Method and system for correcting temperature error of depth camera

1. A method comprising:

obtaining, by at least two depth cameras, a depth image of a current target, wherein two adjacent depth cameras of the at least two depth cameras have a common field of view;

modeling a measurement error of the two adjacent depth cameras caused by a temperature change, wherein the modeling comprises:

determining a gap change Δl caused by the temperature change in a gap l between a lens and an image sensor of an acquisition module in each of the two adjacent depth cameras; and

determining, according to the gap change Δl, an error Δd between a measured deviation value d′ and a true deviation value d, wherein Δd=Δl·tan θ and θ is an angle between an optical axis of the lens and a connecting line between the current target and an optical center of the lens; and

correcting the depth image using the modeled measurement error, wherein the corrected depth image has a minimum depth difference in the common field of view.

2. The method of claim 1 , wherein the modeling of the measurement error further comprises:

determining, according to the error Δd, a depth difference ΔZ between depth values in the common field of view of the two adjacent depth cameras, wherein

Δ

⁢

⁢

Z

=

-

Z

1

2

⁢

Δ

⁢

⁢

l

1

⁢

⁢

tan

⁢

⁢

θ

1

lb

+

Z

2

2

⁢

Δ

⁢

⁢

l

2

⁢

⁢

tan

⁢

⁢

θ

2

lb

,

wherein

Z 1 and Z 2 are respective depth values of the two adjacent depth cameras, Δl 1 and Δl 2 are respective gap changes between lenses and image sensors of acquisition modules in the two adjacent depth cameras, θ 1 and θ 2 , are respective angles between optical axes of the lenses and connecting lines between the current target and optical centers of the lenses of the two adjacent depth cameras, and b is a length of a baseline connecting a projection module and the acquisition module of one of the two adjacent depth cameras.

3. The method of claim 2 , wherein the correcting of the depth image using the modeled measurement error comprises:

calculating depth differences ΔZ (i) =Z 1 (i) −Z 2 (i) between depth values of each pixel in different depth images of the common field of view, wherein i∈[1, m], m is a quantity of a portion of pixels in the common field of view, the depth difference ΔZ (i) is determined by errors of the two adjacent depth cameras, and the errors of the two adjacent depth cameras are the same such that ΔZ 1 (i) =−ΔZ 2 (i) =ΔZ (i) /2;

calculating gap changes Δl 1 (i) and Δl 2 (i) according to the depth difference ΔZ 1 (i) , ΔZ 2 (i) , and a formula

Δ

⁢

⁢

Z

1

(

i

)

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

,

and averaging the gap changes Δl 1 (i) and Δl 2 (i) corresponding to the m pixels to obtain the gap changes Δl 1 and Δl 2 ; and

correcting depth values of pixels of the two adjacent depth cameras using the gap changes Δl 1 and Δl 2 in combination with the depth difference

Δ

⁢

⁢

Z

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

.

4. The method of claim 2 , wherein the correcting of the depth image using the modeled measurement error comprises: correcting the depth image using a least square method.

5. The method of claim 4 , wherein the correcting of the depth image using the least square method comprises:

calculating the gap changes Δl 1 and Δl 2 of the two adjacent depth cameras by minimizing a cost function J=kΣ i=1 m (Z 1 (i) =Z 2 (i) −ΔZ (i) ) 2 , and correcting the depth image of the at least two depth cameras according to the gap changes Δl 1 and Δl 2 , wherein k is a coefficient.

6. The method of claim 5 , wherein the calculating of the gap changes Δl 1 and Δl 2 comprises: calculating the gap changes Δl 1 and Δl 2 by at least one of gradient descent, Newton's method, and a normal equation.

7. A system, comprising:

at least two depth cameras, configured to obtain a depth image of a current target, wherein two adjacent depth cameras of the at least two depth cameras have a common field of view; and

a processor, configured to execute computer instructions stored in a memory to perform operations including:

modeling a measurement error of the two adjacent depth cameras caused by a temperature change, wherein the modeling comprises:

determining a gap change Δl caused by the temperature change in a gap l between a lens and an image sensor of an acquisition module in each of the two adjacent depth cameras; and

determining, according to the gap change Δl, an error Δd between a measured deviation value d′ and a true deviation value d, wherein Δd=Δl·tan θ, and θ is an angle between an optical axis of the lens and a connecting line between the current target and an optical center of the lens; and

correcting the depth image using the modeled measurement error, wherein the corrected depth image has a minimum depth difference in the common field of view.

8. The system of claim 7 , wherein the modeling of the measurement error further comprises:

determining, according to the error Δd, a depth difference ΔZ between depth values in the common field of view of the two adjacent depth cameras, wherein

Δ

⁢

⁢

Z

=

-

Z

1

2

⁢

Δ

⁢

⁢

l

1

⁢

⁢

tan

⁢

⁢

θ

1

lb

+

Z

2

2

⁢

Δ

⁢

⁢

l

2

⁢

⁢

tan

⁢

⁢

θ

2

lb

,

wherein

Z 1 and Z 2 are respective depth values of the two adjacent depth cameras, Δl 1 and Δl 2 are respective gap changes between lenses and image sensors of acquisition modules in the two adjacent depth cameras, θ 1 and θ 2 are respective angles between optical axes of the lenses and connecting lines between the current target and optical centers of the lenses of the two adjacent depth cameras, and b is a length of a baseline connecting a projection module and an acquisition module of one of the two adjacent depth cameras.

