Human tracking and counting using the KINECT range sensor based on Adaboost and Kalman Filter ISVC 2013 Problem • Human tracking Avoid occlusion Human Detection • Observations: – There is an empty space in the front and back of head – The right side of right shoulder and the left side of left shoulder are also empty – There is a height difference between the head and the two shoulders How to describe the spatial information of 3D HASP Human Detection • Those criteria can be formulated as the difference of two pixel areas in the depth map – Haar-like feature • Adaboost is introduced to construct a strong classifiers from those weak criteria Human Detection by Adaboost • Framework Spatial feature • Processing window – 20 redefined sub-windows Spatial feature • Four Haar-like features Depth integral image • The sum of rectangle pixel values from the top-left corner to a pixel in depth image – To speed up the computation of Haar-like features • All pixel intensity values of D: areaValue( D) dd (4) dd (3) dd (2) dd (1) Adaboost algorithm • Construct a strong classifier by a weighted linear combination of weak classifiers 1, j * H ( j ) * j F 0, j * H ( j ) * j where 1, p * h( x) p H(h, x, , p) 1, otherwise Our Classifier • Challenge – Human can stand and face all directions with many postures • Solutions – Combine a horizontal strong classifier and a vertical strong classifier F C ( win) F hor ( win) | F ver ( win) Horizontal Strong Classifier • Formulation 1, j * H ( j ) * j F hor (win) 0, * H ( j ) * j j Vertical Strong Classifier • Formulation 1, j * H ( j ) * j F ver (win) 0, * H ( j ) * j j Training • Took many depth maps of each object by rotating a certain degree • 720 positive images + 288 negative images Results • Testing on three datasets: – Dataset 1: only one human object standing in different directions – Dataset 2: Two human objects – Dataset 3: three or more human objects Results (Dataset 1) Results (Dataset 2) Results (Dataset 3) Choice of window sizes Limitation • Fails if detected humans are standing two very close to each other – Improve tracking accuracy by incorporating Kalman Filter, since the closing time is short in real tracking application. Conclusion • We construct a real-time human detection based the depth image from Kinect sensor • Head and Shoulder Profile described by some Haar-like features is incorporated into Adaboost algorithm to detect human objects. • Detection time for each image is about 33 ms.