Target object throwing distance measuring method based on machine vision
By using a method based on machine vision, we analyze the image characteristics and relative position of the target object during movement, and fit parabolic rules to calculate the throwing distance. This solves the problems of high hardware cost and poor universality in the existing technology, and achieves high efficiency and low cost. Cost of motion measurement.
Patent Information
- Application Number
- CN202310887234.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-19
- Publication Date
- 2025-01-21
AI Technical Summary
The existing motion measurement solutions require auxiliary equipment or sensor burial, which has high hardware costs and does not have good universality and adaptability.
A method based on machine vision is used to obtain multiple images of the target from the initial throw to the landing, analyze the characteristics and relative position of the target, and fit the parabola rule to calculate the throwing distance.
There is no need for traditional hardware equipment layout, which reduces costs, improves the universality and adaptability of measurement, and can effectively digitize and analyze motion data.
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of motion testing based on machine vision, and in particular to a method for measuring a throwing distance of a target object 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 shot put and shot put, sports performance needs to be tested. However, existing sports measurement solutions such as sensor detection require auxiliary equipment or sensor embedding, 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 method for measuring the throwing distance of a target object based on machine vision, comprising the following steps:
[0004] Multiple images of multiple targets are acquired from the time when they are thrown to the ground and then come to rest, and any image contains multiple features of the targets;
[0005] Analyze, from the target features of the multiple images, a plurality of first-type target features belonging to a first target therein and a plurality of first relative positions of the plurality of first-type target features in the images;
[0006] Analyze the landing point of the first target object according to the multiple first relative positions and calculate the landing point of the first target object in the image;
[0007] Acquire a first positional relationship between the identification feature and the image, acquire a second positional relationship between the landing point and the identification feature according to the first positional relationship and the landing point, and calculate a second actual position of the landing point of the first target object according to the second positional relationship and a pre-stored first actual position of the identification feature;
[0008] The first throwing distance is calculated according to the second actual position and the pre-stored actual throwing starting position of the first target object.
[0009] Furthermore, the multiple first relative positions include the landing point of the first target object in the image.
[0010] Furthermore, “analyzing, from the target features of the multiple images, a plurality of first-type target features belonging to the first target therein and a plurality of first relative positions of the plurality of first-type target features in the images” specifically includes:
[0011] Establishing a first relative position corresponding to a first target feature in an initial frame image, and tracking one or more candidate first relative positions belonging to the first target feature in a second frame image having a preset positional relationship with the initial ID;
[0012] Tracking one or more first relative positions in the third frame image that have a preset positional relationship with the one or more first relative positions in the second frame image according to the one or more first relative positions in the second frame image, and excluding one or more candidate first relative positions in the second frame image that do not conform to a preset parabola rule;
[0013] This process is deduced in this way until multiple first-type target object features of all the first targets are acquired;
[0014] A plurality of first relative positions of a plurality of first-type target features in the images are analyzed from the target features of the plurality of images.
[0015] Furthermore, “analyzing, from the target features of the multiple images, a plurality of first-type target features belonging to the first target therein and a plurality of first relative positions of the plurality of first-type target features in the images” specifically includes:
[0016] Extract the relative positions of target features in multiple images;
[0017] The multiple relative positions are fitted into multiple trajectory curves according to a preset parabola rule, the multiple trajectory curves include a first trajectory curve corresponding to the first target object, and the multiple relative positions fitted into the first trajectory curve are multiple first relative positions.
[0018] Furthermore, “analyzing the landing point of the first target object according to the multiple first relative positions” specifically means that the last first relative position involved in fitting the first trajectory line is the landing point of the first target object.
[0019] Furthermore, the one or more first relative positions that cannot be fitted into the first trajectory line due to not complying with the parabola rule but belong to the first target object include at least one first relative position that has a preset first position relationship with the landing point of the first target object.
[0020] Furthermore, there is a preset second positional relationship between two adjacent ones of the plurality of first relative positions involved in fitting the first trajectory line. Preferably, any frame of image contains and only contains one first relative position.
