Animal Biometrics: Quantifying and Detecting Phenotypic Appearance
Total Page:16
File Type:pdf, Size:1020Kb
Online Supplementary Material
Animal biometrics: quantifying and detecting phenotypic appearance
Hjalmar S. Kühl1 and Tilo Burghardt2
1Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, 04103 Leipzig, Germany
2Department of Computer Science, University of Bristol, Woodland Road, Bristol, BS8 1UB, UK
Corresponding author: Kühl, H.S. ([email protected]).
Table S1: Visual Approaches towards Automated Animal Biometrics Categorization and summary of published techniques selected from the reference provided. PART I: SPECIES IDENTIFICATION
A Taxa Identificatio Pattern Recognition Techniques Annotation A S c n Requirement s y q Concept Modeling Model Matching s s F T s u (Profiled and Representati and o e a t i Pattern Learning on Locating c a r e s Information) Strategies (the Strategies i t g m i ‘Profile’) a u e t t r t M i e e L o o d e d n P T v a u y e l T b p l i y li e t p c y e a t S V M A Chimpanzee, Face - Boosting - Boxlet - Cascaded Rules training [38] P I O P Gorilla (Rigid Spatial - LookUp Features - Sliding Windows annotation of E S N P Decomposition Tables - Census facial region C U O E of Facial Features bounding box I A C A Features) - HOG Features E L U R Cat, Tiger, Animal Head - Boosting - Boxlet - Dual Approach training [48] S L A Lion, Panda, (Deformable - SVM Features - Sliding Windows annotation of [49] A N Fox, Cheetah Decomposition - Census head region and R C etc. of Head Shape Features facial features E and Texture) - HOOG Features Cat, Dog Face & Body - Latent SVM - Deformable - Sliding Windows training [75] (Deformable - Color Part-based - Grab Cut annotation of Decomposition Histograms Model - Segment facial region of Face Features - HOG Features Compactness bounding box & Homogeneity - Natural Edges Body Texture) - Pixel Color Various Distinctive - Gradient - SIFT - Scale-Space none [33] Species of Local Body Histograms - Local feature Extrema [34] Insect Features (Local concatenated - k-D Tree Gradient histograms Matching Distributions) African Body - Boosting - Boxlet - Cascaded Rules training [43] penguin (Rigid Spatial - Feature Features - Sliding Windows annotation of Decompositions Prediction key points on of Facial Trees body Features) M Quadrupeds Gait - PCA - Principal - Sparse Kanade- training based [28] O (Spatio- - Normalized Component Lucas-Tomasi on clip T temporal Convolution Vectors of tracking classification I configuration of motion fields O body part N motion) Ivory-Billed Flight - binary - principal axes - Spatio-temporal none – system [15] H Woodpecker characteristics segmentation of silhouette filters priming of Y (speed etc) + (against sky) - velocity and its profile B Silhouette derivatives parameters R information based on I anatomical D knowledge PART II: INDIVIDUAL IDENTIFICATION
A Taxa Identification Pattern Recognition Techniques Annotation A S c Concept s y q (Profiled Pattern Modeling and Model Matching Requireme s F T s u Information) Learning Representation and nts o e a t i Strategies (the ‘Profile’) Locating c a r e s Strategies i t g m i a u e t t r t M i e e L o o d e d n P T v a u y e l T b p l i y li e t p c y e a t I V M A Whale Shark Body Spots - Normalisation - Discrete - Groth’s Algorithm spot [16] N I O P (Configuration of based on Landmarks - I3S annotation or D S N P Discrete Landmarks Similarity verification I U O E on Sides of Body) Assumption V A C A Cheetah Body Spots/Stripes - Spline-based - Discrete - Convolution manual fitting [21] I L U R Tiger (Configuration of Texture Landmarks - Proprietary of spline [22] D L A Discrete Landmarks Backprojection Distance Measure model by U A N on Sides of Body) using body A R C Grey Seal Body Texture - Binarised Surface - covariance-based landmarks as [40] L E (Dense texture of Texture Map texture matching control points main body segment) Leatherback Head/Body Surface - Gradient - SIFT - Scale-Space None [35] Turtle, Masai (Distinctive gradient Histograms - Local feature Extrema [47] giraffe orientation of head or concatenated - k-D Tree Matching body texture) histograms Chimpanzee, Face - Random Faces - Hybrid Global - Random Faces near-frontal [39] Gorilla (Conglomerate of and Local Features face, Pre- local and global face detection characteristics) using work in [48] Salamander Body Texture - Coarse-to-fine - None, the method - Scale-cascaded ground truth [41] (Blob-like shape of decomposition produces match diffeomorphic for learning of markings over main using Gabor decisions for input alignment spectral body segment) filters pairs weights during training n Manta Ray, Body Spots/Blobs - Procrustes’ - Discrete - L -Type distance spot [13] (Configuration of Alignment of I3S Landmarks measures or annotation [76] Raggedtooth discrete landmarks system entropy [77] Shark, Weedy on main body) Seadragons African penguin Chest Spots - Polar Spatial - Shape Contexts - Earth Movers none [43] (Configuration of Histograms Distance [46] discrete landmarks - Phase Curls on chest feathers) Elephants Ears - Edge Tracing - Shape features of - Local edge rough ear- [45] (Shape of ear ear silhouette refinement line silhouette sections) annotation Zebra Body Stripes - Median Filter - Stripe Strings - Dynamic bounding box [25] (Binary pattern - Binarisation Programming around body formed by stripe - Run-Length - Edit Distance side paths) Coding
PART III: BEHAVIOR IDENTIFICATION
Taxa Identification Pattern Recognition Techniques A A Concept n s (Profiled Pattern Modeling and Model Matchi n s A Information) Learning Representation ng o o S c Strategies (the ‘Profile’) and t c y q Locati a i F T s u ng t a e a t i Strate i t a r e s gies o e t g m i n d u e t P r t M i R u e L o o e b e d n q li T v a u c y e l T i a p l i y r t e t p e i y e m o e n n s t s B V M A Mice Home-cage Activity - Spatiotemporal - Motion, position - SVM- none – after [50] E I O P (Spatiotemporal volume analysis and velocity based training H S N P characteristics of features classifier model A U O E particular activities) Birds Nesting Behaviour - Classification - SIFT Scales - Fisher [78] (Bird based on local Information Linear A presence/absence at spatial frequency Discrimi C R nest, egg count) and scale nant U A - Hidden L N Markov A C Models V R E Flocking Behaviour - Tracking by - Particle Cloud and - [53] I (Bird pose and propagation of Subordination Subordin O A & trajectory in flock) positional beliefs ated U L Condens R M ation
O Giant honey Shimmering - Template - 3D position - [29] S T bees matching estimates of Triangul T Behaviour I individual bees ation of E (3D movements of O correspo R hive) N nding E templat O es
ADDITIONAL REFERENCES:
75. Parkhi, O. M. et al. (2011) The truth about cats and dogs. Proceedings of ICCV, 1427-1434. 76. Marshall, A.D. et al. (2008) Morphological measurements of manta rays (Manta birostris) with a description of a foetus from the east coast of Southern Africa. Zootaxa 1717, 24–30.
77. Martin-Smith, K. M. (2011) Photo-identification of individual weedy seadragons Phyllopteryx taeniolatus and its application in estimating population dynamics. Journal of Fish Biology 78, 1757-1768.
78. Ko, T. et al. (2010) Heartbeat of a nest: Using imagers as biological sensors. TOSN 6, 1-30.