CN110517287A - Obtain method, apparatus, equipment and the storage medium of robot fish movement track - Google Patents
Obtain method, apparatus, equipment and the storage medium of robot fish movement track Download PDFInfo
- Publication number
- CN110517287A CN110517287A CN201910410101.9A CN201910410101A CN110517287A CN 110517287 A CN110517287 A CN 110517287A CN 201910410101 A CN201910410101 A CN 201910410101A CN 110517287 A CN110517287 A CN 110517287A
- Authority
- CN
- China
- Prior art keywords
- machine fish
- image
- split
- fish
- neuron
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of method, apparatus, equipment and storage mediums for obtaining robot fish movement track, to solve the problems, such as occurring machine fish position loss when the motion profile to underwater machine fish tracks in the related technology.The method for obtaining robot fish movement track includes: the image for obtaining the machine fish to move about in water；Described image is split, multiple subgraphs are obtained；Identify the machine fish, in the multiple subgraph with position of the determination machine fish in described image；When can not recognize the machine fish in the multiple subgraph, after controlling the machine fish execution deliberate action, the image of the machine fish is obtained again and determines position of the machine fish in the image got again；The motion profile of the machine fish is determined according to position of the machine fish obtained at least twice in described image.The present invention effectively can carry out position tracking to underwater machine fish.
Description
Technical field
The present invention relates to track following technical fields, particularly relate to method, the dress of a kind of acquisition robot fish movement track
It sets, equipment and storage medium.
Background technique
In recent years, being constantly progressive with bionics technology imitates the underwater robot (machine fish) of Push Technology under fish and water
Research increasingly cause to pay close attention to, become one of the hot spot of underwater robot area research.Machine fish can not only be in complex environment
Lower progress underwater operation, marine monitoring, scouting etc. play a role, and provide one for development of new submarine navigation device
The new thinking of kind.
Currently, the research of individual machine fish is focused primarily upon both at home and abroad, and in practical applications, due to the complexity of task
Property, uncertainty, concurrency to need to be cooperated using a plurality of machine fish to complete task.Determine since machine fish itself there is no
Position and telemetering ability, vision system are its unique perception environment " organ ", such as can be by camera (for example, CCD
(Charge Coupled Device, charge-coupled device) camera) acquisition image by processing and analysis, extract effectively letter
Breath is as decision and control foundation.Position and the direction of motion of machine fish and moving target are only quickly and accurately tracked, certainly
Plan control module could make rapidly corresponding decision, it is ensured that the completion of more machine fish collaborative tasks.And complete more machine fish cooperations
The key technology of task first is that multimachine device fish real-time tracking, i.e., find a plurality of machine fish, and will be different in video image
The machine fish of frame shows respective positions sequence after corresponding.But it is tracked in the motion profile to underwater fish
When, the problem of often there is machine fish position loss, machine fish position can not be got.
Summary of the invention
In view of this, it is an object of the invention to propose a kind of method, apparatus, equipment for obtaining robot fish movement track
And storage medium, this method effectively can carry out position tracking to underwater machine fish.
According to the first aspect of the invention, a kind of method for obtaining robot fish movement track is provided, comprising: obtain
The image of the machine fish to move about in water；Described image is split, multiple subgraphs are obtained；In the multiple subgraph
The middle identification machine fish, with position of the determination machine fish in described image；When nothing equal in the multiple subgraph
When method recognizes the machine fish, control after the machine fish executes deliberate action, obtain again the image of the machine fish with
And determine position of the machine fish in the image got again；According to the machine fish obtained at least twice in institute
State the motion profile that the position in image determines the machine fish.
Optionally, described image is split, comprising: using the color- vector of pixel each in image to be split as one
The input vector of a input neuron, using the color- vector of eight pixel adjacent with each pixel as radial function RBF
Each feature vector；Determine the seed neuron in the image to be split, wherein the maximum of pixel to neighbor pixel
When manhatton distance is less than first threshold, which is sub-pixel point, which is seed
Neuron, the seed neuron in the image to be split constitute seed region；By default growing strategy to seed region into
Row growth, obtains multiple packet zones；Calculate the averaged feature vector of each packet zone；Use the average characteristics being calculated
Vector replaces feature vector included in all neurons of packet zone belonging to it；If there is being not attached to any point
The neuron in group region, and the distance of the neuron to its adjacent packet zones is less than second threshold, then connects the neuron
To in away from nearest packet zone；The adjacent packet zone is merged, multiple regions to be split are obtained；According to
The multiple region is split the image to be split, obtains the multiple subgraph.
