CN104182983B (en) | Highway monitoring video definition detection method based on corner features | |
CN103247038A (en) | Overall image information synthetic method driven by visual cognition model | |
CN110334760A (en) | A kind of optical component damage detecting method and system based on resUnet |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160817 |
|
RJ01 | Rejection of invention patent application after publication |