[0021] Furthermore, it also includes:
[0022] Obtaining the relative positions of multiple motion individual features contained in the image;
[0023] A correspondence between the first target object and the first moving individual in the multiple moving individual features is established according to the first trajectory curve and the relative positions of the multiple moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
[0024] Furthermore, “analyzing the landing point of the first target object according to the plurality of first relative positions” specifically includes:
[0025] When the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X+1th frame image, and the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X-1th frame image, then the first relative position appearing in the Xth frame image is the landing point of the first target object.
[0026] Furthermore, it also includes:
[0027] Fitting a first trajectory curve from the plurality of first relative positions according to a parabola rule;
[0028] Obtaining the relative positions of multiple motion individual features contained in the image;
[0029] A correspondence between the first target object and the first moving individual in the multiple moving individual features is established according to the first trajectory curve and the relative positions of the multiple moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
[0030] Furthermore, it also includes:
[0031] Establishing the corresponding relationship between the multiple sub-item regional features and the multiple motion individual features contained in the image;
[0032] Determine the sub-project area where the first target object falls;
[0033] The corresponding relationship between the first throwing distance and the first sports individual is obtained according to the sub-item area where the landing point of the first target object is located and the first corresponding relationship.
[0034] Furthermore, “obtaining a first positional relationship between the identification feature and the image” is specifically:
[0035] Set a visual mark and save the actual position of the visual mark in advance;
[0036] Capturing a first image, the first image containing identification features, the first image and the image being acquired through fixed-point acquisition by the same acquisition terminal;
[0037] The positional relationship between the identification feature extracted from the first image and the first image can be called as the first positional relationship between the dot matrix identification feature and the image. Preferably, the identification is specifically a dot matrix composed of multiple identification points.
[0038] In the present invention, the proposed method for measuring the throwing distance of a target object based on machine vision abandons the traditional method of burying sensors or using wearable devices, and estimates the throwing distance during the motion process based on machine vision. It has low requirements for hardware layout, has strong universality, and is adaptable to aspects such as digital statistical motion data and analysis of the motion process. DETAILED DESCRIPTION
[0039] Example 1
[0040] A method for measuring the throwing distance of a target object based on machine vision proposed in Example 1 includes the following steps:
[0041] S1. Obtain multiple images of multiple targets from the time when they are thrown to the ground and then come to rest at a fixed point, and any image contains multiple target features.
[0042] The actual scene can be multi-person shot put or shot put distance measurement, the target object is a shot put or a solid ball, and multiple images are obtained by fixed-point acquisition using the same image acquisition device.
[0043] S2. Analyze the target features in the multiple images to obtain multiple first-type target features belonging to the first target and multiple first relative positions of the multiple first-type target features in the images.
[0044] S2 specifically includes:
[0045] S21, extracting the relative positions of target features in multiple images.
[0046] S22, fitting multiple relative positions into multiple trajectory curves according to a preset parabola rule, wherein the multiple trajectory curves include a first trajectory curve corresponding to the first target object, and the multiple relative positions fitted into the first trajectory curve are multiple first relative positions.
[0047] The trajectory of the target from the time it is thrown to the time it stops after landing can be divided into two stages:
[0048] The first stage is from the initial throw to the landing point, and the second stage is from the landing point to the stopping point.
[0049] Based on common sense, the first stage of natural flight in the air basically conforms to a certain existing parabolic rule; while the second stage, as a bounce from touching the ground or rolling to a standstill, does not follow any rules and is not applicable to any parabolic rule.
[0050] If you want to get the final distance, the landing point is the key.
[0051] Since the targets fly freely with each other, have their own starting points and end points, and also have their own flight trajectories, multiple trajectory curves can be fitted according to the preset parabola rules based on multiple relative positions.
[0052] Any frame image contains and only contains one first relative position, and two adjacent ones of the multiple first relative positions belong to adjacent frame images. Then, there is a preset second position relationship between two adjacent ones of the multiple first relative positions for fitting the first trajectory line. The second relative position can be a preset distance between the two, a position relationship between the two, etc. The first target object is one of the multiple targets, and the first type of target feature also belongs to one of the multiple target features. The multiple first relative positions naturally include the landing point of the first target object in the image.