Optionally, the adjacent packet zone is merged, comprising: when the area in region to be combined is less than preset area,
And the color distance in the region to be combined is when being less than pre-set color distance threshold, by the region merging technique to be combined to adjacent
Region.
Optionally, the position for determining the machine fish includes: described in the target tracking algorism based on Meanshift calculates
Pixel characteristic value probability in target area and candidate region in image obtains object module description and candidate family description；
The similitude of the candidate family of the object module and present frame is measured using similar function；Selection makes the similar function
Maximum candidate family and the Meanshift vector for obtaining object module；Meanshift vector is iterated to calculate, by restraining
To the position of the machine fish.
Optionally, control the machine fish and execute deliberate action, comprising: control the machine fish according to predetermined angle with
And preset direction is turned to.
According to the second aspect of the invention, a kind of device for obtaining robot fish movement track is provided, comprising: first
Module is obtained, for obtaining the image of the machine fish to move about in water；Segmentation module is obtained for being split to described image
To multiple subgraphs；Identification module, for identifying the machine fish in the multiple subgraph, with the determination machine fish
Position in described image；Second obtains module, when can not recognize the machine fish in the multiple subgraph
When, after controlling the machine fish execution deliberate action, the image of the machine fish is obtained again and determines that the machine fish exists
The position in image got again；Determining module, for according to the machine fish that obtains at least twice in described image
In position determine the motion profile of the machine fish.
Optionally, the segmentation module, comprising: setting unit, for by the color- vector of pixel each in image to be split
The input vector that neuron is inputted as one, using the color- vector of eight pixel adjacent with each pixel as radial
Each feature vector of function RBF；Determination unit, for determining the seed neuron in the image to be split, wherein pixel
When the maximum manhatton distance of point to neighbor pixel is less than first threshold, which is sub-pixel point, the sub-pixel
The corresponding neuron of point is seed neuron, and the seed neuron in the image to be split constitutes seed region；It generates single
Member obtains multiple packet zones for growing by default growing strategy to seed region；First computing unit is used
In the averaged feature vector for calculating each packet zone；Replacement unit, for using the averaged feature vector being calculated to replace it
Feature vector included in all neurons of affiliated packet zone；Connection unit, for any if there is being not attached to
The neuron of packet zone, and the distance of the neuron to its adjacent packet zones is less than second threshold, then connects the neuron
It is connected to away from nearest packet zone；Combining unit obtains to be split for merging the adjacent packet zone
Multiple regions；Cutting unit obtains described more for being split according to the multiple region to the image to be split
A subgraph.
Optionally, the combining unit is used for: and described to be combined when the area in region to be combined is less than preset area
When the color distance in region is less than pre-set color distance threshold, by the region merging technique to be combined to adjacent area.
According to the third aspect of the present invention, it provides a kind of electronic equipment, including memory, processor and is stored in
On memory and the computer program that can run on a processor, the processor are realized when executing described program such as the present invention
Any one method for obtaining robot fish movement track that first aspect provides.
According to the fourth aspect of the present invention, a kind of non-transient computer readable storage medium is provided, it is described non-transient
Computer-readable recording medium storage computer instruction, the computer instruction is for making the computer execute the present invention the
The method that any one provided on one side obtains robot fish movement track.
From the above it can be seen that the method provided by the invention for obtaining robot fish movement track, to swimming in water
When the motion profile of dynamic machine fish is tracked, after being split to the image of the underwater machine fish acquired, then
Machine fish is identified in image after segmentation, the discrimination of machine fish can be improved, so as to be avoided as far as possible in the movement to machine fish
There is the problem of machine fish position loss in track during being tracked.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will to embodiment or
Attached drawing needed to be used in the description of the prior art is briefly described, it should be apparent that, the accompanying drawings in the following description is only
Some embodiments of the present invention, for those of ordinary skill in the art, without creative efforts, also
Other drawings may be obtained according to these drawings without any creative labor.
