CN105787962B - A kind of monocular vision tracking recycled under water based on UUV - Google Patents
A kind of monocular vision tracking recycled under water based on UUV Download PDFInfo
- Publication number
- CN105787962B CN105787962B CN201610104508.5A CN201610104508A CN105787962B CN 105787962 B CN105787962 B CN 105787962B CN 201610104508 A CN201610104508 A CN 201610104508A CN 105787962 B CN105787962 B CN 105787962B
- Authority
- CN
- China
- Prior art keywords
- target
- light source
- frame
- heart
- uuv
- 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.)
- Active
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
Abstract
A kind of monocular vision tracking recycled under water based on UUV, the present invention relates to the monocular vision trackings recycled under water based on UUV.The purpose of the present invention is to solve existing general visual tracking methods, and the low problem of accuracy is tracked in the case where UUV is recycled under water.It is realized by following steps：One, the sequence image of UUV camera acquisitions target light source system；Two, target weighted model is established；Three, candidate target model is calculated in the current frame；Four, Bhattacharyya likeness coefficients are sought；Five, weight coefficient is calculated；Six, the new position in candidate target center is obtained according to weight coefficient；Seven, ρ (y are obtained1)；Eight, when | | y1‑y0||<Then stop when ε；Otherwise make y0=y1, three are returned to, present frame terminates if it is last frame, then tracking；Otherwise continue to read next frame.The present invention is applied to the monocular vision tracking field that UUV is recycled under water.
Description
Technical field
The present invention relates to the monocular vision trackings recycled under water based on UUV.
Background technology
When UUV executes task in ocean, since the limitation of battery capacity makes its recycling particularly important, and the water of UUV
Lower recycling has many advantages, such as to be jolted by stormy waves, and small, concealment is high, strategic value is high；Meanwhile way of recycling of carrying a load on the back under water is by sea
Depressed place cabin platform that midocean or submarine carry recycles UUV, and which does not require boom hoisting and the operation of personnel, has
Vast potential for future development.Video camera has been widely used on UUV, other sensors is compared, with short distance positioning accuracy
Height, the advantages that containing much information, be adaptable, high resolution and cost are smaller of acquisition.
In UUV is recycled under water, monocular vision has many advantages, such as that simple installation, precision are high compared to binocular vision.It considers
Common objects feature has prodigious decaying in water；In practical marine environment, simple Vision Tracking is extremely difficult to
It is required that；And mean shift algorithm calculation amount is small, is suitable for real-time tracking occasion, but it converges on Local Extremum, certain
It cannot be guaranteed that the accuracy of tracking in the case of background interference.
For the underwater recycling docking facilities that the ISiMI AUV that South Korea develops are used for funnel-form, periphery is equipped with 5 guidings
Light source carries out processing by the image to shooting and obtains the relative poses of docking facilities, and by visual servo by AUV track homings
To docking facilities；Optical flow can be used for detect and track moving target, Midland, MI university Negahdaripour et al.
Explore using optical flow carry out estimation underwater robot movement and keep hovering method, but there are measurement error when
Between accumulation and pose drift situation, cause existing general visual tracking method to track standard in the case where UUV is recycled under water
True property is low.
Invention content
The purpose of the present invention is to solve existing general visual tracking methods to be tracked in the case where UUV is recycled under water
The low problem of accuracy, and propose a kind of monocular vision tracking recycled under water based on UUV.
