WO2018113433A1

WO2018113433A1 - Method for screening and spatially locating virtual reality feature points

Info

Publication number: WO2018113433A1
Application number: PCT/CN2017/109794
Authority: WO
Inventors: 李宗乘
Original assignee: 深圳市虚拟现实技术有限公司
Priority date: 2016-12-22
Filing date: 2017-11-07
Publication date: 2018-06-28
Also published as: CN106599929A; CN106599929B

Abstract

A method for screening and spatially locating virtual reality feature points, comprising the following steps S1: after ensuring that all infrared point sources are in an on state, a processing unit controls an infrared camera to capture an image of a virtual reality helmet and calculates the coordinates of the light spot of each infrared point source image; S2: the processing unit carries out ID identification on each light spot in an imaging picture and finds IDs corresponding to all of the light spots; S3: the processing unit controls at least four infrared point sources corresponding to the IDs to be in a lighted state and turns off the other infrared point sources, and the processing unit controls the infrared camera to capture an image of the virtual reality helmet and uses a perspective-n-point (PnP) algorithm to carry out operation positioning of the image; and S4: when the number of light spots in the imaging picture does not meet the number required by the PnP algorithm, steps S1 to S3 are carried out again. Compared to existing technology, said method turns off infrared point sources which complicate calculation, thereby increasing locating efficiency.

Description

Virtual reality feature point screening spatial positioning method

[0001] The present invention relates to the field of virtual reality, and more particularly to a virtual reality feature point screening space positioning method.

Background technique

[0002] Spatial positioning generally uses optical or ultrasonic modes for positioning and measurement, and the model is used to derive the spatial position of the object to be measured. The general virtual reality space positioning system uses the infrared point and the light-sensing camera to determine the spatial position of the object. The infrared point is at the front end of the near-eye display device. After positioning, the light-sensing camera captures the position of the infrared point and then derives the user's position. Physical coordinates. If you know the correspondence between at least three light sources and projections, call the PnP algorithm to get the spatial position of the helmet. The key to achieving this process is to determine the source ID (Identity) of the projection. The current virtual reality spatial positioning is determined by the inaccurate image recognition in a certain distance and direction. The corresponding projection light source I D吋 is too long and the image recognition is inaccurate, which affects the accuracy and efficiency of the positioning.

technical problem

[0003] In order to solve the defect that the current virtual reality space positioning accuracy and efficiency are not high, the present invention provides a virtual reality feature point screening spatial positioning method that can improve positioning accuracy and efficiency.

Problem solution

Technical solution

The technical solution adopted by the present invention to solve the technical problem thereof is as follows: A virtual reality feature point screening space positioning method is provided, which includes the following steps:

[0005] S1: ensuring that all infrared point light sources are in an on state, the processing unit controls the infrared camera to take an image of the virtual reality helmet, and calculates coordinates of the light spots of each of the infrared point source images;

[0006] S2: the processing unit performs ID identification on each light spot in the imaged image to find an ID corresponding to all the light spots;

[0007] S3: the processing image control the infrared point light source corresponding to the ID is at least 4 盏 in a lighting state, turning off the remaining infrared point light sources, and the processing unit controls the infrared camera to capture the virtual reality The image of the helmet is calculated and positioned using the PnP algorithm;

[0008] S4: When the number of spots on the imaged picture does not satisfy the number required by the PnP algorithm, S1 to S3 are re-executed.

[0009] Preferably, the shape of the image is rectangular, and the length of the long side of the image is d, the processing unit calculates the distance between the two spots, and selects the maximum distance d'. d'>d/2吋, the processing unit finds a spot of light closest to the center of the imaged image, and maintains the infrared point source of the spot corresponding to the ID and the three closest to the infrared point source The infrared point source is in a lit state, and the other infrared point sources are turned off at the same time.

[0010] Preferably, the shape of the image is rectangular, and the length of the long side of the image is d, the processing unit calculates the distance between the two spots, and selects the maximum distance d'. d' < d/2吋, the processing unit finds at least four of the infrared point light sources outside the relative positions of the infrared point light sources corresponding to the light spots and keeps lighting, and turns off the other infrared point light sources.

[0011] Preferably, the processing unit combines the known historical information of the previous frame to make a slight translation of the light spot of the image of the previous frame, so that the light spot of the image of the previous frame is corresponding to the light spot of the current frame image. And determining, according to the correspondence relationship and the historical information of the previous frame, a corresponding ID of each light spot having a corresponding relationship on the current frame image.