Fig. 1 is a kind of flow chart of method for obtaining robot fish movement track shown according to an exemplary embodiment；
Fig. 2 is that multiple target positions and tracking in real time in water under a kind of Full vision shown according to an exemplary embodiment
The schematic diagram of system；
Fig. 3 is a kind of top view of machine fish shown according to an exemplary embodiment；
Fig. 4 is a kind of bottom view of machine fish shown according to an exemplary embodiment；
Fig. 5 is the flow chart shown according to an exemplary embodiment being split to image；
Fig. 6 is the flow chart of the method shown according to an exemplary embodiment for obtaining robot fish movement track；
Fig. 7 is a kind of block diagram of device for obtaining robot fish movement track shown according to an exemplary embodiment.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and join
According to attached drawing, the present invention is described in more detail.
It should be noted that all uses " first ", " second ", " third " and " the 4th " in the embodiment of the present invention
Statement is for distinguishing two non-equal entities of same names or non-equal parameter, it is seen that " first ", " second "
" third " and " the 4th " should not be construed as the restriction to the embodiment of the present invention, subsequent embodiment pair only for the convenience of statement
This no longer illustrates one by one.
Fig. 1 is a kind of flow chart of method for obtaining robot fish movement track shown according to an exemplary embodiment, such as
Shown in Fig. 1, this method comprises:
Step 101: obtaining the image of the machine fish to move about in water；
In a step 101, the image of the machine fish to move about in water can be obtained by visual sensor.
Step 102: described image being split, multiple subgraphs are obtained；
Step 103: the machine fish is identified in the multiple subgraph, with the determination machine fish in described image
In position；
In step 103, the son that current existing image recognition algorithm is successively divided in a step 102 can be used
Machine fish is identified in image.
Step 104: when can not recognize the machine fish in the multiple subgraph, controlling the machine fish
After executing deliberate action, the image of the machine fish is obtained again and determines the machine fish in the image got again
In position；
At step 104, machine fish such as can not be recognized in multiple subgraphs, then illustrates that position occurs in machine fish
The case where setting loss, the reason of such case occur is likely to be the water surface and ripple occurs, and there are special illumination or machines for the water surface
The image that the posture of fish swimming makes it be presented on the water surface is too small, in this case, machine fish can control to execute specified move
Make, so that machine fish changes its posture in water, for example, controllable machine fish is turned to the greatest extent to the left or to the right
To maximum turn to is that machine fish is limited to its physical structure, the steering locking angle degree that can be used as.In machine fish
After executing deliberate action, image is obtained again, and identify machine fish in the picture, then can increase what device fish was successfully identified
Probability.
It should be noted that at step 104, after control machine fish executes deliberate action, machine fish can be obtained again
The image to move about in water, and the image obtained again is split, multiple subgraphs are obtained, are obtained in segmentation multiple
The machine fish is identified in subgraph.
Step 105: the machine fish is obtained according to position of the machine fish obtained at least twice in described image
Motion profile.
From the above it can be seen that the method provided by the invention for obtaining robot fish movement track, to swimming in water
When the motion profile of dynamic machine fish is tracked, after being split to the image of the underwater machine fish acquired, then
Machine fish is identified in image after segmentation, the discrimination of machine fish can be improved, so as to be avoided as far as possible in the movement to machine fish
There is the problem of machine fish position loss in track during being tracked.
Fig. 2 is that multiple target positions and tracking in real time in water under a kind of Full vision shown according to an exemplary embodiment
The schematic diagram of system, the above method can be applied in the system.As shown in Fig. 2, the system includes: visual sensor (1), aluminium
Support frame and sink guardrail (2), sink (3) and machine fish (4) processed.Fig. 3 is the top view of machine fish, and Fig. 4 is machine fish
Bottom view, in conjunction with shown in Fig. 3 and Fig. 4, machine fish (4) includes: pectoral fin (5), waterproof fish-skin (6), antenna (7), charging plug
(8), aluminum skeleton (9), communication module (10), fin (11), battery (12), control system (13), the first joint (14),
Two joints (15) and third joint (16).
