CN110363777A - A kind of sea image semantic segmentation method based on reducible space constraint mixed model - Google Patents
A kind of sea image semantic segmentation method based on reducible space constraint mixed model Download PDFInfo
- Publication number
- CN110363777A CN110363777A CN201910592087.9A CN201910592087A CN110363777A CN 110363777 A CN110363777 A CN 110363777A CN 201910592087 A CN201910592087 A CN 201910592087A CN 110363777 A CN110363777 A CN 110363777A
- Authority
- CN
- China
- Prior art keywords
- sky
- sea
- image
- classification
- seashore
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 27
- 239000013535 sea water Substances 0.000 claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 12
- 238000005036 potential barrier Methods 0.000 claims abstract description 9
- 238000009826 distribution Methods 0.000 claims description 17
- 238000005457 optimization Methods 0.000 claims description 11
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 7
- 230000004888 barrier function Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000003595 mist Substances 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000005381 potential energy Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06T7/143—Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
-
- 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
- G06T2207/10024—Color image
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30184—Infrastructure
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The sea image semantic segmentation method based on reducible space constraint mixed model that the invention discloses a kind of, comprising the following steps: (1) input sea color image to be detected；(2) assume that there are sky, seashore/haze, three main semantic regions of seawater and potential barrier regions for sea image, and establish the mixed model of space constraint with this；(3) space constraint mixed model is optimized using expectation-maximization algorithm (EM)；(4) sky, seashore/haze classification Gaussian Profile KL distance (Kullback-Leibler divergence) are calculated, if KL distance is less than the threshold value of setting, abbreviation is carried out to space constraint mixed model；(5) sea image semantic segmentation result is exported.Method of the invention effectively can carry out semantic segmentation to sea image, and have the characteristics that speed is fast, robustness is good.
Description
Technical field
The present invention relates to technical field of image processing, and in particular to a kind of based on reducible space constraint mixed model
Sea image semantic segmentation method.
Background technique
Image, semantic segmentation is the basic technology that computer vision understands, task is to be partitioned into figure from the angle of pixel
Different objects as in simultaneously carry out semantic tagger (classify) to each pixel.Semantic segmentation technology is applied to sea level chart
Picture helps to enhance unmanned water surface ship to the sensing capability of ambient enviroment, to guarantee that it carries out safety work.
In recent years, with the fast development of deep learning, the semantic segmentation method based on convolutional neural networks is driven at nobody
Sail, medical imaging analysis etc. fields obtained it is extensive research and application.However, using convolutional neural networks come to sea image
The correlative study for carrying out semantic segmentation is also seldom, main reason is that lacking a large amount of sea image labeled data.2016,
Kristan et al. is in " Fast Image-Based Obstacle Detection From Unmanned Surface
Vehicles " in propose a kind of sea semantic segmentation method based on Clustering.The method define a space constraints
Mixed model models the pixel characteristic of sea image.Specifically, this process employs three Gaussian Profiles and one
It is uniformly distributed and comes the sky areas to sea image, intermediate seashore/haze Mixed Zone, seawater region and potential barrier respectively
Hinder object area (singular value region) to be described, and passes through the language of expectation-maximization algorithm (EM algorithm) sea Lai Shixian image
Justice segmentation.This method detection performance preferably, fast speed；However, can be with by the sea image of observation unmanned water surface ship shooting
It was found that: when unmanned boat deviates from seashore when driving, there is normally only sky areas, seawater region and potential obstacles for sea image
Object area, there is no Kristan et al. assume intermediate seashore/haze Mixed Zone, thus Kristan et al. proposition method
Semantic segmentation result and physical presence large error in this case.
Summary of the invention
In view of the deficiencies of the prior art, the present invention provides a kind of sea level charts based on reducible space constraint mixed model
As semantic segmentation method, this method effectively can carry out semantic segmentation to sea image, and have that speed is fast, robustness is good
The characteristics of.
In order to achieve the above objectives, the present invention adopts the following technical scheme:
A kind of sea image semantic segmentation method based on reducible space constraint mixed model, comprising the following steps:
(1) sea color image to be detected is inputted；
(2) assume that there are sky, seashore/haze, three main semantic regions of seawater and potential barriers for sea image
Region, and establish with this mixed model of space constraint；
(3) space constraint mixed model is optimized using expectation-maximization algorithm (EM)；
(4) sky, seashore/haze classification Gaussian Profile KL distance (Kullback-Leibler divergence) are calculated, if
KL distance is less than the threshold value of setting, then carries out abbreviation to space constraint mixed model；
(5) sea image semantic segmentation result is exported.
