Interactive input system and method
View Patent ↗A method of determining locations of at least two pointers in a captured image frame comprises generating a vertical intensity profile (VIP) from the captured image frame, the VIP comprising peaks generally corresponding to the at least two pointers; determining if the peaks are closely spaced and, if the peaks are closely spaced, fitting a curve to the VIP; analyzing the fitted curve to determine peak locations of the fitted curve; and registering the peak locations as the pointer locations.
1. A method of determining locations of at least two pointers in a captured image frame, the method comprising:
generating a vertical intensity profile (VIP) from the captured image frame, the VIP comprising peaks generally corresponding to the at least two pointers;
determining if the peaks are closely spaced and, if the peaks are closely spaced:
fitting a curve to the VIP by calculating a nonlinear least squares fit for at least one model, each model comprising at least two peaks, each peak having at least one of a respective peak location, a respective peak half-width and a respective peak amplitude, calculating a goodness-of-fit value for each model and determining a final model by comparing the goodness-of-fit values, the final model being said fitted curve;
analyzing the fitted curve to determine peak locations of the fitted curve; and
registering the determined peak locations as the pointer locations.
2. The method of claim 1 , wherein if the peaks are determined not to be closely spaced, the method further comprising:
analyzing the VIP to determine the peak locations; and
registering the determined peak locations as the pointer locations.
3. The method of claim 1 , wherein the nonlinear least squares fit is calculated using one of a conjugate gradient algorithm, a Levenberg-Marquardt algorithm, and a nonlinear simplex algorithm.
4. The method of claim 1 , wherein the fitting a curve further comprises:
calculating an Akaike Information Criterion (AIC) value for each model; and
determining the final model by comparing the goodness-of-fit values and the AIC values.
5. The method of claim 4 , wherein the final model determined is the model having at least one of a lowest goodness-of-fit value and a lowest AIC value.
6. The method of claim 5 , further comprising:
calculating a value of fit residuals for each model,
wherein the AIC value for each model is calculated based on the value of the fit residuals.
7. The method of claim 1 , wherein the at least one model is based on a point-by-point maximum of the VIP, the point-by-point maximum given by:
max([ƒ(x o , w, σ), g(x o , w, σ), h(x o , w, σ), . . . ])
where ƒ(x, w, σ), g(x, w, σ) and h (x, w, σ) each represent an individual VIP function corresponding to a respective VIP peak, and where x o is the peak location and w is the half-width of each individual VIP function, and σis instrumental profile width.
8. The method of claim 7 , wherein each VIP peak is proportional to an error function erƒ(x) given by:
1
2
·
(
erf
(
x
+
w
2
σ
)
-
erf
(
x
-
w
2
σ
)
)
.
9. The method of claim 1 , further comprising, prior to said determining:
comparing the VIP to a threshold; and
identifying the peaks in the VIP based on the comparing.
10. An interactive input system, comprising:
at least one imaging assembly capturing image frames of a region of interest; and
processing structure in communication with the at least one imaging assembly, the processing structure being configured to:
generate a vertical intensity profile (VIP) from a captured image frame, the VIP comprising peaks generally corresponding to locations of at least two pointers in the captured image frame; and
determine if the peaks are closely spaced, and if the peaks are closely spaced to:
fit a curve to the VIP by calculating a nonlinear least squares fit for at least one model, each model comprising at least two peaks, each peak having at least one of a respective peak location, a respective peak half-width and a respective peak amplitude, calculating a goodness-of-fit value for each model, and determining a final model by comparing the goodness-of-fit values, the final model being the fitted curve;
analyze the fitted curve to determine peak locations of the fitted curve;
register the determined peak locations as the pointer locations.
11. The system of claim 10 , wherein the at least one imaging assembly comprises two or more imaging assemblies, the processing structure being further configured to carry out said generating and said determining for a respective image frame captured by each imaging assembly, and to:
calculate the positions of the at least two pointers by triangulation based on the pointer locations.
12. The system of claim 11 , wherein if the peaks are determined not to be closely spaced, the processing structure is further configured to:
analyze the VIP to determine the peak locations; and
register the determined peak locations as the pointer locations.
13. The system of claim 11 , wherein the nonlinear least squares fit is calculated using one of a conjugate gradient algorithm, a Levenberg-Marquardt algorithm, and a nonlinear simplex algorithm.
14. The system of claim 11 , wherein, during the fitting, the processing structure is further configured to:
calculate an Akaike Information Criterion (AIC) value for each model; and
determine the final model by comparing the goodness-of-fit values and the AIC values.
15. The system of claim 14 , wherein the final model determined is the model having at least one of a lowest goodness-of-fit value and a lowest AIC value.
16. The system of claim 15 , wherein the processing structure is further configured to:
calculate a value of fit residuals for each model,
wherein the AIC value for each model is calculated based on the value of the fit residuals.
17. The system of claim 11 , wherein the at least one model is based on a point-by-point maximum of the VIP, the point-by-point maximum given by:
max([ƒ(x o , w, σ), g(x o , w, σ), h(x o , w, σ) , . . . ])
where ƒ(x, w, σ), g (x, w, σ) and h (x, w, σ) each represent an individual VIP function corresponding to a respective VIP peak, and where x o is the peak location and w is the half-width of each individual VIP function, and σis instrumental profile width.
18. The system of claim 17 , wherein each VIP peak is proportional to an error function erƒ(x) given by:
1
2
·
(
erf
(
x
+
w
2
σ
)
-
erf
(
x
-
w
2
σ
)
)
.
19. The system of claim 11 , wherein the processing structure is further configured to, prior to the determining:
compare the VIP to a threshold; and
identify the peaks in the VIP based on the comparing.
20. A non-transitory computer readable medium having embodied thereon a computer program for determining locations of at least two pointers in a captured image frame, the computer program comprising:
program code for generating a vertical intensity profile (VIP) from the captured image frame, the VIP comprising peaks generally corresponding to the at least two pointers; and
program code for determining if the peaks are closely spaced and, if the peaks are closely spaced:
program code for fitting a curve to the VIP by calculating a nonlinear least squares fit for at least one model, each model comprising at least two peaks, each peak having at least one of a respective peak location, a respective peak half-width and a respective peak amplitude, calculating a goodness-of-fit value for each model and determining a final model by comparing the goodness-of-fit values, the final model being said fitted curve;
program code for analyzing the fitted curve to determine peak locations of the fitted curve; and
program code for registering the determined peak locations as the pointer locations.