Scaled human video tracking
Methods, systems, an apparatus, including computer programs encoded on a storage device, for tracking human movement in video images. A method includes obtaining a first image of a scene captured by a camera; identifying a bounding box around a human detected in the first image; determining a scale amount that corresponds to a size of the bounding box; obtaining a second image of the scene captured by the camera after the first image was captured; and detecting the human in the second image based on both the first image scaled by the scale amount and the second image scaled by the scale amount. Detecting the human in the second image can include identifying a second scaled bounding box around the human detected in the second image scaled by the scale amount.
1. A method, comprising:
generating a bounding box around an object detected in a first image captured by a camera;
determining a scale amount using a size of the bounding box;
scaling, by the scale amount, a second image captured by the camera; and
detecting the object in the second image scaled by the scale amount.
2. The method of claim 1 , wherein detecting the object in the second image scaled by the scale amount comprises:
identifying, within the bounding box, a feature point corresponding to the object detected in the first image scaled by the scale amount; and
identifying a corresponding feature point in the second image scaled by the scale amount.
3. The method of claim 2 , comprising identifying the corresponding feature point in the second image scaled by the scale amount based on a similarity of pixel values between the feature point in the first image scaled by the scale amount and the feature point in the second image scaled by the scale amount.
4. The method of claim 1 , wherein determining the scale amount using the size of the bounding box comprises:
accessing stored data that associates bounding box sizes to respective scale amounts; and
using the stored data, selecting the scale amount associated with the size of the bounding box.
5. The method of claim 1 , wherein detecting the object in the second image comprises:
searching an area of the second image scaled by the scale amount that corresponds to the bounding box of the first image scaled by the scale amount; and
identifying a second bounding box around the object in the second image scaled by the scale amount.
6. The method of claim 1 , wherein the scale amount is based on scaling the bounding box to a scaled size that includes a particular density of feature points within the scaled bounding box.
7. The method of claim 6 , wherein the particular density of feature points comprises a particular range of a number of feature points per pixel within the scaled bounding box.
8. The method of claim 1 , wherein the scale amount is based on a scaling the bounding box to a scaled size that includes a particular quantity of feature points within the scaled bounding box.
9. The method of claim 8 , wherein the particular quantity of feature points comprises a particular range of a number of feature points within the scaled bounding box.
10. The method of claim 1 , wherein the object comprises a human.
11. A method, comprising
determining a range of shape sizes for tracking object movement in images;
determining a first size of a shape around an object detected in a first image, the first size being outside of the range of shape sizes;
scaling the first image by a scale amount to reduce the shape from the first size to a second size that is within the range of shape sizes;
scaling a second image by the scale amount; and
tracking object movement between the first image scaled by the scale amount and the second image scaled by the scale amount.
12. The method of claim 11 , wherein the shape encompasses feature points, and the range of shape sizes includes shape sizes that encompass a quantity of feature points within a particular range of quantities.
13. The method of claim 11 , wherein a shape encompasses feature points, and the range of shape sizes includes shape sizes that encompass a density of feature points within a particular range of densities.
14. The method of claim 13 , wherein a density of feature points comprises a number of feature points per pixel.
15. The method of claim 11 , wherein a shape size comprises at least one of a number of pixels along a first dimension or a number of pixels along a second dimension of the shape.
16. The method of claim 11 , wherein the object has an object shape, and the shape approximates the object shape.
17. The method of claim 16 , wherein the object shape comprises a shape of a human torso or a shape of a standing human body.
18. The method of claim 11 , wherein tracking object movement between the first image scaled by the scale amount and the second image scaled by the scale amount comprises:
identifying, within the shape, a feature point corresponding to the object detected in the first image scaled by the scale amount; and
identifying a corresponding feature point in the second image scaled by the scale amount.
19. The method of claim 18 , comprising identifying the corresponding feature point in the second image scaled by the scale amount based on a similarity of pixel values between the feature point in the first image scaled by the scale amount and the feature point in the second image scaled by the scale amount.
20. The method of claim 11 , wherein the object comprises a human.