Further, in the step (2), it is assumed that mixed model is made of three Gaussian Profiles and being uniformly distributed one by one,
Wherein three Gaussian Profiles are respectively used to description description sky, haze/seashore Mixed Zone, seawater region, and are uniformly distributed use
In the potential barrier region (singular value region) of description.Then, in image ith pixel feature vector yiProbability can be with
It indicates are as follows:
In above formula, N (| m, C) indicates that mean value is m and gauss of distribution function that covariance is C, and U ()=ε indicates uniform
Distribution function (wherein, ε is a very small positive value hyper parameter)；yiIndicate the feature vector of ith pixel in image (also referred to as
Observe data), mainly it is made of the color characteristic of pixel (c1, c2, c3) and coordinate (r, c)；θ indicates all Gausses point in model
Parameter (the i.e. θ={ m of clothk,Ck}K=1,2,3)；π indicates that the category prior of all pixels in image is distributed (i.e. π={ πi}I=1:M,
In, M is the number of pixel in image), πiIndicate that the category prior of ith pixel is distributed (i.e. πi=[πi1,…,πik,…,πi4],
Wherein, πik=p (xi=k) indicate ith pixel classification xiProbability when for k, it is assumed that k=1 indicates sky classification, k=2 table
Show intermediate seashore/haze hybrid category, k=3 indicates seawater classification, and k=4 indicates barrier classification)；
In order to overcome local noise adverse effect caused by image segmentation in the image of sea, by introduce markov with
Airport (Markov Random Field, MRF) to carry out mixed model space constraint, i.e., all pixels in hypothesis image
Category prior is distributed π={ πi}I=1:MAnd Posterior distrbutionp P={ pi}I=1:MIt is a MRF about neighborhood system.According to
The joint probability distribution of Besag method, prior distribution π can be approximated to be:
In above formula, NiFor the neighborhood of pixel i,For neighborhood NiCategory prior distribution:
Wherein, λijFor fixed positive value weight, when pixel j and range pixel i are smaller, λijIt is bigger, and ∑jλij=1.
In addition, the potential-energy function in MRF is (i.e.It can be with is defined as:
In above formula,For KL divergence item, H (πi) it is entropy item.
And Posterior distrbutionp P={ pi}I=1:MJoint probability distribution are as follows:
Wherein, the Posterior distrbutionp p of pixel ii={ pik}K=1:4Calculation formula it is as follows:
Simultaneous formula (1), (2), (4) and (5), the available semantic segmentation model based on the uniform mixed distribution of gaussian sum
Joint probability density function:
In above formula, due toWithIn there are coupled relations, therefore, it is difficult to directly carry out model parameter to it
Estimation.In order to solve this problem, secondary category prior distribution collection s={ s can be introducedi}I=1:MWith auxiliary Posterior distrbutionp collection q=
{qi}I=1:MInto above formula, and peer-to-peer both sides are derived from right logarithm operation simultaneously, to obtain punishing for space constraint mixed model
Penalize log-likelihood function:
In above formula, ° expression Hadamard product operation；And work as si≡πiAnd qi≡piWhen, it can be formula by its abbreviation
(7).In addition, according to maximum a posteriori criterion above-mentioned formula can be maximized by EM algorithm, to realize to the excellent of mixed model
Change.
Further, in the step (3), the specific steps of expectation maximization are as follows:
1. initializing Parameters of Normal Distribution collection θ={ mk,Ck}K=1,2,3.By sea image from top to bottom in proportion { 0,0.3 },
{ 0.3,0.5 } and { 0.5,1 } marks off three regions and calculates separately out sky then according to the feature of these three area pixels
The mean value m of classification1With covariance C1, intermediate seashore/haze hybrid category mean value m2With covariance C2, seawater classification mean value m3
With covariance C3；
2. the category prior for initializing all pixels is distributed π={ πi}I=1:M.Category prior distribution for each pixel
πi, it is as follows to initialize formula:
In above formula, ε is a very small positive value hyper parameter.