Above-mentioned goal of the invention is achieved through the following technical solutions：
Step 1: when being appeared in the visual field of UUV video cameras during target light source system motion, UUV video cameras are adopted
Collect the sequence image of target light source system；
Step 2: heart-shaped light source target is chosen to the first frame image of the sequence image of collected target light source system,
Establish target weighted model
Step 3: calculating candidate target model in the current frame
Step 4: according to the weighted model for choosing heart-shaped light source targetWith present frame candidate target modelIt seeks
Bhattacharyya likeness coefficients；
Step 5: calculating weight coefficient w according to Bhattacharyya likeness coefficientsi；
Step 6: according to weight coefficient wiObtain the new position y in candidate target center1；
Step 7: calculating the candidate target new position y in center1Model { pu(y1)}U=1 ..., m, obtain Bhattacharyya phases
Like property coefficientCompare ρ (y1) and ρ (y0) size, as ρ (y0)>ρ(y1) when,It obtains
ρ(y1)；
Step 8: working as | | y1-y0||<When ε, then stop, ε is the minimum threshold of iteration convergence；Otherwise, make y0=y1, return
To step 3, present frame terminates if it is last frame, then tracking；Otherwise continue to read next frame.
Invention effect
In view of common objects feature has prodigious decaying in water, what depressed place cabin platform was recycled using light source as UUV
Target is guided, interference can be effective against and improves the effect of UUV guiding recycling；The present invention uses a kind of linear target light source battle array
Row, light source arrangement signature carry out target light source when the depressed place cabin platform of movement appears in the visual field of UUV video cameras
Vision tracks, and further completes the recycling of UUV.In practical marine environment, simple Vision Tracking is extremely difficult to want
It asks；And mean shift algorithm calculation amount is small, is suitable for real-time tracking occasion, but it converges on Local Extremum, in certain back ofs the body
It cannot be guaranteed that the accuracy of tracking under scape disturbed condition.In view of actual conditions, using the average drifting side weighted based on target
Method to ensure the accuracy of tracking into line trace to submarine target.
In UUV carrys a load on the back removal process under water, guiding target light source system is installed, when target light source system on the platform of depressed place cabin
When system is appeared in the visual field of video camera, using the average drifting tracking weighted based on target, to light-source system center
Heart-shaped light source carries out vision tracking.By establishing the model of target weighting to heart-shaped light source, the different parts of light source are assigned big
Small not equal weights, and being added in the mathematical model of average drifting track algorithm, the present invention can improve vision tracking accuracy,
Improve tracking effect, as shown in fig. 7, by a kind of monocular vision tracking recycled under water based on UUV using the present invention with
The experimental result of traditional average drifting track algorithm is compared, the method for the present invention and the drift of traditional mean value when 5 frame
The deviation for moving track algorithm is identical；But the deviation of the method for the present invention is 2 when 10 frame, traditional average drifting track algorithm
Deviation be 7；The deviation of the method for the present invention is 1.7 when 20 frame, and the deviation of traditional average drifting track algorithm is 7；70
The deviation of the method for the present invention is 1.7 when frame, and the deviation of traditional average drifting track algorithm is 11.5；It was tracked entirely
Tracking image can be accurately positioned in Cheng Zhong, a kind of monocular vision tracking recycled under water based on UUV of the invention, improve with
Track accuracy.
Description of the drawings
Fig. 1 is UUV monocular-camera pictorial diagrams；
Fig. 2 is target light source system structure diagram；
Fig. 3 is heart-shaped light source schematic diagram；
Fig. 4 is that the present invention is based on the average drifting tracking flow charts that target weights；
Fig. 5 a are for traditional mean shift algorithm in 4 frame to the tracking effect figure of heart-shaped light source；
Fig. 5 b are for traditional mean shift algorithm in 24 frame to the tracking effect figure of heart-shaped light source；
Fig. 5 c are for traditional mean shift algorithm in 45 frame to the tracking effect figure of heart-shaped light source；
Fig. 5 d are for traditional mean shift algorithm in 70 frame to the tracking effect figure of heart-shaped light source；
Fig. 6 a are for the method for the present invention in 4 frame to the tracking effect figure of heart-shaped light source；
Fig. 6 b are for the method for the present invention in 24 frame to the tracking effect figure of heart-shaped light source；
Fig. 6 c are for present invention side in 45 frame to the tracking effect figure of heart-shaped light source；
Fig. 6 d are for the method for the present invention in 70 frame to the tracking effect figure of heart-shaped light source；
Fig. 7 is that the method for the present invention and the tracing positional deviation of traditional mean shift algorithm compare.
