Presentation on theme: "Geoffrey Samuel PhD Researcher"— Presentation transcript:
1 Comparison of complex background subtraction algorithms using a fixed camera Geoffrey SamuelPhD ResearcherIntelligent Systems and Robotics Research Group (ISR)Creative TechnologiesUniversity of Portsmouth
2 IntroBackground subtraction is a important and vital step for computers to understand and interpreter a real-world sceneIt allows a computer to ignore a background so to concentrate on a foreground object
3 HypothesisEach background subtraction algorithm will have its advantages and disadvantages, and that looking and comparing these with a real-world situation, it would be possible to pick one algorithm or a method of combining algorithms to produce a algorithm capable of balancing speed with quality.
4 The GoalTest and evaluate the quality and speed of existing background subtraction algorithms on a complex background with different everyday motions, and to compare the results with those of the extracted “Ground Truth”
5 Complex Background Static Background:- Background does not contain any secondary “unwanted” motion. Controlled environment.Complex Background:-Background contains secondary “unwanted” motion such as the winds effect on trees or blinds.Real-world data.
6 Synthetic Test Data Advantages: Disadvantages: Automatically got the “Ground Truth”.More control over each test clip.Disadvantages:Manual frame by frame “Ground Truth” extraction.Added artefacts from the Chroma keying and compositing.
7 The ExperimentTo Create a set of synthetic data with the “Ground Truth”To test different motions with each background subtraction algorithmTo Compare the results of each algorithm with that of the “Ground Truth”
8 The Motions 7 everyday motions were chosen: DrinkingJoggingPicking up walletScratching headSitting downStanding upWalkingEach motion started on the left of the screen and concluded on the right.
9 Creating the test scenarios Green ScreenGreen Screen with actor“Ground Truth”Final CompositeBack Ground
10 The AlgorithmsBack Plate Difference │framei – backplate│>Ts50
11 The AlgorithmsFrame Difference │framei – framei-1│>Ts50
13 The Algorithms Mixture of Gaussians frame(it = μ) = Σi=1 ωi,t .ț(μ,o) k
14 Measuring the QualityCompare the Per-Pixel value of each frame with the “Ground Truth”(0,576)(768,576)(0,576)(768,576)(0,0)(768,0)(0,0)(768,0)
15 Results - Quality Most correct pixels Most incorrect pixels Test MotionsBackplate DifferenceFrame DifferenceApproximate MedianMixture of Gaussian% of image# of pixelsDrinking90.78%82.12%89.52%83.78%Jogging88.24%88.88%92.14%88.20%Picking up Wallet91.26%88.22%83.40%90.19%Scratch head88.18%84.87%90.56%86.15%Sitting down88.51%80.07%82.28%81.68%Standing up89.40%83.82%80.99%Walking88.47%89.81%94.22%90.01%Most correct pixelsMost incorrect pixels
19 Results - Speed...now ignoring the Mixture of Gaussian speed results
20 ConclusionBackplate difference was the fastest and produce the highest results in 4 out of 7 tests.Frame difference was the ONLY algorithm to correctly remove the complex background, but couldn't correctly identify the foreground element.
22 Taking it furtherA new method that incorporated both the speed of updating to remove the background and yet the knowledge of the background to properly extract the wanted foreground element.Theory Framework idea:ƒFrame DifferenceBackplate DifferenceComplex background removed
23 Where can this lead? Application of this technology could be used in: GamesSurveillanceMesh reconstruction and silhouette extractionVarious computer vision tasks