Wherein, it is directly connected to allow to capture water between the end of visual sensor (1) and the top of system framework
Pond all images, sink guardrail (2) are fixed on the bottom surrounding of system framework, can to played around sink (3) protection and branch
Support effect；
Machine fish (4) can move about in the inside any position of sink (3), and the posture information of its any time is by vision
Sensor (1) is captured；
Communication module (10), battery (12) are located at front end and the rigid connection of aluminum skeleton (9), are used for and the first joint
(14), second joint (15) and third joint (16) balance；
Waterproof fish-skin (6) can fully wrapped around antenna (7), aluminum skeleton (9), communication module (10), battery (12), control
System (13), the first joint (14), second joint (15) and third joint (16), so that these instruments and water segregation.
Wherein, the first joint (14), second joint (15) and third joint (16) can be multistage steering engine.
In a kind of achievable mode, described image is split can include: by the color of pixel each in image to be split
Color vector inputs the input vector of neuron as one, and the color- vector of eight pixels adjacent with each pixel is made
For each feature vector of radial function RBF；Determine the seed neuron in the image to be split, wherein pixel is to adjacent
When the maximum manhatton distance of pixel is less than first threshold, which is sub-pixel point, and the sub-pixel point is corresponding
Neuron is seed neuron, and the seed neuron in the image to be split constitutes seed region；By presetting growing strategy
Seed region is grown, multiple packet zones are obtained；Calculate the averaged feature vector of each packet zone；Using calculating
To averaged feature vector replace feature vector included in all neurons of packet zone belonging to it；If there is not
It is connected to the neuron of any packet zone, and the distance of the neuron to its adjacent packet zones is less than second threshold, then will
The neuron is connected to away from nearest packet zone；The adjacent packet zone is merged, is obtained to be split more
A region；The image to be split is split according to the multiple region, obtains the multiple subgraph.
In above-mentioned image segmentation process, for any pixel point s, meets formula (1), then can be used as sub-pixel point.
μs< θμ
Wherein, s is pixel piont mark, and j is adjacent pixel subscript, | | xs-cj| | it is manhatton distance, expression formula such as formula
(2) shown in, θμFor preset threshold (i.e. above-mentioned first threshold), μsFor pixel s and 8 pixels adjacent thereto it is maximum away from
From.
In above-mentioned image segmentation process, 8 pixels adjacent thereto are attached by pixel, in color space
In, according to above-mentioned formula (1), the maximum manhatton distance of pixel to neighbor pixel is less than preset threshold (i.e. above-mentioned first
Threshold value) when, which is sub-pixel point.As it can be seen that sub-pixel point and its neighborhood territory pixel have similar features, In
In M-PCNN image segmentation algorithm, sub-pixel point is connected to form seed region with its neighborhood territory pixel point, and seed region captures
Same characteristic features pixel constantly extends.It should be noted that the selection of sub-pixel and the extension of seed region are to carry out parallel
, which is the initial stage of region merging technique in image segmentation, repeats segmentation and is used to avoid losing the important thin of image
Section.In above-mentioned formula (1), θμThreshold value set a lesser value.It is theoretical according to color quantizing, it can be by each RGB points
17576 colors, human eye visual perception of this color resolution for 17000 colors are distinguished at 26 quantization levels
For it is sufficiently high.According to these data, θμEqual to the quantized interval radius of rgb space, the numerical value that standardizes is set as 1/52.
Image segmentation process is further described below in conjunction with Fig. 5.As shown in figure 5, image segmentation process can include:
Step 501: after inputting image to be split, using the color- vector x of each pixel of image as an input nerve
The input vector μ of member, each spy of the color- vector of 8 adjacent pixel adjacent with current pixel point as radial basis function RBF
Vector c is levied, determines initial seed point using initial point selection condition；Seed region is grown using growing strategy, will be selected
Seed point as starting point, using formula (1) as judgment condition, grown to 8 pixels adjacent with current pixel point,
Until pixel no longer meets formula (1), group areas number is obtained；
Step 502: calculating the averaged feature vector σ of each packet zoneg, such as formula (6), σ Rg, σ Bg, respectively red, green,
Blue component average value, M are the pixel number in the packet zone that number is g.Obtained averaged feature vector is replaced with into the area
Feature vector included in all neurons in domain；
Step 503: not connected neuron is judged whether there is, if there is not connected neuron, and not connected mind
Difference (distance value i.e. between the two) through member to neighboring region is less than threshold θiWhen, execute step 504 θiFor what is given
Threshold value: being connected in immediate adjacent area using formula (7), otherwise, executes step 505；
Wherein, xμIt is the feature vector for connecting pixel μ, σgjIt is the averaged feature vector that neighboring region number is gj.It is right
All not connected neurons are attached operation simultaneously, and update threshold θi+1=θi+Δθi (θi+1For new threshold value, Δ θiFor threshold
It is worth increment), repeat the above steps 502；
Step 505: zoning area RsWith color distance Rd；
Step 506: judging whether there is the area R in regionsWith color distance RdWhether R is mets< θsAnd Rd< θd, when
When meeting, step 507 is executed, otherwise, process terminates；
Step 507: obtained adjacent area being merged, and all area of space merge parallel；
Merge rule: as the area R in a regionsWith color distance Rd, meet Rs< θsAnd Rd< θdWhen, θsAnd θdRespectively
For preset area size threshold value and color distance threshold value, which is integrated into adjacent area.If the color distance of multiple neighborhoods
When meeting less than given threshold, each step has and only any one region is merged；
Step 507 is repeated, until meeting region merging technique stop condition, completes color images.