Specific implementation mode
Specific implementation mode one：A kind of monocular vision tracking recycled under water based on UUV of present embodiment, specifically
It prepares according to the following steps：
Step 1: appearing in the visual field of UUV (underwater unmanned vehicle) video camera during target light source system motion
When middle, the sequence image of UUV camera acquisition target light source systems；
(target light source system：In view of UUV is carried a load on the back the actual conditions of way of recycling under water, the present invention using it is a kind of towards
To the linear light source array of line traffic control position, " to line traffic control position " principle is that straight line is established by 2 points, and control UUV makes transverse direction, longitudinal direction, bow
To reaching position and posture.The target light source system is characterized as：
(1), point-blank, adjacent light source spacing is equal and is fixed value for the central point arrangement of 9 light sources；
(2), light-source system is centrosymmetric, and heart-shaped light source is located at center, and left and right is respectively uniformly distributed 4 light sources；
(3), heart-shaped light source at the center of linear array of source and is located at depressed place cabin Platform center；
(4), heart-shaped light source includes 4 LED light, wherein 1 does not shine, the heart-shaped light source of remaining 3 compositions.
All light sources do water-proofing treatment, the present invention realize the vision in UUV recycling under water to heart-shaped target light source with
Track.)
Step 2: minimum threshold ε=0.1 of selected iteration convergence, maximum iterations N=10, to collected mesh
The first frame image for marking the sequence image of light-source system chooses heart-shaped light source target (to first frame image manually using rectangle frame choosing
Coring shape light source target, and establish the model of target weighting) establish target weighted model
Step 3: calculating candidate target model in the current frame
Step 4: according to the weighted model for choosing heart-shaped light source targetWith present frame candidate target modelIt seeks
Bhattacharyya likeness coefficients；
Step 5: calculating weight coefficient w according to Bhattacharyya likeness coefficientsi；
Step 6: according to weight coefficient wiObtain the new position y in candidate target center1；
Step 7: calculating the candidate target new position y in center1Model { pu(y1)}U=1 ..., m, obtain Bhattacharyya phases
Like property coefficientCompare ρ (y1) and ρ (y0) size, as ρ (y0)>ρ(y1) when,It obtains
ρ(y1)；
Step 8: working as | | y1-y0||<When ε, then stop, ε is the minimum threshold of iteration convergence；Otherwise, make y0=y1, return
To step 3, present frame terminates if it is last frame, then tracking；Otherwise continue to read next frame.
Specific implementation mode two：The present embodiment is different from the first embodiment in that：It is selected repeatedly in the step 2
For convergent minimum threshold ε=0.1, maximum iterations N=10, to the sequence image of collected target light source system
First frame image chooses heart-shaped light source target and (uses rectangle frame to choose heart-shaped light source target manually first frame image, and establish
The model of target weighting) establish target weighted modelDetailed process is：
Heart-shaped light source target is initialized with rectangle frame, and with the movement of this label target during tracking.