In E-step:
3. θ, π are substituted into formula (6), the Posterior distrbutionp P={ p of all pixels is calculatedi}I=1:M。
4. calculating the secondary category prior distribution s={ s of all pixels according to following formulai}I=1:M；
In above formula, ° expression Hadamard product operation, * indicates convolution algorithm,For normaliztion constant.
5. calculating the auxiliary Posterior distrbutionp q={ q of all pixelsi}I=1:M, calculation formula is as follows:
In above formula,For normaliztion constant.
In M-step:
6. updating category prior distribution collection, calculation formula is as follows:
7. updating Gaussian Distribution Parameters, calculation formula is as follows:
8. judging whether EM algorithm reaches stopping criterion for iteration；If reached, stop iteration, otherwise, then continue 3.~
⑧.Wherein, stopping criterion for iteration is as follows:
Further, in the step (4), when due to unmanned water surface ship towards shore, ship carries the sea that camera takes
There is three sky, intermediate seashore/haze and seawater main semantic regions from top to bottom in face image；However, working as unmanned water surface ship
When away from shore, ship carries the sea image that camera takes and only exists two main semantic spaces of sky and seawater from top to bottom
Domain.Therefore, when sea, image only exists two main semantic regions of sky and seawater, the space for needing to assume step (2) is about
Beam mixed model carries out abbreviation.The specific steps are that:
1. on the basis of step (3), the Gauss point of sky classification and intermediate seashore/haze hybrid category after calculating EM
The KL distance (Kullback-Leibler divergence) of cloth:
Dst=KL (N1||N2) (16)
In above formula, N1Expression sky classification Gaussian Profile N (| m1,C1), N2Indicate intermediate seashore/haze hybrid category
Gaussian Profile N (| m2,C2)。
2. if carrying out abbreviation to mixed model when KL distance dst is less than default fixed threshold T；Otherwise, it directly executes
Step (5).Specifically, as dst < T, the Gaussian Profile of sky classification and intermediate seashore/haze hybrid category is merged,
To form new sky classification Gaussian Profile:
In above formula, m1And C1For the Gaussian parameter of the sky classification after EM optimization, m2And C2For seashore/mist after EM optimization
The Gaussian parameter of haze hybrid category.
Then, mixed model is reduced to be uniformly distributed by 2 Gaussian Profiles and 1 and form, wherein 2 Gaussian Profile difference
For describing description sky and seawater region, and it is uniformly distributed for describing potential barrier region (singular value region).Cause
This, the Posterior distrbutionp p of ith pixel in the image of seai={ pik}K=1:4It can be calculated with following formula:
In above formula, m '1With C '1For the Gaussian parameter of new sky classification, m3And C3For the height of the seawater classification after EM optimization
This parameter, πi3And πi4It is first to belong to seawater, the classification of barrier classification for ith pixel in sea image after respectively EM optimization
Test probability, π 'i1Belong to the category prior probability of new sky classification for ith pixel in the image of sea:
π′i1=πi1+πi2 (19)
In above formula, πi1And πi2Ith pixel belongs to sky, intermediate seashore/mist in sea image after respectively EM optimization
The category prior probability of haze hybrid category.
Further, in the step (5), as dst < T, the Posterior distrbutionp p=that is calculated using formula (18)
{pi}I=1:MObtain sea image semantic segmentation resultIt is as dst >=T, then excellent using step (3) EM
Posterior distrbutionp q={ q after changei}I=1:MObtain sea image semantic segmentation result
Compared with prior art, the invention has the benefit that
Firstly, method proposed by the present invention can be automatically selected whether according to the actual conditions of sea image to space constraint
Mixed model simplified, to improve the accuracy of sea image semantic segmentation.Secondly, sea designed by the present invention
The characteristics of image, semantic dividing method has model structure simple, fast speed is easy to Practical Project deployment.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention.
Fig. 2 is the schematic diagram of the method for the present invention one embodiment, wherein (a) is embodiment mapping to be checked；It (b) is embodiment
Semantic segmentation result figure；
Fig. 3 is sea image semantic segmentation schematic diagram of the embodiment under no seashore/haze background, wherein (a) is to implement
Example mapping to be checked；It (b) is embodiment semantic segmentation result figure.
Specific embodiment
It is clear to be more clear the object, technical solutions and advantages of the present invention, with reference to the accompanying drawing, to tool of the invention
Body embodiment elaborates.Method or step involved in following embodiment is then unless otherwise instructed this technology neck
The conventional method or step in domain, those skilled in the art can make conventional selection or adaptability tune according to concrete application scene
It is whole.Following embodiment is realized using python programming language.