In a kind of achievable mode, the adjacent packet zone is merged can include: when the area in region to be combined
Less than preset area, and when the color distance in the region to be combined is less than pre-set color distance threshold, by the area to be combined
Domain is merged into adjacent area.
In a kind of achievable mode, determine the position of the machine fish can include: based on the target of Meanshift with
Track algorithm calculate described image in target area and candidate region in pixel characteristic value probability, obtain object module description with
And candidate family description；The similitude of the candidate family of the object module and present frame is measured using similar function；Selection
Make the maximum candidate family of the similar function and obtains Meanshift (mean shift algorithm) vector of object module；Iteration
Meanshift vector is calculated, the position of the machine fish is obtained by convergence.
In a kind of achievable mode, controls the machine fish and execute deliberate action can include: control the machine fish and press
It is turned to according to predetermined angle and preset direction.Wherein, which is, for example, that machine fish is limited to its physical structure,
The steering locking angle degree that can be used as.Preset direction can be the left or right in machine fish current kinetic direction.
Fig. 6 is the flow chart of the method shown according to an exemplary embodiment for obtaining robot fish movement track, is tied below
Fig. 6 is closed to illustrate the method for acquisition robot fish movement track of the invention.As shown in Fig. 6, this method comprises:
Step 601: obtaining visual sensor acquired image (also referred to as original image)；
Step 602: the image that step 601 obtains being split, multiple subgraphs are obtained；
Step 603: binary conversion treatment is carried out to multiple subgraphs that step 602 obtains；
Step 604: image dividing processing is carried out to each subgraph of binary conversion treatment；
Step 605: identifying machine fish in image after treatment, obtain recognition result；
Step 606: executing control program, control robot fish movement；
Step 607: judging whether the position of machine fish loses；
Step 608: if the position of machine fish is lost, determining that machine fish stops travelling；
Step 609: control machine fish turns to maximum angle to the left；
Step 610: reacquiring visual sensor acquired image, and return to step 605.
Fig. 7 is a kind of block diagram of device for obtaining robot fish movement track shown according to an exemplary embodiment, is such as schemed
Shown in 7, which includes:
First obtains module 71, for obtaining the image of the machine fish to move about in water；
Divide module 72 and obtains multiple subgraphs for being split to described image；
Identification module 73, for identifying the machine fish in the multiple subgraph, with the determination machine fish in institute
State the position in image；
Second obtains module 74, when can not recognize the machine fish in the multiple subgraph, described in control
After machine fish executes deliberate action, the image of the machine fish is obtained again and determines that the machine fish is being got again
Image in position；
Determining module 75, for determining institute according to position of the machine fish obtained at least twice in described image
State the motion profile of machine fish.
In a kind of achievable mode, the segmentation module, comprising: setting unit is used for picture each in image to be split
The color- vector of element inputs the input vector of neuron as one, by the color of eight pixel adjacent with each pixel
Each feature vector of the vector as radial function RBF；Determination unit, for determining the nerve of the seed in the image to be split
Member, wherein when the maximum manhatton distance of pixel to neighbor pixel is less than first threshold, which is sub-pixel
Point, the corresponding neuron of sub-pixel point are seed neuron, and the seed neuron in the image to be split constitutes seed
Region；Generation unit obtains multiple packet zones for growing by default growing strategy to seed region；First
Computing unit, for calculating the averaged feature vector of each packet zone；Replacement unit, for using the average spy being calculated
Sign vector replaces feature vector included in all neurons of packet zone belonging to it；Connection unit, for if there is
It is not attached to the neuron of any packet zone, and the distance of the neuron to its adjacent packet zones is less than second threshold, then
The neuron is connected to away from nearest packet zone；Combining unit, for closing the adjacent packet zone
And obtain multiple regions to be split；Cutting unit, for dividing according to the multiple region the image to be split
It cuts, obtains the multiple subgraph.