Heart-shaped light source mesh is chosen using rectangle frame to the first frame image of the sequence image of collected target light source system
Mark, and assume that rectangle frame centre coordinate is (Ox,Oy), wherein a and b is respectively the 1/2 of rectangle frame length and width；To make calculating
Simpler and target model features keep stablizing, by pixel in rectangle frameIt normalizes in a unit circle, in rectangle frame
PixelCoordinate is (xi,yi)；
It is 1 that point weights at heart-shaped light source target center are chosen in definition, chooses heart-shaped light source target outermost weights and is intended to
0, (it is different to assign point different in heart-shaped light source target (being to establish its object module to heart-shaped light source, then into line trace)
Weights, make the maximum weight for the model center that target weights, distance center is remoter, and the weights of imparting are smaller) choose heart-shaped light
The weights put between outermost are at the target's center of source：
According to mean shift algorithm, (traditional mean shift algorithm substantially belongs to feature based track algorithm, utilizes target
The histogram in region indicates its feature space, then passes through target area template and candidate target region template
Bhattacharyya similar values obtain best candidate target area using mean shift vectors iteration.It is as follows：
It is respectively h using 1 height and lengthx、hyTrack window initialized target, the track window simultaneously be also kernel function
The place of effect, length and width are its bandwidth.Since target area can be described in histogram, if target image is by n section
Interior m grades of gray-scale pixels composition, each feature space are also designated as bin, and the pixel of target area is { xi}I=1 ... n, b (xi) description
xiThe index value of the pixel histogram at place, x*Indicate that the center of target area, the model of target can be expressed as q={ qu}U=1 ... m,
WhereinPixel coordinate x and y in target zone is passed through into hxAnd hyIt normalizes, the range of k (x) is set as
1, then characteristic value q in object moduleuThe density fonction of estimation is expressed as：
Wherein, δ indicates Kronecker delta functions, δ [b (xi)-u] as pixel value xiWhen belonging to u-th of characteristic value
It is 1, is on the contrary 0.K (x) indicates that kernel function, core window width h are the 1/2 of target area size under normal conditions.C is that normalization is normal
Number, in order to ensure
Candidate target region refers to that mesh target area is indicated in following image sequence, if its center is y,Indicate the location of pixels of target area in present frame, nhIndicate sum of all pixels in region of search.It is similar with formula (3), mesh
Mark feature p in modelu(y) density fonction estimated can be expressed as
Wherein, in order to ensureNormaliztion constant
Bhattacharyya likeness coefficients can state the similitude size between two templates, be represented by
Target-region locating is searching in the picture and the maximum region of object module similitude, that is, is made
Bhattacharyya coefficients are maximum, and the new position y of candidate target region is obtained by mean shift vectors iteration1
Wherein, weight coefficient wiFor
If n is the pixel sum in rectangle frame, m is grey level histogram grid
Sum,It indicatesPixel histogram index value, x*It indicates to choose heart-shaped light source target center, chooses heart-shaped light source mesh
Target model is expressed asAndThe weighted model for then choosing heart-shaped light source target is
Wherein, k (x) indicates that kernel function, h indicate kernel function bandwidth, usually the 1/2 of the target area length of side, using selection
Epanechnikov kernel functions, i.e.,
Wherein, c is constant；
C1For normaliztion constant, in order to ensureThenx*Indicate choosing
Coring shape light source target center；Probability characteristics u=1,2 ..., m；quIt indicates to choose heart-shaped light source target color histogram grid spy
U-th of component in sign vector；δ indicates Kronecker delta functions (Kronecker function)；Working as
It is 1 when belonging to u-th of characteristic value, is on the contrary 0.
Other steps and parameter are same as the specific embodiment one.
Specific implementation mode three：The present embodiment is different from the first and the second embodiment in that：In the step 3
Candidate target model is calculated in present frameDetailed process is：
It may include mesh target area that candidate target region, which refers in each frame, and similar with (1) formula, candidate target model indicates
For：
Wherein, the center of candidate target region is y, can be initialized as in the heart-shaped light source target of previous frame image
The heart；Indicate the pixel coordinate of candidate target region in the current frame, Ch1For normaliztion constant；nhIndicate search window
Interior sum of all pixels；J value ranges are 1≤j≤nh；
Wherein,
Other steps and parameter are the same as one or two specific embodiments.