Embodiment 1
As shown in Figure 1, a kind of sea image semantic segmentation method based on reducible space constraint mixed model, specifically
Realize that steps are as follows:
(1) sea color image to be detected is inputted；
Obtain sea color image by the camera that unmanned water surface ship carries (resolution ratio is 512 × 512).After reducing
The execution time of continuous algorithm, sea image is zoomed into resolution ratio 100 × 100.It is as shown in Figure 2 a the present embodiment sea to be detected
Face image, image used mainly include sky, seashore, waves of seawater and obstruction buoy etc..
(2) assume that there are sky, seashore/haze, three main semantic regions of seawater and potential barriers for sea image
Region, and establish with this mixed model of space constraint；
It is formed assuming that mixed model is uniformly distributed by 3 Gaussian Profiles and 1, wherein 3 Gaussian Profiles are respectively used to retouch
Description sky, haze/seashore Mixed Zone, seawater region are stated, and is uniformly distributed (unusual for describing potential barrier region
It is worth region).In addition, by introducing Markov random field (Markov Random Field, MRF) to carry out mixed model
Space constraint, i.e., the category prior of all pixels is distributed π={ π in hypothesis imagei}I=1:MAnd Posterior distrbutionp P={ pi}I=1:M
It is a MRF about neighborhood system.By derivation, the penalized log-likelihood function of space constraint mixed model is finally obtained:
In addition, the feature vector y of each pixel of sea imageiIt is by color characteristic (c1, c2, c3) and position feature
5 dimensional vector [c1, c2, c3, r, c] that (r, c) is constitutedT.Wherein, color characteristic component is determined by YCbCr color space.
(3) space constraint mixed model is optimized using expectation-maximization algorithm (EM)；
The specific steps of expectation maximization (EM) are as follows:
1. initializing Parameters of Normal Distribution collection θ={ mk,Ck}K=1,2,3.By sea image from top to bottom in proportion { 0,0.3 },
{ 0.3,0.5 } and { 0.5,1 } marks off three regions and calculates separately out sky then according to the feature of these three area pixels
The mean value m of classification1With covariance C1, intermediate seashore/haze hybrid category mean value m2With covariance C2, seawater classification mean value m3
With covariance C3；
2. the category prior for initializing all pixels is distributed π={ πi}I=1:M.Category prior distribution for each pixel
πi, it is as follows to initialize formula:
In above formula, ε is a very small positive value hyper parameter.In the present embodiment, ε=1 × 10-15。
In E-step:
3. θ, π are substituted into formula (6), the Posterior distrbutionp P={ p of all pixels is calculatedi}I=1:M, wherein it is uniformly distributed U
()=ε=1 × 10-15。
4. utilizing following formula, the secondary category prior distribution s={ s of all pixels is calculatedi}I=1:M；
In above formula,Indicating Hadamard product operation, * indicates convolution algorithm,For normaliztion constant.
5. calculating the auxiliary Posterior distrbutionp q={ q of all pixelsi}I=1:M, calculation formula is as follows:
In above formula, ξqiFor normaliztion constant.
In M-step:
6. updating category prior distribution collection, calculation formula is as follows:
7. updating Gaussian Distribution Parameters, calculation formula is as follows:
8. judging whether EM algorithm reaches stopping criterion for iteration；If reached, stop iteration, otherwise, then continue 3.~
⑧.Wherein, stopping criterion for iteration is as follows:
(4) sky, seashore/haze classification Gaussian Profile KL distance (Kullback-Leibler divergence) are calculated, if
KL distance is less than the threshold value of setting, then carries out abbreviation to space constraint mixed model；
When the KL distance dst of sky, seashore/haze classification Gaussian Profile is less than default fixed threshold T, then to mixing
Model carries out abbreviation；Otherwise, step (5) directly are executed.Wherein, the preset fixed threshold T=8 of the present embodiment.
(5) sea image semantic segmentation result is exported.Fig. 2 b is the present embodiment semantic segmentation result figure.
Embodiment 2
Fig. 3 is a preferred embodiment of the method for the present invention under no seashore/haze background.Its specific implementation step and reality
Apply that example 1 is identical, and so it will not be repeated.Can be seen that method of the invention from the semantic segmentation result of embodiment 1 and embodiment 2 can
It automatically selects whether to simplify the mixed model of space constraint according to the actual conditions of sea image, to improve sea
The accuracy of face image, semantic segmentation.