9. The system of claim 8 , wherein the correcting of the depth image using the modeled measurement error comprises:

calculating depth differences ΔZ (i) =Z 1 (i) −Z 2 (i) between depth values of each pixel in different depth images of the common field of view, wherein i∈[1, m], m is a quantity of a portion of pixels in the common field of view, the depth difference ΔZ (i) is determined by errors of the two adjacent depth cameras, and the errors of the two adjacent depth cameras are the same such that ΔZ 1 (i) =−ΔZ 2 (i) =ΔZ (i) /2;

calculating gap changes and Δl 1 (i) , and Δl 2 (i) according to the depth difference ΔZ 1 (i) , ΔZ 2 (i) , and a formula

Δ

⁢

⁢

Z

1

(

i

)

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

,

and averaging the gap changes Δl 1 (i) and Δl 2 (i) corresponding to the m pixels to obtain the gap changes Δl 1 and Δl 2 ; and

correcting depth values of pixels of the two adjacent depth cameras using the gap changes Δl 1 and Δl 2 in combination with the depth difference

Δ

⁢

⁢

Z

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

.

10. The system of claim 8 , wherein the correcting of the depth image using the modeled measurement error comprises: correcting the depth image using a least square method.

11. The system of claim 10 , wherein the correcting of the depth image using the least square method comprises:

calculating the gap changes Δl 1 and Δl 2 of the two adjacent depth cameras by minimizing a cost function J=kΣ i=1 m (Z 1 (i) −Z 2 (i) −ΔZ (i) ) 2 , and correcting the depth image of the at least two depth cameras according to the gap changes Δl 1 and Δl 2 , wherein k is a coefficient.

12. The system of claim 11 , wherein the calculating of the gap changes Δl 1 and Δl 2 comprises: calculating the gap changes Δl 1 and Δl 2 by at least one of gradient descent, Newton's method, and a normal equation.

13. A non-transitory computer-readable storage medium storing computer instructions executable by at least one processor to cause the at least one processor to perform operations comprising:

obtaining, by at least two depth cameras, a depth image of a current target, wherein two adjacent depth cameras of the at least two depth cameras have a common field of view;

modeling a measurement error of the two adjacent depth cameras caused by a temperature change, wherein the modeling comprises:

determining a gap change Δl caused by the temperature change in a gap l between a lens and an image sensor of an acquisition module in each of the two adjacent depth cameras; and

determining, according to the gap change Δl, an error Δd between a measured deviation value d′ and a true deviation value d, wherein Δd=Δl·tan θ, and θ is an angle between an optical axis of the lens and a connecting line between the current target and an optical center of the lens; and

correcting the depth image using the modeled measurement error, wherein the corrected depth image has a minimum depth difference in the common field of view.

14. The non-transitory computer-readable storage medium of claim 13 , wherein the modeling of the measurement error further comprises:

determining, according to the error Δd, a depth difference ΔZ between depth values in the common field of view of the two adjacent depth cameras, wherein

Δ

⁢

⁢

Z

=

-

Z

1

2

⁢

Δ

⁢

⁢

l

1

⁢

⁢

tan

⁢

⁢

θ

1

lb

+

Z

2

2

⁢

Δ

⁢

⁢

l

2

⁢

⁢

tan

⁢

⁢

θ

2

lb

,

wherein

Z 1 and Z 2 are respective depth values of the two adjacent depth cameras, Δl 1 and Δl 2 are respective gap changes between lenses and image sensors of acquisition modules in the two adjacent depth cameras, θ 1 and θ 2 are respective angles between optical axes of the lenses and connecting lines between the current target and optical centers of the lenses of the two adjacent depth cameras, and b is a length of a baseline connecting a projection module and an acquisition module of one of the two adjacent depth cameras.

15. The non-transitory computer-readable storage medium of claim 14 , wherein the correcting of the depth image using the modeled measurement error comprises:

calculating depth differences ΔZ (i) =Z 1 (i) −Z 2 (i) between depth values of each pixel in different depth images of the common field of view, wherein i∈[1, m], m is a quantity of a portion of pixels in the common field of view, the depth difference ΔZ (i) is determined by errors of the two adjacent depth cameras, and the errors of the two adjacent depth cameras are the same such that ΔZ 1 (i) =−ΔZ 2 (i) =ΔZ (i) /2;

calculating gap changes Δl 1 (i) and Δl 2 (i) according to the depth difference ΔZ 1 (i) , ΔZ 2 (i) , and a formula

Δ

⁢

⁢

Z

1

(

i

)

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

,

and averaging the gap changes Δl 1 (i) and Δl 2 (i) corresponding to the m pixels to obtain the gap changes Δl 1 and Δl 2 ; and

correcting depth values of pixels of the two adjacent depth cameras using the gap changes Δl 1 and Δl 2 in combination with the depth difference

Δ

⁢

⁢

Z

=

-

Z

2

⁢

Δ

⁢

⁢

l

⁢

⁢

tan

⁢

⁢

θ

lb

.

16. The non-transitory computer-readable storage medium of claim 14 , wherein the correcting of the depth image using the modeled measurement error comprises:

correcting the depth image using a least square method.

17. The non-transitory computer-readable storage medium of claim 16 , wherein the correcting of the depth image using the least square method comprises:

calculating the gap changes Δl 1 and Δl 2 of the two adjacent depth cameras by minimizing a cost function J=kΣ i=1 m (Z 1 (i) −Z 2 (i) −ΔZ (i) ) 2 , and correcting the depth image of the at least two depth cameras according to the gap changes Δl 1 and Δl 2 , wherein k is a coefficient.