Method and device for extracting key frames in simultaneous localization and mapping and smart device

1. A method for extracting key frames in simultaneous localization and mapping, the method comprises:

acquiring an image frame from an image library storing a plurality of image frames of an unknown environment, and performing feature extraction on the image frame to obtain a quantity of feature points;

acquiring relative motion information including a position variation and an angle variation of the image frame relative to a previous key frame;

by using the acquired relative motion information, determining a binary function relationship of a relative motion index value related to a third constant, the position variation, and the angle variation, and calculating and obtaining a corresponding relative motion index value;

calculating the adaptive threshold currently used according to the relative motion index value, a first constant indicating a relative motion index threshold, and a second constant indicating a feature point quantity threshold;

selecting a key frame according to the information of feature points and the adaptive threshold indicating space information of image frames,

wherein acquiring relative motion information of the image frame relative to a previous key frame comprises:

acquiring three-axis acceleration data and three-axis angular velocity data of a smart device, and acquiring angular displacement data of the smart device;

performing time alignment and spatial alignment on the three-axis acceleration data, the three-axis angular velocity data and the angular displacement data acquired according to a time stamp of the image frame;

after the time alignment and spatial alignment, fusing the three-axis acceleration data, the three-axis angular velocity data, the angular displacement data and the image frame using a predetermined filtering algorithm to obtain a pose value corresponding to the image frame, the pose value including position information and angle information; and

calculating a difference between the pose value corresponding to the image frame and a pose value of the previous key frame to obtain a position variation and an angle variation from a time point of the previous key frame to a current time point,

wherein:

acquiring three-axis acceleration data and three-axis angular velocity data of the smart device comprises: acquiring collected three-axis acceleration data and three-axis angular velocity data from an inertial measurement unit mounted on the smart device;

acquiring angular displacement data of the smart device comprises: acquiring collected angular displacement data from an encoding disk sensor mounted on the smart device;

acquiring the image frame collected by the smart device comprises: acquiring collected image frame from a vision sensor mounted on the smart device.

2. The method according to claim 1 , wherein the method further comprises:

with respect to the preliminary key frame determined, calculating a corresponding relative motion index value by using the relative motion information acquired;

judging whether the corresponding relative motion index value of the preliminary key frame is greater than a preset index threshold; and

when the corresponding relative motion index value of the preliminary key frame is greater than the preset index threshold, determining that the preliminary key frame is a final key frame.

3. The method according to claim 1 , wherein selecting a key frame according to the information of feature points and the adaptive threshold indicating space information of image frames comprises:

judging whether the quantity of feature points is greater than a corresponding adaptive threshold; and

if the quantity of feature points is greater than the corresponding adaptive threshold, determining that the image frame is a preliminary key frame.

4. A smart device, comprising a device body,

the smart device is an intelligent unmanned vehicle;

the device body is a vehicle body having a head and a tail;

the wheel comprises a driving wheel and a driven wheel,

wherein at least one wheel for moving is mounted at a bottom of the device body, and a sensor is provided on the device body;

the sensor comprises a vision sensor which collects an image of an unknown environment to obtain an image frame, stores the acquired image frame into an image library; and

the sensor comprises a motion sensor which acquires relative motion information of the image frame relative to a previous key frame,

the motion sensor comprises an inertial measurement unit for collecting acceleration information and angular velocity information of motion of the vehicle body, and an encoding disk sensor for collecting angular displacement of motion of the vehicle body,

wherein the smart device is configured to:

acquire the image frame from the image library, and perform feature extraction on the image frame to obtain a quantity of feature points;

acquire relative motion information including a position variation and an angle variation of the image frame relative to a previous key frame;

by using the acquired relative motion information, determine a binary function relationship of the relative motion index value related to a third constant, the position variation, and the angle variation, and calculate and obtain a corresponding relative motion index value;

calculate the adaptive threshold currently used according to the relative motion index value, a first constant indicating a relative motion index threshold, and a second constant indicating a feature point quantity threshold;

select a key frame according to the information of feature points and the adaptive threshold indicating space information of image frames,

wherein when acquiring relative motion information including a position variation and an angle variation of the image frame relative to a previous key frame:

acquire three-axis acceleration data and three-axis angular velocity data of a smart device, and acquire angular displacement data of the smart device;

perform time alignment and spatial alignment on the three-axis acceleration data, the three-axis angular velocity data and the angular displacement data acquired according to a time stamp of the image frame;

after the time alignment and spatial alignment, fuse the three-axis acceleration data, the three-axis angular velocity data, the angular displacement data and the image frame using a predetermined filtering algorithm to obtain a pose value corresponding to the image frame, the pose value including position information and angle information; and

calculate a difference between the pose value corresponding to the image frame and a pose value of the previous key frame to obtain a position variation and an angle variation from a time point of the previous key frame to a current time point,

when acquiring three-axis acceleration data and three-axis angular velocity data of the smart device,

acquire collected three-axis acceleration data and three-axis angular velocity data from an inertial measurement unit mounted on the smart device;

when acquiring angular displacement data of the smart device, acquire collected angular displacement data from an encoding disk sensor mounted on the smart device; and

when acquiring the image frame collected by the smart device, acquire collected image frame from a vision sensor mounted on the smart device.

5. The smart device according to claim 4 , wherein the inertial measurement unit is mounted on the tail, and a distance from a center of the vehicle body to the inertial measurement unit is greater than a predetermined distance threshold.

6. The smart device according to claim 4 , wherein the encoding disk sensor is mounted on rotating shafts of the driving wheel and the driven wheel, and is connected to a driving motor of the rotating shafts.

7. The smart device according to claim 4 , wherein the vision sensor is mounted at a position on the head where light is not blocked.