Claims (3)
1. a kind of sea image semantic segmentation method based on reducible space constraint mixed model, which is characterized in that including
Following steps:
(1) sea color image to be detected is inputted；
(2) assume sea image there are sky, seashore/haze, three main semantic regions of seawater and potential barrier region,
And the mixed model of space constraint is established with this；
(3) space constraint mixed model is optimized using expectation-maximization algorithm EM；
(4) sky, seashore/haze classification Gaussian Profile KL distance are calculated, if KL distance is less than the threshold value of setting, to sky
Between constraint mixed model carry out abbreviation；
(5) sea image semantic segmentation result is exported.
2. the sea image semantic segmentation method according to claim 1 based on reducible space constraint mixed model,
It is characterized in that, when due to unmanned water surface ship towards shore, ship carries the sea level chart that camera takes in the step (4)
As there is three sky, intermediate seashore/haze and seawater main semantic regions from top to bottom；However, when unmanned water surface ship deviates from
When shore, ship carries the sea image that camera takes and only exists two main semantic regions of sky and seawater from top to bottom；Cause
This, when sea, image only exists two main semantic regions of sky and seawater, needs to mix the space constraint that step (2) are assumed
Model carries out abbreviation, the specific steps are that:
1. on the basis of step (3), the Gaussian Profile of sky classification and intermediate seashore/haze hybrid category after calculating EM
KL distance:
Dst=KL (N1||N2) (1)
In above formula, N1Expression sky classification Gaussian Profile N (| m1,C1), N2Indicate the Gauss of intermediate seashore/haze hybrid category
Distribution N (| m2,C2)；
2. carrying out abbreviation to mixed model if KL distance dst is less than default fixed threshold T；Otherwise, step is directly executed
(5)；Specifically, as dst < T, the Gaussian Profile of sky classification and intermediate seashore/haze hybrid category is merged, thus
Form new sky classification Gaussian Profile:
In above formula, m1And C1For the Gaussian parameter of the sky classification after EM optimization, m2And C2For the seashore after EM optimization/haze mixing
The Gaussian parameter of classification；
Then, mixed model is reduced to be uniformly distributed by 2 Gaussian Profiles and 1 and form, wherein 2 Gaussian Profiles are respectively used to
Description description sky and seawater region, and be uniformly distributed for describing potential barrier region i.e. singular value region；Therefore, extra large
The Posterior distrbutionp p of ith pixel in the image of facei={ pik}K=1:4It is calculated with following formula:
In above formula, N (| m, C) indicates that mean value is m and gauss of distribution function that covariance is C, and U ()=ε expression is uniformly distributed
Function, wherein ε is a very small positive value hyper parameter；yiThe feature vector for indicating ith pixel in image, is also referred to as observed
Data are mainly made of the color characteristic of pixel (c1, c2, c3) and coordinate (r, c)；m′1With C '1For the height of new sky classification
This parameter, m3And C3For the Gaussian parameter of the seawater classification after EM optimization, πi3And πi4In sea image after respectively EM optimization
Ith pixel belongs to the category prior probability of seawater, barrier classification, π 'i1Belong to new day for ith pixel in the image of sea
The other category prior probability of empty class:
π′i1=πi1+πi2 (4)
In above formula, πi1And πi2It is mixed to belong to sky, intermediate seashore/haze for ith pixel in sea image after respectively EM optimization
Close the category prior probability of classification.