[0053] S3. Analyze the landing point of the first target object according to the multiple first relative positions and calculate the landing point of the first target object in the image.
[0054] The end point of the first stage is the landing point, and the last first relative position of the first target object participating in the fitting in the image is the landing point. Although the multiple first relative positions of the second stage do not participate in the fitting of the first trajectory curve because they do not conform to the parabola rule, the first stage is adjacent to the second stage and the landing point is used as the dividing point. Then, at least one of the multiple first relative positions of the second stage is adjacent to the landing point. The accuracy of the determined landing point can be verified by analyzing whether the multiple first relative positions of the second stage include a first relative position adjacent to the landing point.
[0055] S4. Obtain a first positional relationship between the identification feature and the image.
[0056] Specifically include:
[0057] S41. Setting a visual mark, and pre-storing the actual position of the visual mark.
[0058] The identification is specifically a dot matrix composed of multiple identification points, and multiple calibration points are actually calibrated on the site to form a dot matrix.
[0059] S42: Capture a first image, wherein the first image contains identification features, and the first image and the image are both acquired through fixed-point acquisition by the same acquisition terminal.
[0060] Since the dot matrix marks on the site are visualized, they can be collected. The first image and the image are both obtained through fixed-point collection by the same collection terminal, so that the relative positions of the dot matrix marks and target features in the image and the first image are consistent.
[0061] S43: The positional relationship between the identification feature extracted from the first image and the first image can be called as the first positional relationship between the dot matrix identification feature and the image.
[0062] Even if the identification features become unclear or even disappear on the site, it will not affect.
[0063] S5. Acquire a second positional relationship between the landing point and the identification feature according to the first positional relationship and the landing point, and calculate a second actual position of the landing point of the first target object according to the second positional relationship and the pre-stored first actual position of the identification feature.
[0064] The identification feature is preferably a dot matrix identification feature. When the second position relationship between the landing point and the identification feature is obtained, it is equivalent to knowing between which points in the dot matrix the landing point is located or which point it coincides with. The first actual position of the pre-stored identification feature refers to the actual position relationship between multiple point features contained in the pre-stored dot matrix identification feature and / or the actual positions of multiple point features. When it is known between which points in the dot matrix the landing point is located or which point it coincides with, and the actual position relationship between the multiple point features contained in the dot matrix identification feature and / or the actual positions of multiple point features are known, the second actual position of the landing point can naturally be calculated within the allowable error range.
[0065] S6. Calculate a first throwing distance according to the second actual position and the pre-stored actual throwing starting position of the first target object.
[0066] In multi-person shot put or multi-person shot put events, the actual starting position of the throw is preset.
[0067] S7. Obtaining a corresponding relationship between the first throwing distance and the first moving individual.
[0068] Specifically include:
[0069] S71. Obtain the relative positions of a plurality of motion individual features contained in the image.
[0070] The individual motion features of the thrown object need to be captured in the image.
[0071] S72, establishing a correspondence between the first target object and the first moving individual in the plurality of moving individual features according to the first trajectory curve and the relative positions of the plurality of moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
[0072] Since the target object is thrown by the sports individual, that is, there is a certain positional relationship between the starting point of the trajectory curve of the target object and the sports individual, a correspondence between the first target object and the first sports individual among multiple sports individual characteristics can be established based on this positional relationship. In layman's terms, the correspondence between the sports individual and the performance can also be obtained.
[0073] Example 2
[0074] A method for measuring the throwing distance of a target object based on machine vision proposed in Example 2 includes the following steps:
[0075] S1. Obtain multiple images of multiple targets from the time when they are thrown to the ground and then come to rest at a fixed point, and any image contains multiple target features.
[0076] The actual scene can be multi-person shot put or shot put distance measurement, the target object is a shot put or a solid ball, and multiple images are obtained by fixed-point acquisition using the same image acquisition device.
