Motion estimation method based on machine vision
By acquiring and analyzing images on the motion path, combining the relative position and motion trajectory of the interval area, the problem of high cost of motion measurement hardware in the prior art is solved, and equipment-free motion distance and velocity estimation is achieved, with high general fitness and adaptability.
Patent Information
- Application Number
- CN202310749608.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-25
- Publication Date
- 2024-12-27
AI Technical Summary
The existing motion measurement solutions require wearable devices or embedded sensors, which have high hardware costs and do not have the universality and adaptability of motion paths.
Using a motion estimation method based on machine vision, by arranging the interval areas on the motion path, collecting images containing the characteristics of the interval area, filtering the characteristic images of the moving individuals, estimating the starting and ending positions of the moving individuals, and combining the relative positions of the interval areas and motion trajectory, calculate the motion distance and speed.
It realizes motion distance and speed estimation without wearable devices or buried sensors, which is highly universal and adaptable, reduces hardware costs, and can effectively store and analyze motion data.
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of motion testing based on machine vision, and in particular to a motion estimation method based on machine vision. Background Art
[0002] In the existing technology, the training or assessment of some sports has been automated or even intelligent. For example, in sports such as running and race walking, it is necessary to measure the movement distance, movement time, movement speed, etc. of the movement target. However, the existing movement measurement solutions, such as sensor detection, require wearable devices or sensor embedding, etc., and the hardware cost is relatively high. Summary of the invention
[0003] In order to solve the technical problems existing in the background technology, the present invention proposes a motion estimation method based on machine vision, comprising the following steps:
[0004] Pre-storing the relative positions of a plurality of interval areas arranged at intervals on the motion path, and initiating the acquisition of a first image containing features of the interval areas for the plurality of interval areas;
[0005] Screening out a plurality of first-type images that all contain motion individual features from the first image;
[0006] Filtering out a plurality of second-type images corresponding to the starting motion and the ending motion of the moving individual from the first-type images;
[0007] The starting movement position and the ending movement position of the moving individual are estimated by using the relative positions of the multiple interval areas corresponding to the multiple second type images and the multiple pre-stored interval areas;
[0008] The movement distance of the moving individual is estimated based on the starting movement position and the ending movement position.
[0009] “Estimating the movement distance of a moving individual” specifically includes:
[0010] The motion trajectory of the moving individual is obtained by comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path;
[0011] The movement distance of the moving individual is estimated by comprehensively analyzing the movement trajectory, the relative positions of the pre-stored interval areas, the starting movement position of the moving individual and the ending movement position of the moving individual.
[0012] “Estimating the movement distance of a moving individual” specifically includes:
[0013] The motion trajectory of the moving individual is obtained by comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path;
[0014] When it is analyzed that the above motion trajectory does not match the preset motion trajectory, the effective motion trajectory of the moving individual is analyzed, and a comprehensive analysis is performed on the effective motion trajectory, the relative positions of the pre-stored interval areas, the starting motion position of the moving individual, and the ending motion position of the moving individual to estimate the movement distance of the moving individual.
[0015] Furthermore: the start time and the end time of the movement of the individual movement are obtained, and the movement speed of the individual movement is estimated in combination with the movement distance of the individual movement.
[0016] The “relative position of the spacing area” specifically refers to: a plurality of relative position coordinates of a plurality of calibration points on the motion path, and a one-to-one correspondence between the positions of the plurality of calibration points and the plurality of calibration point features falling within the spacing area in the second type of image.
[0017] “Estimating the starting and ending positions of a moving individual” specifically includes:
[0018] Acquire a first positional relationship between a moving individual feature in a second type of image and a plurality of calibration point features falling in the second type of image;
[0019] The starting movement position and the ending movement position of the moving individual are estimated according to the multiple relative position coordinates of the multiple calibration points corresponding to the pre-stored multiple calibration point features and the first position relationship.