Specific implementation mode four：Unlike one of present embodiment and specific implementation mode one to three：The step 4
The middle weighted model according to the heart-shaped light source target of selectionWith present frame candidate target modelSeek Bhattacharyya
Likeness coefficient；
With Bhattacharyya likeness coefficients, (similarity function is used for judging object module in present frame and next
Matching degree between candidate target model in frame, Bhattacharyya likeness coefficients belong to divergence form measurement, the original that it is indicated
Reason meaning is to solve for the cosine value of angle between two vectors, its utilization is with the obvious advantage in Mean Shift algorithms, can indicate
ForWherein, Bhattacharyya coefficients (BShi coefficients) indicate to press from both sides cosine of an angle between two vectors in formula
Value, when the two angle is 0, likeness coefficient value is big) it describes to choose the weighted model of heart-shaped light source targetWith work as
Previous frame candidate target modelSimilarity degree, Bhattacharyya likeness coefficients are expressed as
The cosine value of angle between two m dimensional vectors, when two vector angles are zero, two vectors phase the most
Seemingly.
Other steps and parameter are identical as one of specific implementation mode one to three.
Specific implementation mode five：Unlike one of present embodiment and specific implementation mode one to four：The step 5
It is middle that weight coefficient w is calculated according to Bhattacharyya likeness coefficientsi；
Positioning candidate target center is to find and target weighted model in the current frameThe maximum candidate of likeness coefficient
Target area, if the center of the candidate target region of previous frame is y0, Bhattacharyya likeness coefficients are existedPlace carries out Taylor series expansion：
Wherein, ChFor normaliztion constant；For the candidate target model of previous frame；
Obtain weight coefficient
Other steps and parameter are identical as one of specific implementation mode one to four.
Specific implementation mode six：Unlike one of present embodiment and specific implementation mode one to five：The step 6
It is middle according to weight coefficient wiObtain the new position y in candidate target center1；y1Formula be：
Other steps and parameter are identical as one of specific implementation mode one to five.
Beneficial effects of the present invention are verified using following embodiment：
Embodiment one：
A kind of monocular vision tracking recycled under water based on UUV of the present embodiment is specifically prepared according to the following steps
's：
It is closely underwater in UUV in conjunction with a kind of 4 monocular vision tracking recycled under water based on UUV of the present invention of attached drawing
In removal process, when target light source system (such as attached drawing 2) appears in the visual field of UUV monocular-cameras (such as attached drawing 1), camera shooting
Machine collects sequence image, and heart-shaped light source (such as attached drawing 3) position that first frame is chosen using rectangle frame carries out vision tracking, if square
Shape frame centre coordinate (Ox,Oy), target area pixel is(coordinate is (xi,yi)), a and b are respectively rectangle frame length and width
The 1/2 of degree.Specifically comprise the following steps:
The first step：Minimum threshold ε=0.1 of selected iteration convergence, maximum iterations n=10, in initial frame, really
The initial position of vertical heart light source, defining arbitrary point weights in target area is：
Model { the q of target weighting is established to heart-shaped light sourceu}U=1 ... m：
Wherein, x*Indicate that the center of target area, k (x) indicate kernel function, h indicates kernel function bandwidth, using selection
Epanechnikov kernel functions, i.e.,
Second step：Candidate target model { p is calculated in the current frameu(y0)}U=1 ..., m：
Wherein, the center of candidate target model isIndicate the pixel of target area in the current frame
Coordinate, nhIndicate the sum of all pixels in region of search.
And then obtain Bhattacharyya likeness coefficients
Third walks：ByTo calculate weight coefficient.
4th step：The new position y in candidate target center is obtained by following formula1：
5th step：Calculate { pu(y1)}U=1 ..., m, obtain Bhattacharyya coefficients
Compare ρ (p (y0), q) and ρ (p (y1), q) size, as ρ (p (y0),q)>ρ(p(y1), q) when,Obtain ρ (p
(y1),q)。
6th step：When | | y1-y0||<When ε, then stop；Otherwise, make y0=y1, return to second step.
7th step：Present frame terminates if it is last frame, then tracking.Otherwise it is present frame to read next frame.