[0077] S2. Analyze, from the object features of the multiple images, a plurality of first-type object features belonging to the first object therein and a plurality of first relative positions of the plurality of first-type object features in the images.
[0078] Specifically include:
[0079] S21, establishing a first relative position corresponding to a first target feature in an initial frame image, and tracking one or more candidate first relative positions belonging to the first target feature in a second frame image having a preset position relationship with the initial ID.
[0080] S22. Track one or more first relative positions in the third frame image that have a preset positional relationship with the one or more first relative positions in the second frame image based on the one or more first relative positions in the second frame image, and exclude one or more candidate first relative positions in the second frame image that do not conform to the preset parabola rule.
[0081] S23, and so on, until all the first-type target object features belonging to the first target object are acquired.
[0082] S24: Analyze the target features of the multiple images to obtain multiple first relative positions of multiple first-type target features in the images.
[0083] Among them, the multiple first relative positions include the landing point of the first target object.
[0084] S3. Analyze the landing point of the first target object according to the multiple first relative positions and calculate the landing point of the first target object in the image.
[0085] Specifically: when the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X+1th frame image, and the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X-1th frame image, then the first relative position appearing in the Xth frame image is the landing point of the first target object.
[0086] S31. Setting a visual mark and pre-storing the actual position of the visual mark.
[0087] The identification is specifically a dot matrix composed of multiple identification points, and multiple calibration points are actually calibrated on the site to form a dot matrix.
[0088] S32, capturing a first image, wherein the first image contains identification features, and the first image and the image are both captured by the same acquisition terminal at a fixed point.
[0089] Since the dot matrix marks on the site are visualized, they can be collected. The first image and the image are both obtained through fixed-point collection by the same collection terminal, so that the relative positions of the dot matrix marks and target features in the image and the first image are consistent.
[0090] S33: The positional relationship between the identification feature extracted from the first image and the first image can be called as the first positional relationship between the dot matrix identification feature and the image.
[0091] Even if the identification features become unclear or even disappear on the site, it will not affect.
[0092] S4. Acquire a second positional relationship between the landing point and the identification feature according to the first positional relationship and the landing point, and calculate a second actual position of the landing point of the first target object according to the second positional relationship and the pre-stored first actual position of the identification feature.
[0093] The identification feature is preferably a dot matrix identification feature. When the second position relationship between the landing point and the identification feature is obtained, it is equivalent to knowing between which points in the dot matrix the landing point is located or which point it coincides with. The first actual position of the pre-stored identification feature refers to the actual position relationship between multiple point features contained in the pre-stored dot matrix identification feature and / or the actual positions of multiple point features. When it is known between which points in the dot matrix the landing point is located or which point it coincides with, and the actual position relationship between the multiple point features contained in the dot matrix identification feature and / or the actual positions of multiple point features are known, the second actual position of the landing point can naturally be calculated within the allowable error range.
[0094] S5. Calculate a first throwing distance according to the second actual position and the pre-stored actual throwing starting position of the first target object.
[0095] In multi-person shot put or multi-person shot put events, the actual starting position of the throw is preset.
[0096] S6. Obtaining a corresponding relationship between the first throwing distance and the first moving individual.
[0097] Specifically include:
[0098] S61, fitting a first trajectory curve from a plurality of first relative positions according to a parabola rule.
[0099] S62: Obtain the relative positions of a plurality of motion individual features contained in the image.
[0100] S63, establishing a correspondence between the first target object and the first moving individual in the plurality of moving individual features according to the first trajectory curve and the relative positions of the plurality of moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
[0101] Since the target object is thrown by the sports individual, that is, there is a certain positional relationship between the starting point of the trajectory curve of the target object and the sports individual, a correspondence between the first target object and the first sports individual among multiple sports individual characteristics can be established based on this positional relationship. In layman's terms, the correspondence between the sports individual and the performance can also be obtained.
[0102] Example 3
[0103] A method for measuring the throwing distance of a target object based on machine vision proposed in Example 3 includes the following steps:
[0104] S1. Obtain multiple images of multiple targets from the time when they are thrown to the ground and then come to rest at a fixed point, and any image contains multiple target features.
