Device, computer program and method

1 . A device-system for detecting a goaltending event, comprising:

a plurality of cameras; and

circuitry configured to:

receive a video stream from the cameras;

determine a real-life position of a basketball from a video stream by determining the pixels in the video stream that correspond to the basketball;

detect an impact on the basketball from the movement of the basketball captured in the video stream;

determine a real-life skeletal feature position; and

output a signal indicating a detected goaltending event based on the detected impact and the real-life position of the basketball;

wherein the signal indicating the detected goaltending event is further based on the real-life position of the basketball relative to a real-life position of a basketball rim and the real-life skeletal feature position.

2 . The system according to claim 1 , wherein the signal includes a time segment indicative of the position of the detected goaltending event within the video stream.

3 . The system according to claim 1 , wherein the impact is detected based upon the deviation of movement of the basketball from a polynomial path.

4 . The system according to claim 1 , wherein the real-life position of the basketball is determined by the steps of:

performing object detection on a frame from the video stream to identify the position of one or more candidate basketballs within the frame using a machine learning technique.

5 . The system according to claim 1 , further comprising a content providing device wherein the content providing device is configured to generate a clip of video based upon the signal output from the device.

6 . The system of claim 1 , wherein the circuitry is configured to determine the real-life position of the basketball from the video stream by determining whether each pixel in the video stream is the center of the basketball.

7 . The system according to claim 1 , wherein the real-life position of the basketball rim is determined by the steps of:

converting one or more segments of a frame in the video stream into a colour space reference image, the reference colour being the colour of the basketball rim;

performing edge detection on the colour space reference image; and

determining the real-life position of the basketball rim based upon the edge detected colour space reference image.

8 . The system according to claim 7 , wherein the one or more segments include a segment bounding the basketball rim.

9 . The system of claim 1 , wherein the circuitry is configured to determine the real-life position of the basketball from the video stream by placing reference circles having different radii over each pixel and determining whether there is an edge of a circle at a number of points distributed along the circumference of the reference circles.

10 . The system of claim 9 , wherein the circuitry is configured to determine the real-life position of the basketball from the video stream by further applying a score to each pixel to indicate a likelihood of each pixel being the center of the basketball.

11 . A method for detecting a goaltending event, comprising:

receiving a video stream from cameras;

determining a real-life position of a basketball from a video stream by determining the pixels in the video stream that correspond to the basketball;

detecting an impact on the basketball from the movement of the basketball captured in the video stream;

determining a real-life skeletal feature position; and

outputting a signal indicating a detected goaltending event based on the detected impact and the real-life position of the basketball, wherein the signal indicating the detected goaltending event is further based on the real-life position of the basketball relative to a real-life position of a basketball rim and the real-life skeletal feature position.

12 . The method according to claim 11 , wherein the signal includes a time segment indicative of the position of the detected goaltending event within the video stream.

13 . The method according to claim 11 wherein the impact is detected based upon the deviation of movement of the basketball from a polynomial path.

14 . The method according to claim 11 , wherein the real-life position of the basketball is determined by the steps of:

performing object detection on a frame from the video stream to identify the position of one or more candidate basketballs within the frame using a machine learning technique.

15 . A method comprising generating a clip of video based upon the output signal and a method according to claim 11 .

16 . A non-transitory computer readable medium storing a program product comprising computer readable instructions which, when loaded onto a computer, configures the computer to perform the method according to claim 11 .

17 . The method of claim 11 , wherein determining the real-life position of the basketball from the video stream includes placing reference circles having different radii over each pixel and determining whether there is an edge of a circle at a number of points distributed along the circumference of the reference circles.

18 . The system of claim 11 , wherein determining the real-life position of the basketball from the video stream includes determining whether each pixel in the video stream is the center of the basketball.

19 . The method according to claim 11 , wherein the real-life position of the basketball rim is determined by the steps of:

converting one or more segments of a frame in the video stream into a colour space reference image, the reference colour being the colour of the basketball rim;

performing edge detection on the colour space reference image; and

determining the real-life position of the basketball rim based upon the edge detected colour space reference image.

20 . The method according to claim 19 , wherein the one or more segments include a segment bounding the basketball rim.