Example two：
It chooses underwater picture sequence and carries out experimental verification, '+' number indicates tracking rectangle frame center, choose the heart of first frame
Light source target position is marked into line trace and with rectangle frame, choose respectively image sequence the 4th, 24,45,70 frames.Attached drawing 5a,
5b, 5c and 5d are tracking result of the heart-shaped light source target under traditional mean shift process, and tracking box drift occurs in when 24 frame
The situation of shifting, drift conditions always exist in image sequence behind therewith, and target light source is tracked in 70 frame and has been gone out
Very serious drift conditions are showed；Attached drawing 6a, 6b, 6c and 6d are heart-shaped light source target in the average drifting side weighted based on target
Tracking result under method.Average drifting tracking based on target weighting keeps target signature more preferable due to the introducing of weights
Ground protrudes during tracking, so can be easier to iteration finds target, shows good tracking effect.From 7 hairs of attached drawing
Bright method and the tracing positional deviation of traditional mean shift algorithm compare it is found that during entire tracking, are weighted based on target
Mean shift process tracking image can be accurately positioned, obtained tracking effect well.
The present invention can also have other various embodiments, without deviating from the spirit and substance of the present invention, this field
Technical staff makes various corresponding change and deformations in accordance with the present invention, but these corresponding change and deformations should all belong to
The protection domain of appended claims of the invention.
Claims (4)
1. a kind of monocular vision tracking recycled under water based on UUV, it is characterised in that：The method includes：
Step 1: when being appeared in the visual field of UUV video cameras during target light source system motion, UUV camera acquisition mesh
Mark the sequence image of light-source system；
Step 2: choosing heart-shaped light source target to the first frame image of the sequence image of collected target light source system, establish
Target weighted model
Step 3: calculating present frame candidate target model in the current frame
Step 4: according to target weighted modelWith present frame candidate target modelSeek Bhattacharyya similitudes
Coefficient；
U is probability characteristics, and value u=1,2 ..., m, m are grey level histogram grid sum；Y is in candidate target region
Heart position；
Step 5: calculating weight coefficient w according to Bhattacharyya likeness coefficientsi；
The value of i is 1-n, and n is the pixel sum in rectangle frame；
Step 6: according to weight coefficient wiObtain the new position y in candidate target center1；
Step 7: calculating the candidate target new position y in center1Model { pu(y1)}U=1 ..., m, obtain Bhattacharyya similitudes
CoefficientCompare ρ (y1) and ρ (y0) size, as ρ (y0) > ρ (y1) when,Obtain ρ
(y1)；
quIt indicates to choose u-th of component in heart-shaped light source target color histogram grid feature vector；
y0For the center of the candidate target region of previous frame；
Step 8: working as | | y1-y0| | when < ε, then stop, ε is the minimum threshold of iteration convergence；Otherwise, make y0=y1, back to step
Rapid three, present frame terminates if it is last frame, then tracking, otherwise continues to read next frame；
Heart-shaped light source target is chosen to the first frame image of the sequence image of collected target light source system in the step 2,
Establish target weighted modelDetailed process is：
Heart-shaped light source target is chosen using rectangle frame to the first frame image of the sequence image of collected target light source system, and
It is assumed that rectangle frame centre coordinate is (Ox,Oy), wherein a and b is respectively the 1/2 of rectangle frame length and width；By picture in rectangle frame
Vegetarian refreshmentsIt normalizes in a unit circle, pixel in rectangle frameCoordinate is (xi,yi)；
It is 1 that point weights at heart-shaped light source target center are chosen in definition, chooses heart-shaped light source target outermost weights and is intended to 0, choosing
The weights put between point and outermost at coring shape light source target center are：
According to mean shift algorithm, if n is the pixel sum in rectangle frame, m is grey level histogram grid sum,ForPixel histogram index value, x*To choose heart-shaped light source target center, the model for choosing heart-shaped light source target is expressed asAndThe weighted model for then choosing heart-shaped light source target is
Wherein, k () indicates that kernel function, h indicate kernel function bandwidth, is the 1/2 of the target area length of side, using selection
Epanechnikov kernel functions, i.e.,
Wherein, c is constant；
C1For normaliztion constant, in order to ensureThenx*It indicates to choose the heart
Shape light source target center；δ indicates Kronecker delta functions；Working asIt is when belonging to u-th of characteristic value
1, it is otherwise 0；
Candidate target model is calculated in the step 3 in the current frameDetailed process is：
It includes mesh target area that candidate target region, which refers in each frame, and candidate target model is expressed as：
Wherein, the center of candidate target region is y, can be initialized as the heart-shaped light source target center of previous frame image；Indicate the pixel coordinate of candidate target region in the current frame, Ch1For normaliztion constant；nhIt indicates in search window
Sum of all pixels；J value ranges are 1≤j≤nh；
Wherein,
2. a kind of monocular vision tracking recycled under water based on UUV according to claim 1, it is characterised in that：It is described
According to the weighted model for choosing heart-shaped light source target in step 4With present frame candidate target modelIt seeks
Bhattacharyya likeness coefficients；
The weighted model of the heart-shaped light source target of selection is described with Bhattacharyya likeness coefficientsWith present frame candidate target
ModelSimilarity degree, Bhattacharyya likeness coefficients are expressed as
Wherein,The cosine value of angle between two m dimensional vectors, when two vector angles are zero, two vectors phase the most
Seemingly.