In a kind of achievable mode, the combining unit can be used for: when the area in region to be combined is less than default face
Product, and the color distance in the region to be combined be less than pre-set color distance threshold when, by the region merging technique to be combined to phase
Neighbouring region.
Optionally, the position of the machine fish is determined can include: the target tracking algorism based on Meanshift calculates institute
Pixel characteristic value probability in the target area and candidate region in image is stated, object module description is obtained and candidate family is retouched
It states；The similitude of the candidate family of the object module and present frame is measured using similar function；Selection makes the similar letter
The maximum candidate family of number simultaneously obtains the Meanshift vector of object module；Meanshift vector is iterated to calculate, convergence is passed through
Obtain the position of the machine fish.
Optionally, control the machine fish and execute deliberate action can include: control the machine fish according to predetermined angle with
And preset direction is turned to.
The device of above-described embodiment has corresponding method real for realizing method corresponding in previous embodiment
The beneficial effect of example is applied, details are not described herein.
Based on the same inventive concept, the embodiment of the invention also provides a kind of electronic equipment, including memory, processor
And the computer program that can be run on a memory and on a processor is stored, the processor is realized when executing described program
The method of robot fish movement track is obtained described in any embodiment as above.
Based on the same inventive concept, the embodiment of the invention also provides a kind of non-transient computer readable storage medium, institutes
Non-transient computer readable storage medium storage computer instruction is stated, the computer instruction is for executing the computer
The method of robot fish movement track is obtained described in any embodiment as above.
It should be understood by those ordinary skilled in the art that: the discussion of any of the above embodiment is exemplary only, not
It is intended to imply that the scope of the present disclosure (including claim) is limited to these examples；Under thinking of the invention, above embodiments
Or can also be combined between the technical characteristic in different embodiments, step can be realized with random order, and be existed such as
Many other variations of the upper different aspect of the invention, for simplicity, they are not provided in details.
In addition, to simplify explanation and discussing, and in order not to obscure the invention, in provided attached drawing
It can show or can not show and be connect with the well known power ground of integrated circuit (IC) chip and other components.In addition,
Device can be shown in block diagram form, to avoid obscuring the invention, and this has also contemplated following facts, i.e.,
The details of embodiment about these block diagram arrangements be height depend on will implementing platform of the invention (that is, these are thin
Section should be completely within the scope of the understanding of those skilled in the art).Detail (for example, circuit) is being elaborated to describe
In the case where exemplary embodiment of the present invention, it will be apparent to those skilled in the art that can there is no this
Implement the present invention in the case where a little details or in the case that these details change.Therefore, these descriptions should be by
It is considered illustrative rather than restrictive.
Although having been incorporated with specific embodiments of the present invention, invention has been described, according to retouching for front
It states, many replacements of these embodiments, modifications and variations will be apparent for those of ordinary skills.Example
Such as, discussed embodiment can be used in other memory architectures (for example, dynamic ram (DRAM)).
What the embodiment of the present invention was intended to cover fall within the broad range of appended claims all such replaces
It changes, modifications and variations.Therefore, all within the spirits and principles of the present invention, any omission for being made, modification, equivalent replacement,
Improve etc., it should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of method for obtaining robot fish movement track characterized by comprising
Obtain the image of the machine fish to move about in water；
Described image is split, multiple subgraphs are obtained；
Identify the machine fish, in the multiple subgraph with position of the determination machine fish in described image；
When can not recognize the machine fish in the multiple subgraph, controls the machine fish and execute deliberate action
Afterwards, the image of the machine fish is obtained again and determines position of the machine fish in the image got again；
The motion profile of the machine fish is determined according to position of the machine fish obtained at least twice in described image.
