Post on 10-Feb-2016
description
transcript
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
A Model-Based Approach to the Detection A Model-Based Approach to the Detection and Classification of Mines in Side-scan and Classification of Mines in Side-scan
SonarSonar
S.Reed, Y.Petillot, J.BellS.Reed, Y.Petillot, J.Bell
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
ContentsContents
• Why Use Unsupervised Techniques?• Our Proposed CAD/CAC algorithm.• The Sonar Process.• Automated Object Detection.• Extraction of Object Features.• Automated Object Classification.• Future Research.• Conclusions.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Unsupervised TechniquesUnsupervised Techniques
• Rapid Advances in AUV Technology.
• On-board analysis now required.
• Large amounts of data quickly available for analysis.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Unsupervised Techniques Unsupervised Techniques
• Future automated systems will require all available information (navigation data, image processing models .etc.) to be fused.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
CAD/CAC ProposalCAD/CAC Proposal
Detect MLO’s(MRF-based Model)
Fuse Other Views
ExtractHighlight/Shadow
(CSS Model)
Classify Object(Dempster-Shafer)
FalseAlarm?
PositiveClassification?
1 2 YES
YES
NO
NOMINE
REMOVE FALSE ALARM
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
The Sonar ProcessThe Sonar Process
• Sonar images represent the time of flight of the sound rather than distance.
• Objects appear as a highlight/shadow pair in the sonar image.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
The Detection ModelThe Detection Model
• A Markov Random Field(MRF) model framework is used.
• MRF models operate well on noisy images.• A priori information can be easily incorporated.
• They are used toretrieve the underlying label field (e.g shadow/non-shadow)
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Basic MRF TheoryBasic MRF Theory
A pixel’s class is determined by 2 terms:
– The probability of being drawn from each classes distribution.
– The classes of its neighbouring pixels.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Incorporating A Priori InfoIncorporating A Priori Info
• Object-highlight regions appear as small, dense clusters.
• Most highlight regions have an accompanying shadow region.
),()(ln),()],(1[),(),,(,
2
Ss ts Ss Ss
sssXstsstsss oxsexxxyxoyxU Segment by minimising:
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Initial Detection ResultsInitial Detection Results
• Initial Results Good.• Model sometimes detects false alarms due to clutter
such as the surface return – requires more analysis!
DETECTED OBJECT
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Object Feature ExtractionObject Feature Extraction
• The object’s shadow is often extracted for classification.
• The shadow region is generally more reliable than the object’s highlight region for classification.
• Most shadow extraction models operate well on flat seafloors but give poor results on complex seafloors.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
The CSS ModelThe CSS Model
• 2 Statistical Snakes segment the mugshot image into 3 regions : object-highlight, object-shadow and background.
A priori information is modelled:
• The highlight is brighter than the shadow
• An object’s shadow region can only be as wide as its highlight region.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
CSS ResultsCSS ResultsCSS ModelStandard Model
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
The Combined ModelThe Combined Model
• Objects detected by MRF model are put through the CSS model.
• The CSS snakes are initialised using the label field from the detection result. This ensures a confident initialisation each time.
• The CSS can detect MANY of the false alarms. False alarms without 3 distinct regions ensure the snakes rapidly expand, identifying the detection as a false alarm.
• Navigation info is also used to produce height information which can also remove false alarms.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
ResultsResults
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Results 2
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Results 3
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Result 4
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
BP ’02 ResultsBP ’02 Results
• The combined detection/CSS model was run on 200 BP’02 data files containing 70 objects.
• 80% of the objects where detected and features extracted(for classification).
• 0.275 false alarms per image.
• The surface return resulted in some of the objects not being detected. Dealing with this would produce a detection rate of ~ 91%.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Object ClassificationObject Classification• The extracted object’s shadow can be used for
classification.
• We extend the classic mine/not-mine classification to provide shape and dimension information.
• The non-linear nature of the shadow-forming process ensures finding relevant invariant features is difficult.