[0105] The actual scene can be multi-person shot put or shot put distance measurement, the target object is a shot put or a solid ball, and multiple images are obtained by fixed-point acquisition using the same image acquisition device.
[0106] S2. Analyze, from the object features of the multiple images, a plurality of first-type object features belonging to the first object therein and a plurality of first relative positions of the plurality of first-type object features in the images.
[0107] The implementation of S2 may refer to Example 1 or Example 2, which will not be described in detail here.
[0108] The plurality of first relative positions include a point where the first target object falls in the image.
[0109] S3. Analyze the landing point of the first target object according to the multiple first relative positions and calculate the landing point of the first target object in the image.
[0110] S4. Obtain a first positional relationship between the identification feature and the image.
[0111] Specifically include:
[0112] S41. Setting a visual mark, and pre-storing the actual position of the visual mark.
[0113] The identification is specifically a dot matrix composed of multiple identification points, and multiple calibration points are actually calibrated on the site to form a dot matrix.
[0114] S42: Capture a first image, wherein the first image contains identification features, and the first image and the image are both acquired through fixed-point acquisition by the same acquisition terminal.
[0115] Since the dot matrix marks on the site are visualized, they can be collected. The first image and the image are both obtained through fixed-point collection by the same collection terminal, so that the relative positions of the dot matrix marks and target features in the image and the first image are consistent.
[0116] S43: The positional relationship between the identification feature extracted from the first image and the first image can be called as the first positional relationship between the dot matrix identification feature and the image.
[0117] Even if the identification features become unclear or even disappear on the site, it will not affect.
[0118] S5. Acquire a second positional relationship between the landing point and the identification feature according to the first positional relationship and the landing point, and calculate a second actual position of the landing point of the first target object according to the second positional relationship and the pre-stored first actual position of the identification feature.
[0119] The identification feature is preferably a dot matrix identification feature. When the second position relationship between the landing point and the identification feature is obtained, it is equivalent to knowing between which points in the dot matrix the landing point is located or which point it coincides with. The first actual position of the pre-stored identification feature refers to the actual position relationship between multiple point features contained in the pre-stored dot matrix identification feature and / or the actual positions of multiple point features. When it is known between which points in the dot matrix the landing point is located or which point it coincides with, and the actual position relationship between the multiple point features contained in the dot matrix identification feature and / or the actual positions of multiple point features are known, the second actual position of the landing point can naturally be calculated within the allowable error range.
[0120] S6. Calculate a first throwing distance according to the second actual position and the pre-stored actual throwing starting position of the first target object.
[0121] In multi-person shot put or multi-person shot put events, the actual starting position of the throw is preset.
[0122] S7. Obtaining a corresponding relationship between the first throwing distance and the first moving individual.
[0123] Specifically include:
[0124] S71. Establishing a correspondence between a plurality of sub-item regional features and a plurality of motion individual features contained in the image.
[0125] S72, determining the sub-item area where the landing point of the first target object is located;
[0126] S73. Obtain a correspondence between the first throwing distance and the first sports individual according to the sub-item area where the landing point of the first target object is located and the first correspondence.
[0127] In layman's terms, we can also obtain the corresponding relationship between individual athletes and their performance.
[0128] 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 method for measuring the throwing distance of a target object based on machine vision, characterized in that: The steps include: Multiple images of multiple targets are acquired from the time when they are thrown to the ground and then come to rest, and any image contains multiple features of the targets; Analyze, from the target features of the multiple images, a plurality of first-type target features belonging to a first target therein and a plurality of first relative positions of the plurality of first-type target features in the images; Analyze the landing point of the first target object according to the multiple first relative positions and calculate the landing point of the first target object in the image; Acquire a first positional relationship between the identification feature and the image, acquire a second positional relationship between the landing point and the identification feature according to the first positional relationship and the landing point, and calculate a second actual position of the landing point of the first target object according to the second positional relationship and a pre-stored first actual position of the identification feature; The first throwing distance is calculated according to the second actual position and the pre-stored actual throwing starting position of the first target object.
2. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 1, characterized in that: The plurality of first relative positions include a point where the first target object falls in the image.
3. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 1, characterized in that: “Analyzing, from the target features of the multiple images, a plurality of first-type target features belonging to the first target therein and a plurality of first relative positions of the plurality of first-type target features in the images” specifically means: Establishing a first relative position corresponding to a first target feature in an initial frame image, and tracking one or more candidate first relative positions belonging to the first target feature in a second frame image having a preset positional relationship with the initial ID; Tracking one or more first relative positions in the third frame image that have a preset positional relationship with the one or more first relative positions in the second frame image according to the one or more first relative positions in the second frame image, and excluding one or more candidate first relative positions in the second frame image that do not conform to a preset parabola rule; This process is deduced in this way until multiple first-type target object features of all the first targets are acquired; A plurality of first relative positions of a plurality of first-type target features in the images are analyzed from the target features of the plurality of images.
4. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 1, characterized in that: “Analyzing, from the target features of the multiple images, a plurality of first-type target features belonging to the first target therein and a plurality of first relative positions of the plurality of first-type target features in the images” specifically means: Extract the relative positions of target features in multiple images; The multiple relative positions are fitted into multiple trajectory curves according to a preset parabola rule, the multiple trajectory curves include a first trajectory curve corresponding to the first target object, and the multiple relative positions fitted into the first trajectory curve are multiple first relative positions.
5. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 4, characterized in that: “Analyzing the landing point of the first target object based on multiple first relative positions” specifically means: the last first relative position involved in fitting the first trajectory line is the landing point of the first target object.
6. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 5, characterized in that: The one or more first relative positions that cannot be fitted into the first trajectory line due to not complying with the parabola rule but belong to the first target object include at least one first relative position having a preset first position relationship with the landing point of the first target object.
7. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 4, characterized in that: There is a preset second positional relationship between two adjacent first relative positions among the plurality of first relative positions involved in fitting the first trajectory line. Preferably, any frame of image contains and only contains one first relative position.
8. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 4, characterized in that: Also includes: Obtaining the relative positions of multiple motion individual features contained in the image; A correspondence between the first target object and the first moving individual in the multiple moving individual features is established according to the first trajectory curve and the relative positions of the multiple moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
9. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 1, characterized in that: “Analyzing the landing point of the first target object according to the plurality of first relative positions” specifically includes: When the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X+1th frame image, and the first relative position appearing in the Xth frame image is lower than the first relative position appearing in the X-1th frame image, then the first relative position appearing in the Xth frame image is the landing point of the first target object.
10. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 9, characterized in that: Also includes: Fitting a first trajectory curve from the plurality of first relative positions according to a parabola rule; Obtaining the relative positions of multiple motion individual features contained in the image; A correspondence between the first target object and the first moving individual in the multiple moving individual features is established according to the first trajectory curve and the relative positions of the multiple moving individual features in the image, thereby obtaining a correspondence between the first throwing distance and the first moving individual.
11. The method for measuring the throwing distance of a target object based on machine vision according to claim 1, characterized in that: Also includes: Establishing the corresponding relationship between the multiple sub-item regional features and the multiple motion individual features contained in the image; Determine the sub-project area where the first target object falls; The corresponding relationship between the first throwing distance and the first sports individual is obtained according to the sub-item area where the landing point of the first target object is located and the first corresponding relationship.
12. The method for measuring the throwing distance of a target object based on machine vision as claimed in claim 1, characterized in that: “Obtaining the first positional relationship between the identification feature and the image” is specifically: Set a visual mark and save the actual position of the visual mark in advance; Capturing a first image, the first image containing identification features, the first image and the image being acquired through fixed-point acquisition by the same acquisition terminal; The positional relationship between the identification feature extracted from the first image and the first image can be called as the first positional relationship between the dot matrix identification feature and the image. Preferably, the identification is specifically a dot matrix composed of multiple identification points.