2. the method according to claim 1, wherein being split to described image, comprising:
The input vector that neuron is inputted using the color- vector of pixel each in image to be split as one, will be with each pixel
Each feature vector of the color- vector of eight adjacent pixels as radial function RBF；
Determine the seed neuron in the image to be split, wherein the maximum manhatton distance of pixel to neighbor pixel
When less than first threshold, which is sub-pixel point, which is seed neuron, described
Seed neuron in image to be split constitutes seed region；
Seed region is grown by default growing strategy, obtains multiple packet zones；
Calculate the averaged feature vector of each packet zone；
Using the averaged feature vector being calculated replace feature included in all neurons of packet zone belonging to it to
Amount；
If there is the neuron for being not attached to any packet zone, and the distance of the neuron to its adjacent packet zones is less than
The neuron is then connected to away from nearest packet zone by second threshold；
The adjacent packet zone is merged, multiple regions to be split are obtained；
The image to be split is split according to the multiple region, obtains the multiple subgraph.
3. according to the method described in claim 2, it is characterized in that, the adjacent packet zone is merged, comprising:
When region to be combined area be less than preset area, and the color distance in the region to be combined be less than pre-set color distance
When threshold value, by the region merging technique to be combined to adjacent area.
4. the method according to claim 1, wherein determining that the position of the machine fish includes:
Target tracking algorism based on Meanshift calculates pixel characteristic value in target area and candidate region in described image
Probability obtains object module description and candidate family description；
The similitude of the candidate family of the object module and present frame is measured using similar function；
Selection makes the maximum candidate family of the similar function and obtains the Meanshift vector of object module；
Meanshift vector is iterated to calculate, the position of the machine fish is obtained by convergence.
5. method according to any one of claims 1 to 4, which is characterized in that it controls the machine fish and executes deliberate action,
Include:
The machine fish is controlled to be turned to according to predetermined angle and preset direction.
6. a kind of device for obtaining robot fish movement track characterized by comprising
First obtains module, for obtaining the image of the machine fish to move about in water；
Divide module and obtains multiple subgraphs for being split to described image；
Identification module, for identifying the machine fish in the multiple subgraph, with the determination machine fish in described image
In position；
Second acquisition module controls the machine fish when can not recognize the machine fish in the multiple subgraph
After executing deliberate action, the image of the machine fish is obtained again and determines the machine fish in the image got again
Position；
Determining module, for determining the machine fish according to position of the machine fish obtained at least twice in described image
Motion profile.
7. device according to claim 6, which is characterized in that the segmentation module, comprising:
Setting unit, for inputting the input vector of neuron using the color- vector of pixel each in image to be split as one,
Using the color- vector of eight pixel adjacent with each pixel as each feature vector of radial function RBF；
Determination unit, for determining the seed neuron in the image to be split, wherein pixel to neighbor pixel most
When big manhatton distance is less than first threshold, which is sub-pixel point, which is kind
Sub- neuron, the seed neuron in the image to be split constitute seed region；
Generation unit obtains multiple packet zones for growing by default growing strategy to seed region；
First computing unit, for calculating the averaged feature vector of each packet zone；
Replacement unit, for using the averaged feature vector being calculated to replace institute in all neurons of packet zone belonging to it
The feature vector for including；
Connection unit, for if there is the neuron for being not attached to any packet zone, and the neuron is to its adjacent packets
The distance in region is less than second threshold, then the neuron is connected to away from nearest packet zone；
Combining unit obtains multiple regions to be split for merging the adjacent packet zone；
Cutting unit obtains the multiple subgraph for being split according to the multiple region to the image to be split.