[0020] “Acquiring a first positional relationship between a moving individual feature in a second type of image and a plurality of calibration point features falling in the second type of image” specifically includes:
[0021] Multiple groups of calibration points are set correspondingly in multiple interval areas, and the relative position coordinates of all calibration points on the motion path are pre-stored;
[0022] Collecting multiple original images, each of the multiple original images contains multiple interval area features in a one-to-one correspondence, each of the original images contains a set of calibration point features, and the original images and the first image containing the same interval area features are obtained through fixed-point acquisition by the same acquisition terminal;
[0023] Extracting a second positional relationship between a plurality of calibration point features contained in the original image and the original image;
[0024] The second positional relationship between the original image obtained by fixed-point acquisition by the same acquisition terminal as the second type of image and the multiple calibration point features therein is called to establish the first positional relationship between the moving individual features in the second type of image and the multiple calibration point features.
[0025] Furthermore: the motion path is a circular runway.
[0026] Furthermore: a plurality of interval areas arranged at intervals on the motion path are adjacent to each other.
[0027] In the present invention, the proposed motion estimation method based on machine vision abandons the traditional method of burying sensors or using wearable devices to estimate the distance and speed during exercise. The sports individual can use any position on the motion path as the starting position and the ending position, and can also practice or test any distance. Similarly, for free walking or running, it can also effectively estimate and store relevant data. It has extremely strong universality and is adaptable to aspects such as digital statistical motion data and analysis of the motion process. DETAILED DESCRIPTION
[0028] The present invention proposes a motion estimation method based on machine vision, comprising the following steps:
[0029] S1. Pre-store relative positions of a plurality of interval areas arranged at intervals on a motion path, and initiate acquisition of a first image containing features of the interval areas for the plurality of interval areas.
[0030] The motion path may be a runway or a trail, or a circular runway. The interval area is a partial area on the motion path. The interval area fully covers the motion path in the width direction. The acquisition of the first image containing the characteristics of the interval area for multiple interval areas can be achieved by a terminal arranged on the motion path or on the side of the motion path. The terminal may be an image acquisition device or a video acquisition device. One or more terminals may be arranged for one interval area.
[0031] The multiple interval areas arranged at intervals on the motion path are adjacent to each other, so that there is no visual blind area in the entire motion path.
[0032] S2. Filter out a plurality of first-type images that all contain motion individual features from the first image.
[0033] Determining whether the first image is a first type image including the moving individual features may be achieved through existing technologies such as face recognition, and all the moving individual features correspond to the same moving individual.
[0034] S3. Filter out a plurality of second-type images corresponding to the starting motion and the ending motion of the moving individual from the first-type images.
[0035] The method of identifying the second type of image corresponding to the start and end of motion from the first image belongs to the existing technology, and the corresponding technical configuration can be carried out according to different scenarios, such as by identifying and analyzing multiple consecutive frames of first images, judging by speed changes, and assisting in judging entry or exit through image recognition, or obtaining it through motion triggering and other methods.
[0036] S4. Estimate the starting movement position and the ending movement position of the moving individual through the relative positions of the multiple interval areas corresponding to the multiple second type images and the multiple pre-stored interval areas.
[0037] Specific implementation method 1 of S4 is:
[0038] After acquiring the second type of image, the moving individual can be judged from which interval area the movement starts and which interval area the movement ends by tracing back the interval areas contained in the second type of image, and the starting and ending movement positions of the moving individual can be estimated based on the relative positions of multiple pre-stored interval areas.
[0039] In the specific implementation 2 of S4, the "relative position of the spacing area" is specifically: multiple relative position coordinates of multiple calibration points on the motion path, and the positions between the multiple calibration points and the multiple calibration point features falling into the spacing area in the second type of image correspond one to one.
[0040] Specific implementation method 2 of S4 includes the following steps:
[0041] S41, obtaining a first positional relationship between a moving individual feature in a second type of image and a plurality of calibration point features falling in the second type of image.
[0042] S41 specifically includes the following steps:
[0043] S411, setting a plurality of groups of calibration points in a plurality of interval areas, and pre-storing the relative position coordinates of all the calibration points on the motion path.
[0044] The multiple groups of calibration points are set for visual calibration in interval areas on the motion path.
[0045] S412, collecting multiple original images, each of which contains multiple interval area features in a one-to-one correspondence, each of which contains a set of calibration point features, and the original images and the first image containing the same interval area features are obtained through fixed-point collection by the same collection terminal.
[0046] S413: Extracting a second positional relationship between a plurality of calibration point features contained in the original image and the original image.