Advantageous effects of the invention

Beneficial effect

[0012] Compared with the prior art, the present invention increases the efficiency of positioning by turning off the infrared point source which complicates the calculation, and uses the relative position of the infrared point source on the imaged image to filter the infrared point source that needs to be turned off. A screening method is given. The method of comparing the maximum distance between the light spots with the long side distance of the imaged picture to distinguish the closing of the infrared point light source corresponding to the light spot is simple and easy, and the operability is strong. When the maximum distance between the light spots is greater than half of the length of the long side of the image, the infrared point source corresponding to the four light spots selected in the middle is illuminated, which can be better calculated by the PnP algorithm, and the same is also ensured. The positioned light spots do not quickly move out of the imaged image, preventing repeated ID recognition and costing a lot of time. When the maximum distance between the light spots is less than half of the length of the long side of the image, the infrared point source corresponding to at least 4 light spots selected from the outside is illuminated, and the PnP algorithm can be used for calculation, and the light is also ensured. The distance between the spots is large enough that it will not be produced due to pixels or the like. A large error is caused by making a slight translation of the light spot to make the current light spot correspond to the light spot of the previous frame image, thereby avoiding repeated ID recognition and saving a lot of time.

Brief description of the drawing

DRAWINGS

[0013] The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

1 is a schematic diagram showing the principle of a virtual reality feature point screening spatial positioning method according to the present invention;

2 is a schematic diagram of an infrared point source distribution of a virtual reality feature point screening spatial positioning method according to the present invention;

[0016] FIG. 3 shows one of images taken by an infrared camera;

[0017] FIG. 4 illustrates one of the imaged images presented after the infrared point source is turned off;

[0018] FIG. 5 shows two images taken by an infrared camera;

[0019] FIG. 6 shows the second image of the image presented after the infrared point source is turned off.

BEST MODE FOR CARRYING OUT THE INVENTION

[0020] In order to solve the defect that the current virtual reality spatial positioning accuracy and efficiency are not high, the present invention provides a virtual reality feature point screening spatial positioning method that can improve positioning accuracy and efficiency.

[0021] In order to more clearly understand the technical features, objects and effects of the present invention, the embodiments of the present invention are described in detail with reference to the accompanying drawings.

[0022] Please refer to FIG. 1 and FIG. 2. The virtual reality feature point screening spatial positioning method comprises a virtual reality helmet 10, an infrared camera 20 and a processing unit 30, and the infrared camera 20 is electrically connected to the processing unit 30. The virtual reality helmet 10 includes a front panel 11, and a plurality of infrared point light sources 13 are distributed on the front panel 11 of the virtual reality helmet 10 and the four side panels of the upper, lower, left, and right sides. The number of infrared point sources 13 must be at least the minimum number that the PnP algorithm can operate. The shape of the infrared point light source 13 is not particularly limited. For the sake of illustration, we take the number of infrared point light sources 13 on the front panel 11 to be seven, and the seven infrared point light sources form a shape of approximately "w". A plurality of infrared point sources 13 can be illuminated or turned off as needed by the firmware interface of the virtual reality helmet 10. The infrared point light source 13 on the virtual reality helmet 10 forms a light spot on the image by the shooting of the infrared camera 20. Due to the band pass characteristic of the infrared camera, only the infrared point light source 13 can form a spot projection on the image, and the remaining portions are uniformly formed. Background image. The infrared point source 13 on the virtual reality helmet 10 can form a spot of light on the image. Referring to FIG. 3 and FIG. 4, FIG. 3 shows an image 4 of the infrared point source 13 captured by the infrared camera 20. The image of the image 41 is rectangular, and the length of the longer side of the rectangle is d. Ensuring that all of the infrared point sources are in an on state, the processing unit 30 controls the infrared camera 20 to take an image of the virtual reality helmet 10 with seven spots on the image 41. The processing unit 30 calculates the coordinates of each light spot based on the position of the light spot on the imaged picture 41, and measures the distance between the two light spots, from which the maximum distance d' is selected. When d'>d/2吋, it indicates that the range of the pupil spot on the image 41 is large, since each spot is sequentially ID

(Identity, serial number) The identification and operation of the PnP algorithm can take a lot of time. Only a part of the points can satisfy the needs of the PnP algorithm. Therefore, the processing unit 30 first performs ID identification on each spot in the image 41. Finding the ID corresponding to all the light spots, and then finding the light spot closest to the center position of the imaged image 41 as the center point, maintaining the infrared spot light source 13 corresponding to the ID of the light spot and the three infrared rays closest to the infrared point light source The point light source 13 is in a lighting state, and the other infrared point light sources 13 are turned off. Thereafter, there are only 4 light spots on the image 41 of the next frame, and the processing unit 30 can track each light spot and calibrate the corresponding ID. The method is: in spatial positioning, since the sampling time of each frame is small enough, generally 30 ms, in general, the position difference of each light spot of the previous frame and each light spot on the current frame is small, and the processing is small. The unit 30 combines the known historical information of the previous frame to make a slight translation of the light spot of the previous frame image to make the light spot of the previous frame image Frame image before the light spot is generated corresponding relationship, on the basis of the correspondence and a history information is determined to have a correspondence between each light spot corresponding to the ID of the current frame image. In the case where all the light spot corresponding IDs are known, the processing unit 30 directly calls the PnP algorithm to obtain the spatial positioning position of the virtual reality helmet 10. When the virtual reality helmet 10 causes the number of spots in the image 41 to be smaller than the number of spots required by the PnP algorithm, the above method is re-executed to select a new infrared point source 13 to be lit.