8. device according to claim 7, which is characterized in that the combining unit is used for:
When region to be combined area be less than preset area, and the color distance in the region to be combined be less than pre-set color distance
When threshold value, by the region merging technique to be combined to adjacent area.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor realizes such as acquisition described in any one of claim 1 to 5 when executing described program
The method of robot fish movement track.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction is for making the computer perform claim require 1 to 5 described in any item acquisition machines
The method of device fish movement track.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910410101.9A CN110517287A (en) | 2019-05-17 | 2019-05-17 | Obtain method, apparatus, equipment and the storage medium of robot fish movement track |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910410101.9A CN110517287A (en) | 2019-05-17 | 2019-05-17 | Obtain method, apparatus, equipment and the storage medium of robot fish movement track |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110517287A true CN110517287A (en) | 2019-11-29 |
Family
ID=68622512
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910410101.9A Pending CN110517287A (en) | 2019-05-17 | 2019-05-17 | Obtain method, apparatus, equipment and the storage medium of robot fish movement track |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110517287A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140266205A1 (en) * | 2013-03-12 | 2014-09-18 | MRI Interventions, Inc. | Intra-body medical devices for use in mri environments |
| CN104599262A (en) * | 2014-12-18 | 2015-05-06 | 浙江工业大学 | Multichannel pulse coupling neural network based color image segmentation technology |
| CN104931091A (en) * | 2015-06-24 | 2015-09-23 | 金陵科技学院 | Bionic robot fish measuring platform and using method thereof |
| CN107186708A (en) * | 2017-04-25 | 2017-09-22 | 江苏安格尔机器人有限公司 | Trick servo robot grasping system and method based on deep learning image Segmentation Technology |
| CN109241985A (en) * | 2017-07-11 | 2019-01-18 | 普天信息技术有限公司 | A kind of image-recognizing method and device |
-
2019
- 2019-05-17 CN CN201910410101.9A patent/CN110517287A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140266205A1 (en) * | 2013-03-12 | 2014-09-18 | MRI Interventions, Inc. | Intra-body medical devices for use in mri environments |
| CN104599262A (en) * | 2014-12-18 | 2015-05-06 | 浙江工业大学 | Multichannel pulse coupling neural network based color image segmentation technology |
| CN104931091A (en) * | 2015-06-24 | 2015-09-23 | 金陵科技学院 | Bionic robot fish measuring platform and using method thereof |
| CN107186708A (en) * | 2017-04-25 | 2017-09-22 | 江苏安格尔机器人有限公司 | Trick servo robot grasping system and method based on deep learning image Segmentation Technology |
| CN109241985A (en) * | 2017-07-11 | 2019-01-18 | 普天信息技术有限公司 | A kind of image-recognizing method and device |
Non-Patent Citations (1)
| Title |
|---|
| 张鑫 等: "UAV目标跟踪预测算法研究", 《计算机与数学工程》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105245841B (en) | A kind of panoramic video monitoring system based on CUDA | |
| CN110782477A (en) | Moving target rapid detection method based on sequence image and computer vision system | |
| CN106296725A (en) | Moving target detects and tracking and object detecting device in real time | |
| CN101587591B (en) | Visual accurate tracking technique based on double parameter thresholds dividing | |
| US10572762B2 (en) | Image processing method for performing pattern matching for detecting a position of a detection target | |
| CN103426182A (en) | Electronic image stabilization method based on visual attention mechanism | |
| CN106709878B (en) | A kind of rapid image fusion method | |
| CN112184604B (en) | Color image enhancement method based on image fusion | |
| CN106530267B (en) | A kind of fusion method for avoiding panoramic picture from misplacing | |
| CN108875621A (en) | Image processing method, device, computer equipment and storage medium | |
| CN106485658A (en) | A kind of optimum suture line search method for image mosaic | |
| CN110544269A (en) | twin network infrared target tracking method based on characteristic pyramid | |
| CN112837344A (en) | Target tracking method for generating twin network based on conditional confrontation | |
| JP2003271933A (en) | Face detector, face detecting method, and robot device | |
| CN107351080A (en) | A kind of hybrid intelligent research system and control method based on array of camera units | |
| CN114979489A (en) | Gyroscope-based heavy equipment production scene video monitoring and image stabilizing method and system | |
| CN109949338A (en) | A kind of target following recognition methods based on the bionical underwater fish of CPG model | |
| CN106780326A (en) | A kind of fusion method for improving panoramic picture definition | |
| Shao et al. | Real-time tracking of moving objects on a water surface | |
| CN102426478A (en) | Information processing device, information processing method, program and electronic apparatus | |
| CN109658441A (en) | Foreground detection method and device based on depth information | |
| CN110517287A (en) | Obtain method, apparatus, equipment and the storage medium of robot fish movement track | |
| Aziz et al. | Evaluation of visual attention models for robots | |
| CN106485731B (en) | A kind of picture position method for registering based on improvement Hough transformation | |
| CN108171746A (en) | The club bead alignment system and method for a kind of view-based access control model |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191129 |
|
| RJ01 | Rejection of invention patent application after publication |