[0047] The second positional relationship has nothing to do with the content of the original image, and only the positions of the group of calibration point features in the original image need to be calibrated.
[0048] S414, calling the second positional relationship between the original image obtained by fixed-point acquisition by the same acquisition terminal as the second type of image and the multiple calibration point features therein to establish the first positional relationship between the moving individual features in the second type of image and the multiple calibration point features.
[0049] Since the original image and the first image containing the same interval area features are obtained through fixed-point acquisition by the same acquisition terminal, the positions where the group of calibration point features appear in the first image are the same as the positions where the group of calibration point features appear in the original image. That is to say, even if the calibration points disappear for some reason, the first positional relationship between the moving individual features and the multiple calibration point features in the second type of image can be established by calling the second positional relationship.
[0050] S42, estimating the starting movement position and the ending movement position of the moving individual according to the multiple relative position coordinates of the multiple calibration points corresponding to the pre-stored multiple calibration point features and the first position relationship.
[0051] After establishing the first positional relationship between the moving individual features and multiple calibration point features in the second type of image, the position coordinates of the moving individual at the start of movement and the position coordinates of the moving individual at the end of movement can be estimated by analyzing the first positional relationship and the known position coordinates of the multiple calibration point features.
[0052] S5. Estimate the movement distance of the moving individual according to the starting movement position and the ending movement position.
[0053] Specific implementation mode 1 of S5 is:
[0054] Since the relative positions of a plurality of interval areas arranged at intervals on the motion path are pre-stored, the interval area corresponding to the starting motion position and the interval area corresponding to the ending motion position can be calculated to estimate the motion distance of the moving individual.
[0055] Specific implementation method 2 of S5 includes the following steps:
[0056] S51, obtaining a motion trajectory of a moving individual by comprehensively analyzing a plurality of acquisition times corresponding to a plurality of first-type images, a plurality of interval regions corresponding to interval region features contained in a plurality of first-type images, and an arrangement rule of the plurality of interval regions on a motion path.
[0057] In the case of free running or free walking, the motion trajectory of the moving individual may not be a direct route from the starting motion position to the ending motion position, and there may be round trips, etc. By comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path, the motion trajectory of the moving individual can be determined by the order in which the moving individual passes through each interval area.
[0058] S52. When it is analyzed that the above-mentioned motion trajectory does not match the preset motion trajectory, the effective motion trajectory of the moving individual is analyzed, and a comprehensive analysis is performed on the effective motion trajectory, the relative positions of the pre-stored interval areas, the starting motion position of the moving individual, and the ending motion position of the moving individual to estimate the movement distance of the moving individual.
[0059] For example, when an individual is exercising on a circular track, the individual's motion trajectory is analyzed and it is determined that the individual's motion trajectory does not match the preset motion trajectory. Therefore, it is judged that the individual may deviate from the motion path and take a shortcut. The effective motion trajectory of the individual is re-identified by removing the part of the motion path that does not conform to the rules, and then the effective motion trajectory of the individual is obtained.
[0060] After analyzing the effective motion trajectory of the moving individual, since the position of the interval area is determined, the distance between the interval areas can be accurately calculated. Combined with the starting motion position and the ending motion position of the moving individual, the movement distance of the moving individual can be estimated.
[0061] Specific implementation method 3 of S5 is:
[0062] S51, obtaining a motion trajectory of a moving individual by comprehensively analyzing a plurality of acquisition times corresponding to a plurality of first-type images, a plurality of interval regions corresponding to interval region features contained in a plurality of first-type images, and an arrangement rule of the plurality of interval regions on a motion path.
[0063] In the case of free running or free walking, the motion trajectory of the moving individual may not be a direct route from the starting motion position to the ending motion position, and there may be round trips, etc. By comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path, the motion trajectory of the moving individual can be determined by the order in which the moving individual passes through each interval area.
[0064] S52, by comprehensively analyzing the movement trajectory, the relative positions of the pre-stored interval areas, the starting movement position of the movement individual, and the ending movement position of the movement individual, the movement distance of the movement individual is estimated.
[0065] After analyzing the motion trajectory of the moving individual, since the position of the interval area is determined, the distance between the interval areas can be accurately calculated. Even if the moving individual runs back and forth, the moving distance of the moving individual can be estimated as long as the motion trajectory of the moving individual and the starting and ending positions of the moving individual are obtained.