[0024] Please refer to FIG. 5 and FIG. 6. In FIG. 5, there are seven light spots on the imaged image 41. The processing unit 30 calculates the coordinates of each light spot based on the position of the light spot on the imaged picture 41, and measures the distance between the two light spots, from which the maximum distance d' is selected. When d'< d/2吋, it indicates that the range of the calender spot on the image 41 is small, and it takes a lot of time to perform ID recognition and PnP algorithm operation for each spot, and only some of them are taken. The PnP algorithm can also meet the needs of the PnP algorithm. Therefore, the processing unit 30 first ID identifies each spot in the image, finds the ID corresponding to all the light spots, and then finds the relative position of the infrared point source 13 corresponding to the IDs. Keeping these infrared point sources at least 4 infrared point sources 13 outside 13 is in the light state, and the other infrared point light sources 13 are turned off. This ensures that the light spots on the imaged image 41 are not too dense, thereby affecting the accuracy of the measurement. The processing unit 30 directly calls the PnP algorithm to obtain the spatial positioning position of the virtual reality helmet 10. When the virtual reality helmet 10 causes the number of spots in the imaged picture 41 to be less than the number of spots required by the PnP algorithm due to the movement, the above method is re-executed to select a new infrared point source 13 that needs to be illuminated.

[0025] After the ID identification is completed, the processing unit 30 calls the PnP algorithm to obtain the spatial positioning position of the helmet, Pn.

The P algorithm belongs to the prior art, and the present invention will not be described again.

Compared with the prior art, the present invention increases the efficiency of positioning by turning off the infrared point source 13 which complicates the calculation, and uses the relative position of the infrared point source 13 on the imaged picture 41 to filter the need to be closed. The infrared point source 13 gives a screening method. The method of comparing the maximum distance between the light spots with the long side distance of the imaged image 41 to distinguish the closing of the infrared point light source 13 corresponding to the light spot is simple and easy, and the operability is strong. When the maximum distance between the light spots is greater than half of the length of the long side of the image 41, the infrared point light source 13 corresponding to the four light spots selected in the middle portion is illuminated, and the PnP algorithm can be used for calculation, and the same is ensured. The spot of light used for positioning does not quickly move out of the imaged picture 41, preventing repeated ID recognition and taking a lot of time. When the maximum distance between the light spots is less than half of the length of the long side of the image 41, the infrared point light source 13 corresponding to at least four light spots selected from the outside is illuminated, and the PnP algorithm can be used for calculation, and the same is also ensured. The distance between the light spots is large enough, and there is no large error due to pixels or the like. By making a slight translation of the light spots, the current light spots are corresponding to the light spots of the previous frame image, thereby avoiding repeated ID recognition and saving. A lot of time.

The embodiments of the present invention have been described above with reference to the drawings, but the present invention is not limited to the specific embodiments described above, and the specific embodiments described above are merely illustrative and not restrictive. The skilled person is able to make various forms within the scope of the present invention without departing from the spirit and scope of the invention as claimed.

Claims

Claim

[Claim 1] A virtual reality feature point screening spatial positioning method, comprising the following steps

S1: ensuring that all the infrared point light sources are in an on state, the processing unit controls the infrared camera to take an image of the virtual reality helmet, and calculates coordinates of the light spots of each of the infrared point source images;

S2: the processing unit performs ID identification on each light spot in the imaged image to find an ID corresponding to all the light spots;

S3: the infrared light source of the processing image control corresponding ID is at least 4 盏 in a lighting state, and the remaining infrared point light sources are turned off, and the processing unit controls the infrared camera to capture an image of the virtual reality helmet and The PnP algorithm is used to perform operation and positioning; S4: When the number of spots on the imaged picture does not satisfy the number required by the PnP algorithm, S1 to S3 are re-executed.

The method for locating a virtual reality feature point screening space according to claim 1, wherein the shape of the image is rectangular, and the length of the long side of the image is d, the processing unit Calculating the distance between the two spots of the light spot, and selecting the maximum distance d', when d'>d/2吋, the processing unit finds the spot of light closest to the center position of the imaged image, and keeps the spot corresponding to the spot The infrared point source of the ID and the three infrared point sources closest to the infrared point source are in a lighting state, and the other infrared point sources are turned off.

The method for locating a virtual reality feature point screening space according to claim 1, wherein the shape of the image is rectangular, and the length of the long side of the image is d, the processing unit Calculating a distance between the two spots of the light spot, and selecting a maximum distance d′, wherein when d′ < d/2吋, the processing unit finds at least 4 of the relative positions of the infrared point light sources corresponding to the light spots. The infrared point source is kept lit, and the other infrared point sources are turned off.

The method for locating a virtual reality feature point screening space according to any one of claims 1 to 3, wherein the processing unit combines the historical information of the previous frame with the image of the previous frame. The light spot makes a slight translation so that the light spot of the previous frame image is corresponding to the light spot of the current frame image, and each light having a corresponding relationship on the current frame image is determined according to the correspondence relationship and the history information of the previous frame. The corresponding ID of the spot.