3. the sea image semantic segmentation method according to claim 1 based on reducible space constraint mixed model,
It is characterized in that, in the step (5), the Gaussian Profile of sky classification and intermediate seashore/haze hybrid category after EM
When KL distance dst < T, the Posterior distrbutionp p={ p that is calculated using formula (3)i}I=1:MObtain sea image semantic segmentation resultAs dst >=T, then the Posterior distrbutionp q={ q after step (3) EM optimization is utilizedi}I=1:MObtain sea
Face image, semantic segmentation result
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910592087.9A CN110363777A (en) | 2019-07-03 | 2019-07-03 | A kind of sea image semantic segmentation method based on reducible space constraint mixed model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910592087.9A CN110363777A (en) | 2019-07-03 | 2019-07-03 | A kind of sea image semantic segmentation method based on reducible space constraint mixed model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110363777A true CN110363777A (en) | 2019-10-22 |
Family
ID=68217730
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910592087.9A Pending CN110363777A (en) | 2019-07-03 | 2019-07-03 | A kind of sea image semantic segmentation method based on reducible space constraint mixed model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110363777A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613515A (en) * | 2020-11-23 | 2021-04-06 | 上海眼控科技股份有限公司 | Semantic segmentation method and device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102799646A (en) * | 2012-06-27 | 2012-11-28 | 浙江万里学院 | Multi-view video-oriented semantic object segmentation method |
| CN104036503A (en) * | 2014-06-05 | 2014-09-10 | 哈尔滨工程大学 | Image segmentation method based on spatial location information |
| CN109284663A (en) * | 2018-07-13 | 2019-01-29 | 上海大学 | A kind of sea obstacle detection method based on normal state and uniform Mixture Distribution Model |
-
2019
- 2019-07-03 CN CN201910592087.9A patent/CN110363777A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102799646A (en) * | 2012-06-27 | 2012-11-28 | 浙江万里学院 | Multi-view video-oriented semantic object segmentation method |
| CN104036503A (en) * | 2014-06-05 | 2014-09-10 | 哈尔滨工程大学 | Image segmentation method based on spatial location information |
| CN109284663A (en) * | 2018-07-13 | 2019-01-29 | 上海大学 | A kind of sea obstacle detection method based on normal state and uniform Mixture Distribution Model |
Non-Patent Citations (1)
| Title |
|---|
| KRISTAN，ET AL: "Fast Image-Based Obstacle Detection From Unmanned Surface Vehicles", 《IEEE TRANSACTIONS ON CYBERNETICS》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613515A (en) * | 2020-11-23 | 2021-04-06 | 上海眼控科技股份有限公司 | Semantic segmentation method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11488308B2 (en) | Three-dimensional object detection method and system based on weighted channel features of a point cloud | |
| Zhou et al. | Efficient road detection and tracking for unmanned aerial vehicle | |
| CN111681197B (en) | Remote sensing image unsupervised change detection method based on Siamese network structure | |
| CN106023257B (en) | A kind of method for tracking target based on rotor wing unmanned aerial vehicle platform | |
| CN103985133B (en) | Search method and system for optimal splicing lines among images based on graph-cut energy optimization | |
| CN110287837A (en) | Sea obstacle detection method based on prior estimate network and space constraint mixed model | |
| Delmerico et al. | Building facade detection, segmentation, and parameter estimation for mobile robot localization and guidance | |
| US9256619B2 (en) | Method for detecting and recognizing boats | |
| CN105022990A (en) | Water surface target rapid-detection method based on unmanned vessel application | |
| CN107808386A (en) | A kind of sea horizon detection method based on image, semantic segmentation | |
| CN107169972B (en) | Non-cooperative target rapid contour tracking method | |
| CN103077531A (en) | Grayscale target automatic tracking method based on marginal information | |
| CN104463877A (en) | Shoreline registration method based on information of radar image and electronic sea chart | |
| CN105787962B (en) | A kind of monocular vision tracking recycled under water based on UUV | |
| Zhang et al. | Research on unmanned surface vehicles environment perception based on the fusion of vision and lidar | |
| CN109919026A (en) | A kind of unmanned surface vehicle local paths planning method | |
| CN111881725A (en) | Optical remote sensing image ship target detection method fusing space-frequency domain features | |
| CN109284663A (en) | A kind of sea obstacle detection method based on normal state and uniform Mixture Distribution Model | |
| CN108830224A (en) | A kind of high-resolution remote sensing image Ship Target Detection method based on deep learning | |
| CN110298855B (en) | Sea-sky-line detection method based on Gaussian mixture model and texture analysis | |
| Alla et al. | Vision-based Deep Learning algorithm for Underwater Object Detection and Tracking | |
| Petković et al. | An overview on horizon detection methods in maritime video surveillance | |
| CN102509308A (en) | Motion segmentation method based on mixtures-of-dynamic-textures-based spatiotemporal saliency detection | |
| CN112801021B (en) | Method and system for detecting lane line based on multi-level semantic information | |
| Hashmani et al. | A survey on edge detection based recent marine horizon line detection methods and their applications |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191022 |