3. a kind of monocular vision tracking recycled under water based on UUV according to claim 2, it is characterised in that：It is described
Weight coefficient w is calculated according to Bhattacharyya likeness coefficients in step 5i；
Positioning candidate target center is to find and target weighted model in the current frameThe maximum candidate target area of likeness coefficient
Domain, if the center of the candidate target region of previous frame is y0, Bhattacharyya likeness coefficients are existedPlace into
Row Taylor series expansion：
Wherein, Ch1For normaliztion constant；For the candidate target model of previous frame；
Obtain weight coefficient
4. a kind of monocular vision tracking recycled under water based on UUV according to claim 3, it is characterised in that：It is described
According to weight coefficient w in step 6iObtain the new position y in candidate target center1；y1Formula be：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610104508.5A CN105787962B (en) | 2016-02-25 | 2016-02-25 | A kind of monocular vision tracking recycled under water based on UUV |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610104508.5A CN105787962B (en) | 2016-02-25 | 2016-02-25 | A kind of monocular vision tracking recycled under water based on UUV |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105787962A CN105787962A (en) | 2016-07-20 |
| CN105787962B true CN105787962B (en) | 2018-10-30 |
Family
ID=56403660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610104508.5A Active CN105787962B (en) | 2016-02-25 | 2016-02-25 | A kind of monocular vision tracking recycled under water based on UUV |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105787962B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107065882B (en) * | 2017-05-17 | 2020-04-03 | 哈尔滨工程大学 | Method for dynamically and autonomously recovering UUV on water surface of USV |
| CN108170976B (en) * | 2018-01-11 | 2021-06-01 | 哈尔滨工程大学 | Safety analysis method in AUV dynamic recovery process of underwater submarine |
| CN108415441B (en) * | 2018-03-05 | 2020-03-10 | 中国海洋大学 | Following method of underwater robot target following system based on monocular vision |
| CN109238291B (en) * | 2018-10-26 | 2019-07-12 | 河海大学 | A kind of planing method of water surface unmanned boat guiding cable recycling Autonomous Underwater Vehicle |
| CN110686669B (en) * | 2019-09-23 | 2021-03-30 | 中国海洋大学 | ROV sea cucumber distribution statistical method and device based on positioning compensation and visual perception |
| CN112013774B (en) * | 2019-09-30 | 2021-10-22 | 中国科学院西安光学精密机械研究所 | Distance measuring system and distance measuring method |
| CN113139986A (en) * | 2021-04-30 | 2021-07-20 | 东风越野车有限公司 | Integrated environment perception and multi-target tracking system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010041034A1 (en) * | 2008-10-09 | 2010-04-15 | Isis Innovation Limited | Visual tracking of objects in images, and segmentation of images |
| CN103116896A (en) * | 2013-03-07 | 2013-05-22 | 中国科学院光电技术研究所 | Visual saliency model based automatic detecting and tracking method |
| CN103345751A (en) * | 2013-07-02 | 2013-10-09 | 北京邮电大学 | Visual positioning method based on robust feature tracking |
| CN103903279A (en) * | 2014-03-21 | 2014-07-02 | 上海大学 | Parallel tracking system and method based on bionic binocular vision onboard platform |
| CN104298996A (en) * | 2014-08-08 | 2015-01-21 | 中国科学院自动化研究所 | Underwater active vision tracking method applied to bionic robot fish |
-
2016