[0066] S6. Obtain the movement start time and movement end time of the movement individual, and estimate the movement speed of the movement individual based on the movement distance of the movement individual.
[0067] The average speed of the individual can be estimated based on knowing the movement distance and movement time.
[0068] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A motion estimation method based on machine vision, characterized in that: The following steps are involved: Pre-storing the relative positions of a plurality of interval areas arranged at intervals on the motion path, and initiating the acquisition of a first image containing features of the interval areas for the plurality of interval areas; Screening out a plurality of first-type images that all contain motion individual features from the first image; Filtering out a plurality of second-type images corresponding to the starting motion and the ending motion of the moving individual from the first-type images; The starting movement position and the ending movement position of the moving individual are estimated by using the relative positions of the multiple interval areas corresponding to the multiple second type images and the multiple pre-stored interval areas; The movement distance of the moving individual is estimated based on the starting movement position and the ending movement position.
2. The motion estimation method based on machine vision as claimed in claim 1, characterized in that: "Estimating the movement distance of a moving individual" specifically includes: The motion trajectory of the moving individual is obtained by comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path; The movement distance of the moving individual is estimated by comprehensively analyzing the movement trajectory, the relative positions of the pre-stored interval areas, the starting movement position of the moving individual and the ending movement position of the moving individual.
3. The motion estimation method based on machine vision as claimed in claim 1, characterized in that: "Estimating the movement distance of a moving individual" specifically includes: The motion trajectory of the moving individual is obtained by comprehensively analyzing the multiple acquisition times corresponding to the multiple first-type images, the multiple interval areas corresponding to the interval area features contained in the multiple first-type images, and the arrangement rules of the multiple interval areas on the motion path; When it is analyzed that the above motion trajectory does not match the preset motion trajectory, the effective motion trajectory of the moving individual is analyzed, and a comprehensive analysis is performed on the effective motion trajectory, the relative positions of the pre-stored interval areas, the starting motion position of the moving individual, and the ending motion position of the moving individual to estimate the movement distance of the moving individual.
4. The motion estimation method based on machine vision as claimed in claim 1, characterized in that: The start time and end time of the movement of the individual are obtained, and the movement speed of the individual is estimated based on the movement distance of the individual.
5. The method for motion estimation based on machine vision as claimed in claim 1, characterized in that: The “relative position of the spacing area” specifically refers to: a plurality of relative position coordinates of a plurality of calibration points on the motion path, and a one-to-one correspondence between the positions of the plurality of calibration points and the plurality of calibration point features falling within the spacing area in the second type of image.
6. The method for motion estimation based on machine vision as claimed in claim 5, characterized in that: "Estimating the starting and ending positions of a moving individual" specifically includes: Acquire a first positional relationship between a moving individual feature in a second type of image and a plurality of calibration point features falling in the second type of image; The starting movement position and the ending movement position of the moving individual are estimated according to the multiple relative position coordinates of the multiple calibration points corresponding to the pre-stored multiple calibration point features and the first position relationship.
7. The motion estimation method based on machine vision as claimed in claim 6, characterized in that: “Acquiring a first positional relationship between a moving individual feature in a second type of image and a plurality of calibration point features falling in the second type of image” specifically includes: Multiple groups of calibration points are set correspondingly in multiple interval areas, and the relative position coordinates of all calibration points on the motion path are pre-stored; Collecting multiple original images, each of the multiple original images contains multiple interval area features in a one-to-one correspondence, each of the original images contains a set of calibration point features, and the original images and the first image containing the same interval area features are obtained through fixed-point acquisition by the same acquisition terminal; Extracting a second positional relationship between a plurality of calibration point features contained in the original image and the original image; The second positional relationship between the original image obtained by fixed-point acquisition by the same acquisition terminal as the second type of image and the multiple calibration point features therein is called to establish the first positional relationship between the moving individual features in the second type of image and the multiple calibration point features.
8. The method for motion estimation based on machine vision as claimed in claim 1, characterized in that: The motion path is a circular runway.
9. The motion estimation method based on machine vision as claimed in claim 1, characterized in that: A plurality of interval areas arranged at intervals on the moving path are adjacent to each other.