Shadows from the same objectShadows from the same object
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Modelling the Sonar ProcessModelling the Sonar Process
• Mines can be approximated as simple shapes – cylinders, spheres and truncated cones.
• Using Nav data to slant-range correct, we can generate synthetic shadows under the same sonar conditions as the object was detected.
• Simple line-of-sight sonar simulator. Very fast.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Comparing the ShadowsComparing the Shadows
• Iterative Technique is required to find best fit. Parameter space limited by considering highlight and shadow length.
• Synthetic and real shadow compared using the Hausdorff Distance.
• It measures the mismatch of the 2 shapes.
HAUSDORFFDISTANCE
),(),,(max),( ABhBAhBAH
||||min max),( babahBbAa
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Incorporating KnowledgeIncorporating Knowledge
• As the technique is model-based, information on likely mine dimensions can be incorporated.
• Limited information from the highlight region can also be used to distinguish between the tested classes.
• We obtain an overall membership function for each class.
)()()(),,( elongjj
parjj
hausjjj
finalj HH
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
The Classification DecisionThe Classification Decision
• A decision could be made by simply defining a ‘Positive Classification Threshold’. This is a ‘hard’ decision and non-changeable.
• The ‘lawnmower’ nature of Sidescan surveys ensures the same object is often viewed multiple times. The model should ideally be capable of multi-view classification.
• We use DEMPSTER-SHAFER theory.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Mono-view ResultsMono-view Results
• Dempster-Shafer allocates a BELIEF to each class.
• Unlike Bayesian or Fuzzy methods, D-S theory can also consider union of classes.
Bel(cyl)=0.83Bel(sph)=0.0Bel(cone)=0.0Bel(clutter)=0.08
Bel(cyl)=0.0Bel(sph)=0.303Bel(cone)=0.45Bel(clutter)=0.045
Bel(cyl)=0.42Bel(sph)=0.0Bel(cone)=0.0Bel(clutter)=0.46
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Mono-view ResultsMono-view ResultsModel was tested on 66 mugshots containing cylinders,Spheres, Truncated cones and clutter objects.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Multi-view AnalysisMulti-view AnalysisDempster-Shafer allows results from multiple views to be fused.
Mono-Image Belief Fused BeliefObj Cyl Sph Cone Clutt Objs
FusedCyl Sph Cone Clutt
1 0.70 0.00 0.00 0.21 1 0.70 0.00 0.00 0.21
2 0.83 0.00 0.00 0.08 1,2 0.93 0.00 0.00 0.05
3 0.83 0.00 0.00 0.08 1,2,3 0.98 0.00 0.00 0.01
4 0.17 0.00 0.00 0.67 1,2,3,4 0.96 0.00 0.00 0.03
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Multi-Image AnalysisMulti-Image Analysis
Mono-Image Belief Fused BeliefObj Cyl Sph Cone Clutt Objs
FusedCyl Sph Cone Clutt
5 0.00 0.17 0.23 0.45 5 0.00 0.17 0.23 0.45
6 0.00 0.00 0.37 0.44 5,6 0.00 0.00 0.30 0.60
7 0.00 0.303 0.45 0.045 5,6,7 0.00 0.02 0.67 0.17
8 0.00 0.32 0.23 0.31 5,6,7,8 0.00 0.01 0.62 0.20
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Future ResearchFuture ResearchThe current detection model considers objects as a Highlight/Shadow pair. An object can also be considered as a discrepancy in the surrounding texture field.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
ConclusionsConclusions
• Automated Detection/Feature Extraction model has been developed and tested on a large amount of data. Good Results obtained, improvements expected when surface returns removed.
• Classification model uses a simple sonar simulator and Dempster-Shafer theory to classify the objects. Extends mine/not-mine classification to provide shape and size information.
• Future research is focusing on texture segmentation to complement the current work.
Automated Detection and Classification Automated Detection and Classification ModelsModels
S.Reed@hw.ac.uk
Acknowledgements
We would like to thank the following institutions for
their support and for providing data:
DRDC–Atlantic, Canada
Saclant Centre, Italy
GESMA, France