- 2016-02-25 CN CN201610104508.5A patent/CN105787962B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010041034A1 (en) * | 2008-10-09 | 2010-04-15 | Isis Innovation Limited | Visual tracking of objects in images, and segmentation of images |
| CN103116896A (en) * | 2013-03-07 | 2013-05-22 | 中国科学院光电技术研究所 | Visual saliency model based automatic detecting and tracking method |
| CN103345751A (en) * | 2013-07-02 | 2013-10-09 | 北京邮电大学 | Visual positioning method based on robust feature tracking |
| CN103903279A (en) * | 2014-03-21 | 2014-07-02 | 上海大学 | Parallel tracking system and method based on bionic binocular vision onboard platform |
| CN104298996A (en) * | 2014-08-08 | 2015-01-21 | 中国科学院自动化研究所 | Underwater active vision tracking method applied to bionic robot fish |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105787962A (en) | 2016-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105787962B (en) | A kind of monocular vision tracking recycled under water based on UUV | |
| Wu et al. | Vision-based real-time aerial object localization and tracking for UAV sensing system | |
| CN107818326A (en) | A kind of ship detection method and system based on scene multidimensional characteristic | |
| CN110782481A (en) | Unmanned ship intelligent decision method and system | |
| CN111461023A (en) | Method for quadruped robot to automatically follow pilot based on three-dimensional laser radar | |
| CN106780560B (en) | Bionic robot fish visual tracking method based on feature fusion particle filtering | |
| CN101916446A (en) | Gray level target tracking algorithm based on marginal information and mean shift | |
| Aykin et al. | On feature extraction and region matching for forward scan sonar imaging | |
| CN111126116A (en) | Unmanned ship river channel garbage identification method and system | |
| Zhang et al. | Research on unmanned surface vehicles environment perception based on the fusion of vision and lidar | |
| Liu et al. | Real-time monocular obstacle detection based on horizon line and saliency estimation for unmanned surface vehicles | |
| Zhan et al. | Effective waterline detection for unmanned surface vehicles in inland water | |
| CN110490903B (en) | Multi-target rapid capturing and tracking method in binocular vision measurement | |
| Hashmani et al. | A survey on edge detection based recent marine horizon line detection methods and their applications | |
| CN113792593A (en) | Underwater close-range target identification and tracking method and system based on depth fusion | |
| US11948344B2 (en) | Method, system, medium, equipment and terminal for inland vessel identification and depth estimation for smart maritime | |
| Shi et al. | Obstacle type recognition in visual images via dilated convolutional neural network for unmanned surface vehicles | |
| Zhou et al. | A real-time algorithm for visual detection of high-speed unmanned surface vehicle based on deep learning | |
| CN112417948B (en) | Method for accurately guiding lead-in ring of underwater vehicle based on monocular vision | |
| Li et al. | Vision-based target detection and positioning approach for underwater robots | |
| Deng et al. | Underwater circular object positioning system based on monocular vision | |
| Sun et al. | Autonomous underwater vehicle docking system for energy and data transmission in cabled ocean observatory networks | |
| CN105931268A (en) | Mean shift tracking method based on scale adaption in UUV underwater recovery process | |
| Xu et al. | An effective stereo SLAM with high-level primitives in underwater environment | |
| Duarte et al. | Multiple vessel detection in harsh maritime environments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |