CN116625409A - Dynamic positioning performance evaluation method, device and system - Google Patents
Dynamic positioning performance evaluation method, device and system Download PDFInfo
- Publication number
- CN116625409A CN116625409A CN202310863675.8A CN202310863675A CN116625409A CN 116625409 A CN116625409 A CN 116625409A CN 202310863675 A CN202310863675 A CN 202310863675A CN 116625409 A CN116625409 A CN 116625409A
- Authority
- CN
- China
- Prior art keywords
- positioning
- value
- error
- camera
- dynamic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 131
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 68
- 230000007246 mechanism Effects 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000003062 neural network model Methods 0.000 claims abstract description 29
- 238000006073 displacement reaction Methods 0.000 claims abstract description 14
- 238000013210 evaluation model Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims description 47
- 230000033001 locomotion Effects 0.000 claims description 38
- 238000005259 measurement Methods 0.000 claims description 27
- 230000000007 visual effect Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 10
- 230000009471 action Effects 0.000 claims description 7
- 238000007621 cluster analysis Methods 0.000 claims description 7
- 230000005284 excitation Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 239000003550 marker Substances 0.000 claims description 3
- 101150048216 tag-131 gene Proteins 0.000 description 18
- 230000006870 function Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000014509 gene expression Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M11/00—Testing of optical apparatus; Testing structures by optical methods not otherwise provided for
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Manufacturing & Machinery (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Analytical Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application provides a dynamic positioning performance evaluation method, equipment and a system. The method comprises the following steps: setting a driving mechanism and a vision module and obtaining position information data of the vision module corresponding to a first time stamp; setting a calibration module, and acquiring an image of the calibration module under a preset condition through the vision module to obtain positioning information data corresponding to the image and a positioning time stamp; comparing the first time stamp with the positioning time stamp by adopting a dynamic displacement compensation algorithm to calculate and obtain various positioning errors; the weight of the preset parameter is analyzed by adopting a dynamic positioning performance evaluation algorithm to obtain a corresponding evaluation value, wherein the dynamic positioning performance evaluation algorithm is established based on an expert system and a gray clustering evaluation model; and designing and training a BP neural network model, and inputting the evaluation value into the trained BP neural network model to output a dynamic positioning performance evaluation scheme.
Description
Technical Field
The present application relates to the field of image processing technologies, and in particular, to a dynamic positioning performance evaluation method, device, and system.
Background
With the continuous development of artificial intelligence, the application of visual positioning technology is increasingly widespread, and the visual positioning technology can shoot a target image (such as a label image) in a visual field range through a camera and convert the image into target data for transmission by an image processing system. Visual positioning requires the effectiveness of positioning, which results in positioning speed and accuracy that pose a high challenge.
In a dynamic positioning scene compared with a static scene, the relative motion of the camera and the tag greatly increases the difficulty of the tag positioning result calibration and performance evaluation, and increases the difficulty of the camera and the tag specification in the type selection. Therefore, how to find an optimal dynamic positioning performance evaluation scheme in a dynamic scene to meet the requirement of visual positioning becomes a problem to be solved.
It should be appreciated that the description in this background section is only for aiding in the understanding of the disclosed aspects of the application and is not necessarily prior art prior to the filing date of this application.
Disclosure of Invention
In one aspect, the present application provides a dynamic positioning performance evaluation method, including: setting a driving mechanism and a visual module, wherein the driving mechanism is configured to control the movement of the visual module so as to obtain position information data of the visual module corresponding to a first time stamp through feedback information calculation of the driving mechanism; setting a calibration module, and acquiring an image of the calibration module under a preset condition through the vision module to obtain positioning information data corresponding to the image and a positioning time stamp, wherein the vision module and the calibration module move relatively; comparing the first time stamp with the positioning time stamp by adopting a dynamic displacement compensation algorithm to calculate and obtain various positioning errors; analyzing the weight of preset parameters by adopting a dynamic positioning performance evaluation algorithm to obtain corresponding evaluation values, wherein the preset parameters comprise a plurality of positioning errors, and the dynamic positioning performance evaluation algorithm is established based on an expert system and a gray clustering evaluation model; and designing and training a BP neural network model, and inputting the evaluation value into the trained BP neural network model to output a dynamic positioning performance evaluation scheme.
In one embodiment, the vision module comprises a camera; the driving mechanism comprises a motor and a driving belt, the motor drives the driving belt to move, and the camera is positioned on the driving belt and moves together with the driving belt; and the calibration module comprises a label, and the positioning error comprises a label positioning error, wherein the label and the camera move relatively in the process of the joint movement of the camera and the driving belt.
In one embodiment, the step of setting the preset condition includes: providing a plurality of said labels having different sizes; and configuring a shooting specification of the camera to different resolutions and/or different frame rates to acquire images of the tag of different sizes under the shooting specification.
In one embodiment, the plurality of positioning errors comprises: a first positioning error in a first direction, a second positioning error in a second direction, and a third positioning error in a third direction, wherein the first direction, the second direction, and the third direction are perpendicular to each other; and the step of obtaining the first positioning error comprises: obtaining the movement speed, the movement acceleration and the first timestamp of the camera at a first position through the feedback information of the driving mechanism, and calculating to obtain a first position parameter of the camera; inputting the first position parameter, the first timestamp, the movement speed and the movement acceleration; inputting the positioning time stamp, wherein the first position is the position with the minimum time deviation from the label corresponding to the positioning time stamp; and calculating to obtain a difference value between the positioning timestamp and the first timestamp, and calculating to obtain a second position parameter of the camera corresponding to the positioning timestamp, wherein the first positioning error is obtained based on the difference value between the second position parameter and positioning information data corresponding to the positioning timestamp.
In one embodiment, the step of obtaining the second positioning error and the third positioning error comprises: measuring to obtain a first positioning measurement value and a second positioning measurement value of the camera at the first position; obtaining a first positioning calculation value and a second positioning calculation value based on a self-recognition visual marker positioning algorithm; and comparing the first positioning calculation value with the first positioning measurement value to obtain the second positioning error, and comparing the second positioning calculation value with the second positioning measurement value to obtain the third positioning error, wherein the first positioning measurement value is a measurement value along the second direction, the second positioning measurement value is a measurement value along the third direction, the first positioning calculation value is a calculation value along the second direction, and the second positioning calculation value is a calculation value along the third direction.
In one embodiment, the step of analyzing the weights of the preset parameters to obtain the corresponding evaluation values further includes locating the calculated speed average value: analyzing the weight of the positioning calculation speed average value by adopting the dynamic positioning performance evaluation algorithm so as to obtain a corresponding evaluation value of the positioning calculation speed average value; and analyzing the weights of the positioning errors by adopting the dynamic positioning performance evaluation algorithm so as to obtain corresponding evaluation values of each positioning error.
In one embodiment, the first positioning error includes a first error mean, a first error standard deviation, a first maximum error value, a first minimum error value, a first error absolute value maximum, a first error absolute value minimum; the second positioning error comprises a second error mean value, a second error standard deviation, a second maximum error value, a second minimum error value, a second error absolute value maximum value and a second error absolute value minimum value; and the third positioning error comprises a third error mean value, a third error standard deviation, a third maximum error value, a third minimum error value, a third error absolute value maximum value and a third error absolute value minimum value.
In one embodiment, the dynamic positioning performance evaluation algorithm based on expert system and gray cluster evaluation model establishment includes: dividing the influence degree of the performance index of the preset parameter on the dynamic positioning performance evaluation scheme into a plurality of grades, and setting a corresponding score for each grade; determining a sample matrix of performance index scores of the experts in the expert system on the preset parameters; determining gray class and constructing a probability function, and calculating a gray evaluation matrix; and outputting a gray cluster analysis result.
In one embodiment, the BP neural network model includes an input layer, a hidden layer, and an output layer, and the steps of designing and training the BP neural network model include: setting an initial weight parameter and a critical value of the BP neural network model, wherein the initial weight parameter and the critical value are random values; inputting the preset parameters and the corresponding evaluation values thereof to calculate to obtain an intermediate output value; the intermediate output value is processed through an excitation function to obtain an output value of the hidden layer; the output value of the hidden layer is input to the output layer after being calculated and processed, and the calculated evaluation value is output after being processed by the excitation function so as to complete forward propagation; adopting a gradient descent method to adjust weights and thresholds of the input layer, the hidden layer and the output layer so as to finish back propagation; and reciprocally performing the forward propagation and the backward propagation until the error is less than a preset desired value.
In one embodiment, the camera captures more than one cycle of the label image under the preset condition, wherein the camera reciprocates once in the direction of the conveyor belt from an initial position and returns to the initial position by one cycle.
Another aspect of the present application provides a dynamic positioning performance evaluation apparatus, including: a calibration module; the visual module is used for acquiring the image of the calibration module under the preset condition to obtain positioning information data corresponding to the image and the positioning time stamp, wherein the visual module and the calibration module move relatively; the driving mechanism is configured to control the movement of the vision module so as to calculate position information data of the vision module corresponding to the first time stamp through feedback information of the driving mechanism; the first algorithm module is used for comparing the first timestamp with the positioning timestamp based on a dynamic displacement compensation algorithm so as to calculate and obtain various positioning errors; the second algorithm module is used for analyzing the weight of preset parameters based on a dynamic positioning performance evaluation algorithm to obtain corresponding evaluation values, wherein the preset parameters comprise a plurality of positioning errors, and the dynamic positioning performance evaluation algorithm is established based on an expert system and a gray clustering evaluation model; and the trained BP neural network model is used for processing the evaluation value and outputting a dynamic positioning performance evaluation scheme.
In one embodiment, the vision module comprises a camera; the driving mechanism comprises a motor and a driving belt, the motor drives the driving belt to move, and the camera is positioned on the driving belt and moves together with the driving belt; and the calibration module comprises a label, and the positioning error comprises a label positioning error, wherein the label and the camera move relatively in the process of the joint movement of the camera and the driving belt.
In one embodiment, the preset condition includes: the number of the labels is a plurality of, and the labels are respectively of different sizes; and the shooting specification of the camera is configured to be different in resolution and/or different in frame rate, wherein the camera acquires the image of the tag under the shooting specification.
In one embodiment, the preset parameters further include a location calculation speed average.
In still another aspect, the present application provides a dynamic positioning performance evaluation system, including: the dynamic positioning performance evaluation apparatus according to any one of the above; and the mechanical arm is used for completing the preset action, wherein the preset action is completed under the condition of the dynamic positioning performance evaluation scheme output by the dynamic positioning performance evaluation equipment.
The dynamic performance evaluation method provided by the application can have at least one of the following beneficial effects:
according to the method in some embodiments of the application, the position of the vision module at a certain moment in the movement process of the vision module can be accurately measured;
according to the method in some embodiments of the application, the position information data of the vision module and the positioning information data of the calibration module are accurately synchronized in time, so that measurement deviation caused by time dislocation of the vision module and the positioning information data is reduced;
According to the method in some embodiments of the application, the automatic calibration of the positioning of the calibration module under the preset condition is realized by adopting a dynamic displacement compensation algorithm, so that the workload of manual calibration is reduced; and
according to the method in some embodiments of the application, the specification schemes of the optimal vision module and the calibration module in the appointed dynamic scene can be rapidly output through the dynamic positioning performance evaluation algorithm, so that the workload of manual data analysis and scheme selection is greatly reduced.
Drawings
Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, taken in conjunction with the accompanying drawings. In the drawings:
FIG. 1 is a flowchart of a dynamic positioning performance evaluation method according to an exemplary embodiment of the present application;
FIG. 2 is a functional schematic of the modules in the dynamic positioning performance evaluation method according to an exemplary embodiment of the present application;
fig. 3 is a schematic structural view of a dynamic positioning performance evaluation apparatus according to an exemplary embodiment of the present application;
FIG. 4 is a schematic diagram of tag positioning according to an exemplary embodiment of the present application;
FIG. 5 is a schematic diagram of a BP neural network model according to an exemplary embodiment of the application; and
Fig. 6 is a dynamic performance evaluation flowchart of a dynamic positioning performance evaluation apparatus according to an exemplary embodiment of the present application.
Detailed Description
For a better understanding of the application, various aspects of the application will be described in more detail with reference to the accompanying drawings. It should be understood that the detailed description is merely illustrative of exemplary embodiments of the application and is not intended to limit the scope of the application in any way. Like reference numerals refer to like elements throughout the specification. The expression "and/or" includes any and all combinations of one or more of the associated listed items.
It should be noted that in this specification, the expressions "first", "second", "third", etc. are used only to distinguish one feature from another feature, and do not denote any limitation of the features, particularly do not denote any order of precedence. Thus, a first direction discussed in this disclosure may also be referred to as a second direction and vice versa without departing from the teachings of the present disclosure.
In the description, references to "one embodiment," "an example embodiment," "some embodiments," etc., indicate that the embodiment described may include a particular feature, structure, or characteristic, but every embodiment may not necessarily include the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Furthermore, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the relevant art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
In the drawings, the thickness, size, and shape of the components have been slightly adjusted for convenience of description. The figures are merely examples and are not drawn to scale. For example, the dimensions of the labels drawn in the figures in the present application are not to scale in actual production. As used herein, "about," "approximately," and similar terms are used as terms of a table approximation, not as terms of a table degree, and are intended to illustrate inherent deviations in measured or calculated values that would be recognized by one of ordinary skill in the art.
It should be understood that expressions such as "comprising," "including," "having," "containing," and/or "comprising" are open-ended, rather than closed-ended, which indicates the presence of stated features, elements and/or components, but does not preclude the presence or addition of one or more other features, elements, components and/or groups thereof. Furthermore, when a statement such as "at least one of the following" appears after a list of features listed, it modifies the entire list of features rather than just modifying the individual elements in the list. Furthermore, when describing embodiments of the application, use of "may" means "one or more embodiments of the application. Also, the term "exemplary" is intended to refer to an example or illustration.
It will also be understood that meanings such as "on", "over" and "over" should be interpreted in the broadest sense such that "on" means not only "directly on" something but also includes "on" and having an intermediate feature or layer therebetween, and "over" or "over" means not only "over" or "over" something but also may include "over" or "over" something and having no intermediate feature or layer therebetween (i.e., directly on something).
Unless otherwise defined, all terms (including engineering and technical terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application pertains. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. In addition, unless explicitly defined or contradicted by context, the particular steps included in the methods described herein need not be limited to the order described, but may be performed in any order or in parallel.
The features, principles, and other aspects of the present application are described in detail below.
The inventors of the present application have found that dynamic positioning of tags in dynamic scenes where the camera and tag are moving relative to each other is faced with a number of problems compared to static scenes. For example:
firstly, the accurate measurement of the position of a camera or a tag is difficult to implement by a manual mode under the condition of movement;
secondly, the time synchronization of the positioning of the camera to the tag and the camera position is difficult to ensure, and the dislocation of the camera position and positioning data in the time direction easily occurs, so that the positioning error is increased;
third, under specified dynamic scenes and specified camera motion speeds, system conditions may include different tag sizes, camera specifications to take pictures at different resolutions and/or frame rates, etc., which makes the measurement effort of tag positioning errors large;
fourth, in a dynamic scenario, the tag positioning error may fluctuate in different forms due to systematic errors, random errors, and the like; on the other hand, because the conditions of mutual restriction and mutual influence exist in various indexes of the positioning performance (including positioning calculation speed, positioning error mean value, positioning error fluctuation and the like), different emphasis requirements are placed on various indexes of the positioning performance under different scenes.
The application provides a dynamic positioning performance evaluation method, which can find out an optimal dynamic positioning performance evaluation scheme under a dynamic scene to meet the requirement of visual positioning, thereby at least partially improving or solving the problems.
Fig. 1 is a flowchart of a dynamic positioning performance evaluation method 1000 according to an embodiment of the present application. As shown in fig. 1, the present application provides a dynamic positioning performance evaluation method 1000, which includes:
step S1100, setting a driving mechanism and a vision module, wherein the driving mechanism is configured to control the movement of the vision module so as to obtain position information data of the vision module corresponding to the first timestamp through feedback information calculation of the driving mechanism;
step 1200, setting a calibration module, and acquiring an image of the calibration module under a preset condition through a vision module to obtain positioning information data corresponding to the image and a positioning timestamp, wherein the vision module and the calibration module move relatively;
step S1300, comparing the first time stamp with the positioning time stamp by adopting a dynamic displacement compensation algorithm to calculate and obtain various positioning errors;
step S1400, analyzing weights of preset parameters to obtain corresponding evaluation values by adopting a dynamic positioning performance evaluation algorithm, wherein the preset parameters comprise various positioning errors, and the dynamic positioning performance evaluation algorithm is built based on an expert system and a gray clustering evaluation model; and
Step S1500, designing and training a BP neural network model, and inputting the evaluation value into the trained BP neural network model to output a dynamic positioning performance evaluation scheme.
It should be understood that the steps shown in method 1000 are not exclusive and that other steps may be performed before, after, or between any of the steps shown. Further, some of the illustrated steps may be performed concurrently or may be performed in a different order than that shown in fig. 1. The above steps S1100 to S1500 are further described below in conjunction with fig. 1 to 5.
Step S1100, setting a driving mechanism and a vision module, wherein the driving mechanism is configured to control the movement of the vision module
And (3) calculating and obtaining the position information data of the vision module corresponding to the first time stamp through the feedback information of the driving mechanism.
Fig. 2 is a functional schematic diagram of each module in the dynamic positioning performance evaluation method according to an exemplary embodiment of the present application, and in combination with fig. 2 and 3, in step S1100, a driving mechanism 110 and a vision module 120 may be provided, and the driving mechanism 110 is configured to control the movement of the vision module 120. In some embodiments, an upper computer 140 may be further provided, the driving mechanism 110 may feed back its own working information to the upper computer 140, and perform information processing through the upper computer 140 to obtain the vision module 120 and the first timestamp t1 corresponding to the position information dataP1, thereby being beneficial to realizing accurate measurement of the position information of the vision module 120 in the dynamic scene;
the first time stamptReference numeral 1 denotes unique data corresponding to the first time, which is generated by the host computer 140 based on the operation information fed back by the driving mechanism 110.
Fig. 3 is a schematic structural view of a dynamic positioning performance evaluation apparatus according to an exemplary embodiment of the present application. As shown in fig. 3, the driving mechanism 110 may include a motor 111 and a driving belt 112, and the vision module 120 may include a camera 121. The motor 111 is used for driving the motion of the driving belt 112, and the camera 121 is disposed on the driving belt 112. Illustratively, the camera 121 is fixed to the conveyor belt and remains co-moving with the conveyor belt 112 as it moves.
In some embodiments, the motor 111 begins to operate upon receipt of a start command, driving the belt 112 along, for exampleyAnd the direction is driven, and the conveyor belt 112 drives the cameras 121 to move together. The motor 111 can feed back its own working information to the host computer 140, and calculate and obtain the camera 121 and the first timestamp via the host computer 140t1 corresponding to the position information dataP1. Illustratively, the motor 111 comprises a stepper motor and the conveyor 112 comprises a belt.
Step S1200, setting a calibration module, and collecting images of the calibration module under preset conditions through a vision module to obtain the image
And obtaining positioning information data corresponding to the image and the positioning time stamp, wherein the vision module and the calibration module move relatively.
With continued reference to fig. 2 and 3, the calibration module 130 may be provided, and an image of the calibration module 130 may be collected by the vision module 120, and the vision module 120 may transmit the collected image information to the upper computer 140, and may obtain the image and the positioning timestampt2 corresponding positioning information dataP2. It should be noted that the positioning time stampt2 is unique data corresponding to the second time generated by the upper computer 140 based on the operation information of the vision module 110. Illustratively, calibration module 130 includes a label 131, label 131 being securable to substrate 160. The vision module 120 includes a camera 120, and the tag 131 and the camera 121 move relatively during the common movement of the camera 121 with the conveyor belt 112.
In a dynamic scenario, when the camera 121 needs to perform image acquisition at a first time, setting of its own acquisition mode (for example, acquiring images at a fixed frame rate) may cause the time that the camera 121 actually records the images to be delayed or advanced to a second time, thereby causing misalignment of the position and positioning data of the camera 121 in the time direction, and increasing the positioning error.
In some embodiments, preset conditions may be set and images of the tag 131 are acquired using the camera 121 under the preset conditions. The step of setting the preset condition may include: first, the number of the tags 131 is set to be plural, and the sizes of different tags 131 may be set to be different; second, the photographing specifications of the camera 121 are configured to different resolutions and/or different frame rates to capture images of the different-sized tags 131 under the different photographing specifications.
The camera 121 starts from the initial position and followsyDirection movement to the designated position and then toyThe reverse direction movement and return to the initial position, and the camera 121 reciprocates once from the initial position and returns to the initial position again, which is denoted as one round. For example, the camera 121 may capture images of the tag 131 corresponding to a plurality of time stamps during one rotation, and position the time stamp among the plurality of time stampst2 corresponds to a first image among the images of the photographed tag 131. The upper computer 140 processes the first image and obtains corresponding positioning information dataP2, positioning information dataP2 and location time stampt2 corresponds to.
Illustratively, the tag 131 includes a rimyThe first, second and third tags 1311, 1312 and 1313 are sequentially arranged in the direction. The resolution specification of the camera 121 at the time of shooting includes at least one of 1920×1080, 1280×720, and 640×480, and the frame rate specification includes at least one of 15fps, 30fps, and 60 fps. The camera 121 may take at least one of the first, second, and third tags 1311, 1312, and 1313 for multiple rounds to acquire multiple sets of data.
In some rounds, the shooting specification of the camera 121 is set to 1920×1080, for example, the frame rate is 30fps, and the camera 121 captures an image of the first tag 1311 under the specification, the first tag 1311 is square, for example, and the side length thereof is 0.05m, for example; in other rounds, the shooting specification of the camera 121 is set to 640×480, for example, and the frame rate is 60fps, and the camera 121 captures an image of the second tag 1312 under the specification, and the second tag 1312 is square, for example, and the side length thereof is 0.04m, for example; in still other rounds, the shooting specification of the camera 121 is set to, for example, 1280×720 in resolution and 60fps in frame rate, and the camera 121 captures an image of the first tag 1311 under the specification, the first tag 1311 being, for example, a square, and the side length thereof being, for example, 0.05m.
It should be noted that the description of the resolution, the frame rate, and the sizes of the first tag 1311 and the second tag 1312 in the shooting specification of the camera 121 in the context of the present application is merely illustrative, and not a limitation of the combination of the two, and those skilled in the art may select other suitable specifications for free combination according to different application conditions and different schemes.
Step S1300, comparing the first time stamp with the positioning time stamp by using dynamic displacement compensation algorithm to calculate
A variety of positioning errors are obtained.
With continued reference to FIG. 2, the upper computer 140 may include a first algorithm module 141, the first algorithm module 141 being built based on a dynamic displacement compensation algorithm, the first algorithm module 141 stamping a first time stampt1 and location time stamptAnd 2, outputting various positioning errors after the comparison processing.
In some embodiments, the plurality of positioning errors includes a first positioning error in a first direction, a second positioning error in a second direction, and a third positioning error in a third direction, the first direction, the second direction, and the third direction being perpendicular to each other. Illustratively, the first direction is opposite toxThe direction is parallel, the second directionyDirection is parallel, the third directionzThe directions are parallel.
Fig. 4 is a schematic diagram of tag positioning according to an exemplary embodiment of the present application. As shown in fig. 4, in the working state, the driving mechanism 110 feeds back its own working information to the upper computer 140, and obtains the movement speed of the camera 121 at the first position a after processing by the upper computer 140vAcceleration of movementaAnd a first timestampt1, and obtain a first positional parameter of the camera 121 P1. First position A is atyThe direction is at a distance from the tag 131 ofy1, inzThe distance from the tag 131 in the direction isz1. First position parameterP1. First time stampt1. Speed of movementvAnd motion accelerationaIs input to the firstAn algorithm module 141 and is processed by a dynamic displacement compensation algorithm. Furthermore, a time stamp is locatedt2 is also input to the algorithm module 141.
In some embodiments, the dynamic displacement compensation algorithm includes searching for a timestamp corresponding to the position of the camera 121 closest to the positioning timestamp t2, where the timestamp is the first timestamp t1. In other words, the first position a is the position with the smallest time deviation from the tag 131 corresponding to the positioning time stamp t 2. The difference between the positioning timestamp t2 and the first timestamp t1 is Δt, which satisfies Δt=t2-t 1. From the motion velocity v and the difference Δt corresponding to the first position a, a second position parameter P2 of the camera 121 corresponding to the positioning time stamp t2 can be calculated, which satisfies p2=p1+vΔt+1/2aΔt 2 . The first positioning error (including the positioning error in the x-direction) is calculated based on the difference between the second position parameter P2 and the positioning information data of the first image.
In some embodiments, a second positioning error is obtained (including along yDirectional positioning errors) and a third positioning error (including alongzDirectional positioning error) includes measuring a first positioning measurement and a second positioning measurement of the camera 121 at the first position a; obtaining a first positioning calculation value and a second positioning calculation value based on a self-recognition visual marker positioning algorithm; and comparing the first positioning calculation value with the first positioning measurement value to obtain a second positioning error, and comparing the second positioning calculation value with the second positioning measurement value to obtain the third positioning error. Illustratively, the first positioning measurement is in the second directionyDirection), the second positioning measurement value is a measurement value along the third directionzDirection), the first positioning calculation being a calculation in the second direction, the second positioning calculation being a calculation in the third direction.
The application can acquire images of the labels 131 with different sizes under different resolutions and frame rates by multiple times, and calculate and obtain various positioning errors of the labels 131 by adopting a dynamic displacement compensation algorithm, can combine parameters of different shooting specifications of the camera 121, different sizes of the labels 131 and the like, and automatically process the parameters by the first algorithm module 141, thereby optimizing the measuring workload and the measuring efficiency with larger positioning errors of the labels.
Step S1400, analyzing the weights of the preset parameters by adopting a dynamic positioning performance evaluation algorithm to obtain the corresponding values
Evaluating values, wherein the preset parameters comprise various positioning errors, and the dynamic positioning performance evaluation algorithm is based on an expert system and gray
And (5) establishing a clustering evaluation model.
With continued reference to fig. 2, the upper computer 140 may further include a second algorithm module 142, where the second algorithm module 142 is established based on a dynamic performance evaluation algorithm, and the second algorithm module 141 processes and analyzes weights of preset parameters. Illustratively, the preset parameters include a variety of positioning errors, including, for example, a first positioning error, a second positioning error, and a third positioning error. Further, the first positioning error comprises a first error mean value, a first error standard deviation, a first maximum error value, a first minimum error value, a first error absolute value maximum value and a first error absolute value minimum value;
the second positioning error comprises a second error mean value, a second error standard deviation, a second maximum error value, a second minimum error value, a second error absolute value maximum value and a second error absolute value minimum value; and the third positioning error comprises a third error mean value, a third error standard deviation, a third maximum error value, a third minimum error value, a third error absolute value maximum value and a third error absolute value minimum value. The weights of the various positioning errors can be analyzed by adopting a dynamic positioning performance evaluation algorithm, so that corresponding evaluation values of each positioning error are obtained.
It can be appreciated that the first error mean is the label positioning atxThe second error mean value is the error mean value of the direction and is that the label is positioned atyThe third error mean value is the error mean value of the direction and is that the label is positioned atzError mean value of direction;
the first error standard deviation is that the label is positioned atxThe standard deviation of the error of the direction and the second standard deviation of the error are that the label is positioned atyThe standard deviation of the error of the direction and the third standard deviation of the error are that the label is positioned atzStandard deviation of error in direction;
the first maximum error value is that the label is positioned atxThe maximum error value of the direction and the second maximum error value are that the label is positioned atyThe maximum error value of the direction and the third maximum error value are that the label is positioned atzThe error maximum of the direction;
the first minimum error value is that the label is positioned atxThe maximum error value and the second minimum error value of the direction are that the label is positioned atyThe maximum error value and the third minimum error value of the direction are that the label is positioned atzThe error minimum of the direction;
the maximum value of the absolute value of the first error is that the label is positioned atxThe maximum value of the absolute value of the error of the direction and the maximum value of the absolute value of the second error are that the label is positioned atzThe maximum value of the absolute value of the error of the direction and the maximum value of the absolute value of the third error are that the label is positioned at zThe absolute value of the error of the direction is maximum;
the first error absolute value minimum value is that the label is positioned atxThe maximum value of the absolute value of the error of the direction and the minimum value of the absolute value of the second error are that the label is positioned atzThe maximum value of the absolute value of the error of the direction and the minimum value of the absolute value of the third error are that the label is positioned atzThe absolute value of the error of the direction is maximum.
In some embodiments, the preset parameters further include a location calculation speed average of the tag 131. The analysis of the weight of the preset parameter further comprises the steps of adopting the dynamic positioning performance evaluation algorithm to analyze the weight of the average value of the positioning calculation speed, and obtaining the corresponding evaluation value of the average value of the positioning calculation speed.
The dynamic positioning performance evaluation algorithm is built based on an expert system and a gray clustering evaluation model. The influence degree of the performance index of the preset parameter on the dynamic positioning performance evaluation scheme can be divided into a plurality of grades, and corresponding scores are set for each grade; determining a sample matrix and gray class of performance index scores of experts in the expert system on preset parameters, constructing a probability function, calculating a gray evaluation matrix, and finally outputting a gray cluster analysis result. The calculation flow comprises the steps of inputting preset parameters and inputting the resolution, frame rate and label size data information of a camera. The weight of the positioning performance index can be analyzed through an expert system and a gray clustering evaluation model, and an analysis result is obtained.
Illustratively, performance metrics are presentedR ij Is classified into 5 classes, namely: very low, medium, high, very high. By [0,1 ]]The values between which score the corresponding classes to represent the magnitude of their affected extent. Table 1 shows the scores corresponding to the impact levels (impact levels between two levels, expressed as values between the two levels, which may be below 0.1 or above 0.9, with minimum not below 0 and maximum not exceeding 1).
TABLE 1
Can pass throughwExpert-to-performance indexR ij (i=1,2...N,j=1,2...Q) The evaluation was carried out, wherein,jrepresent the firstiThe class performance index includes the number of factors. First, themExpertm=1,2...w) For performance indexR ij The evaluation observations of (2) are respectivelyd ijm ThenmThe sample matrix D formed by the bit expert is:
d 111 d 112 ... d m11 ... d w11
d 121 d 122 ... d m12 ... d w12
. . . .
. . . .
. . . .
D = d ij1 d ij2 ... d ijm ... d ijw
. . . .
. . . .
. . . .
d NQ1 d NQ2 ... d NQm ... d NQw
due to expertJudging that subjective randomness exists, comparing the data base of the judging matrix with each other in pairs, wherein the data base is possibly inconsistent or missing, and an expert is difficult to determine each performance indexR ij Indirect influence relation between the two, so that the application can obtain the performance indexR ij The relationship and gray cluster analysis model set correct the indirect relationship between the factors which are difficult to be determined by the expert, and reduce the error caused by subjective judgment of the expert to a certain extent.
In addition, because the gray cluster analysis has low requirements on samples, no special distribution rule of sample data is needed, the actual situation can be well reflected in a few or more than ten small sample ranges, the calculation is relatively uncomplicated, the gray weighting cluster analysis method can be adopted to firstly determine the gray construction probability function, calculate the gray evaluation coefficient, determine the gray evaluation matrix and calculate the performance index R ij Corresponding weight is used as a fixed weight coefficient of each gray class in the dynamic positioning process, and the performance index can be realized by the composite operation of the evaluation matrix and the fixed weight coefficientR ij And converting subjective language evaluation into objective digitization, so that risks which cannot be quantized subjectively are quantized into comparable objective values through fuzzy languages, and dynamically positioned evaluation quantification is realized.
Table 2 shows scoring results obtained by the expert system based on the influence of the tag locating performance parameters.
TABLE 2
It should be noted that the number of experts in the expert systemwThere may be a plurality, expert 1 in Table 2 is merely illustrative, in an embodiment number of expertswOther values are possible, for examplew=5、w=10、w=20, etc., to which the present application is not limited.
Table 3 shows the basisw=10, the expert system scores the analysis results of the gray cluster analysis model of the results.
TABLE 3 Table 3
In table 3, the position calculation speed average has an important role in dynamic positioning, which has a great influence on the timeliness of the motion planning based on the tag positioning. It should be noted that the secondary index influence weight is a normalized weight.
Step S1500, designing and training a BP neural network model, and inputting the evaluation value into the trained BP neural network model to output a dynamic positioning performance evaluation scheme.
Fig. 5 is a schematic diagram of a BP neural network model according to an exemplary embodiment of the present application. As shown in fig. 5, a BP neural network model 150 may be designed, which includes an input layer 151, a hidden layer 152, and an output layer 153. The BP neural network model 150 is designed to improve the objectivity in the comprehensive evaluation of error performance.
In some embodiments, the input layer 151 of the BP neural network model 150 is designed to contain 19 neurons (I1 to I19) according to the localization scene and the error model, and its input parameters are, for example, performance parameters (including localization calculation speed average;y、x、zpositioning error mean values in three directions;y、x、zstandard deviation of positioning errors in three directions;y、x、zpositioning maximum error values in three directions;y、x、zpositioning minimum error values in three directions;y、x、zthe absolute value of the positioning error in three directions is maximum;y、x、zabsolute minimum of positioning errors in three directions). The number of layers of the hidden layer 152 is 1, which is set to include 12 neurons (H1 to H12), and the output layer 153 includes 1 neuron for outputting the final evaluation value.
For example, the tag locating performance result samples and their evaluation quantized values may be input into the BP neural network model 150 for training, as follows:
Setting initial weight of BP neural networkωAnd critical valueθ. Initial weightωAnd critical valueθAre all set to smaller values, which can be achieved byThe random generator program generates a set of random numbers between-0.5 and 0.5 as initial weightsωAnd critical valueθ ;
Inputting preset parameters and calculating corresponding evaluation values to obtain intermediate output valuesO j1 Which satisfies the following requirementsO j1 =ΣW ij *y i - q j . Wherein,,y i is the first of the 19 performance parametersiThe number of the two-dimensional space-saving type,W ij is the first of the input layer 151iThe first neuron and hidden layer 152jThe weight between the individual neurons is such that,q j is a bias term. Illustratively, a random generator program may be used to generate a set of random numbers between-0.001 and 0.001 as bias termsq j ;
The hidden layer 152 is processed by the excitation function sigmod as an output value of the hidden layer 152H j . Illustratively, the excitation function sigmod isf(y)=1/(1+e y- ),H j Satisfy the following requirementsH j =f(O j1 )=1/(1+e O j-1 )。
The output value of the hidden layer 152H j The weighted sum is input to the output layer 153 again, and the actual output value is obtained after processing by the excitation function sigmodO k To complete forward propagation, actual output valueO k Namely, calculating an evaluation value;
the weights and thresholds of the input layer 151, hidden layer 152, output layer 153 are adjusted using a gradient descent method to accomplish back propagation. Illustratively, the counter-propagating value of the error uses the output value O k And the actual evaluation value of the sampleT k Error square sum of (2)E p Which satisfies 2E p =Σ(T k -O k ) 2 . And if the error value is larger than the preset expected value, performing a back propagation process of the error. The weight and threshold of each layer are adjusted by gradient descent method. Reciprocating forward and backwardThe back propagation and the back propagation are performed until the error is less than a preset expected value, for example, 0.001, to complete training of the BP neural network model 150.
The input value and the evaluation output value of the tag positioning performance evaluation sample are processed through the trained BP neural network model 150, so that an optimal dynamic performance evaluation scheme can be obtained, and the problem that the shooting specification of the camera 121, the size of the tag 131 and the like are difficult to select due to mutual restriction and mutual influence of various indexes (including positioning calculation speed, positioning error mean value, positioning error fluctuation and the like) of the positioning performance under different scenes is solved.
Still another aspect of the present application provides a dynamic positioning performance evaluation apparatus 100, with continued reference to fig. 2 and 3, the dynamic positioning performance evaluation apparatus 100 includes a driving mechanism 110, a vision module 120, a calibration module 130, a first algorithm module 141, a second algorithm module 142, and a trained BP neural network model 150. Illustratively, the driving mechanism 110 may control the movement of the vision module 120 and may feed back its own working information to be calculated to obtain the vision module 120 and the first timestamp t1 corresponding to the position information dataP1. The vision module 120 is used for acquiring the image of the calibration module 130 under the preset condition to obtain the image and the positioning time stampt2 corresponding positioning information dataP2. The calibration module 130 may be configured to be fixed such that the vision module 120 moves relative to the calibration module 130 when the vision module 120 is driven by the driving mechanism 110.
In some implementations, the first algorithm module 141 is built based on a dynamic displacement compensation algorithm, which may be established by a first timestampt1 and second timestampt2 to calculate and obtain various positioning errors.
In some embodiments, the second algorithm module 142 analyzes the weights of the preset parameters based on a dynamic positioning performance evaluation algorithm to obtain corresponding evaluation values. Illustratively, the preset parameters may include the various alignment errors described above, and may also include the position calculation speed average of the calibration module 130. It should be noted that the dynamic positioning performance evaluation algorithm is built based on an expert system and a gray clustering evaluation model, the expert system can be utilized to score the weight of each performance index, the gray clustering evaluation model is adopted to analyze based on the scoring condition, the multi-level index weight is obtained, and the weight evaluation values of various performance indexes of preset parameters can be effectively obtained in different scenes.
In some embodiments, the trained BP neural network model 150 may be used to process the weight evaluation values of the various performance indicators obtained based on the second algorithm module 142 and ultimately output a dynamic positioning performance evaluation scheme, which is the optimal choice scheme.
In some embodiments, vision module 120 includes a camera 121, drive mechanism 110 includes a motor 111 and a belt 112, motor 111 drives belt 112 in motion, camera 111 is positioned on belt 112 and moves with belt 112, calibration module 130 includes a label 131, and the positioning error includes a label positioning error. It will be appreciated that the camera 121 moves relative to the tag 131 and the camera 111 during movement in concert with the belt 112.
In some embodiments, the preset conditions include that the number of tags 131 is plural and each has a different size; the photographing specification of the camera 121, under which the camera 121 captures the image of the tag 131, is configured to different resolutions and/or different frame rates. Illustratively, the resolution specification of the camera 121 at the time of photographing includes at least one of 1920×1080, 1280×720, and 640×480, and the frame rate specification includes at least one of 15fps, 30fps, and 60 fps. During operation of the dynamic positioning performance evaluation apparatus 100, the camera 121 may take at least one of the first, second, and third tags 1311, 1312, and 1313 in a plurality of passes, wherein the camera 131 reciprocates once from the initial position and returns to the initial position again, denoted as one pass.
In some embodiments, the dynamic positioning performance evaluation apparatus 100 further includes a host computer 140, and the host computer 140 includes a first algorithm module 141, a second algorithm module 142, and a third algorithm module 143. The third algorithm module 143 is built based on the BP neural network model and can be used for positioning performance evaluation in dynamic scenarios.
Fig. 6 is a dynamic performance evaluation flowchart of a dynamic positioning performance evaluation apparatus according to an exemplary embodiment of the present application. The dynamic performance evaluation flow of the dynamic positioning performance evaluation apparatus 100 is described below with reference to fig. 3 and 6.
The dynamic positioning performance evaluation device 100 may include a control layer 210, an algorithm layer 220, and an analysis layer 230. Calibration control software in, for example, a calibration device (not shown) may be turned on to establish a connection of the calibration control software to the motor 111, camera 121. The calibration control software reads the configuration file, obtains calibration configuration data and issues movement instructions to the motor 111. The motor 111 drives the camera 121 along the belt 112yDirection movement, and acquires position data, movement speed, movement acceleration and other data of the camera 121 which is driven to return by the motor 111 in real time.
Illustratively, the calibration control software controls the camera 121 to take images at regular time and process each target data in the images in real time, such as calculating the positioning information data of the tag 131 and counting the time elapsed for positioning calculation. The position information data of the camera 121 corresponding to the currently processed image frame can be obtained by a dynamic displacement compensation algorithm, and the position information data of the camera 121 and the corresponding position information data of the camera 121 can be calculated according to the positioning information data of the tag 131 and the corresponding position information data of the camera 121, respectively, to obtain the positions of the camera 121 and the tag 131 along the edge y、x、zAnd calculating the positioning errors of the positioning information data of the tag 131 relative to the actual measured value.
Illustratively, the positioning error may include a first positioning error (edgeyDirectional positioning error), second positioning error (alongxDirectional positioning error) and a third positioning error (alongzDirectional positioning errors). Further, the first positioning error comprises a first error mean value, a first error standard deviation, a first maximum error value, a first minimum error value, a first error absolute value maximum value and a first error absolute value minimum value; the second positioning error comprises a second error mean value, a second error standard deviation, a second maximum error value, a second minimum error value, a second error absolute value maximum value and a second error absolute value minimum value; the third positioning error comprises a third error mean value and a third error standard deviationA third maximum error value, a third minimum error value, a third absolute error value maximum value, and a third absolute error value minimum value.
When the camera 121 reaches the specified position, the calibration control software controls the motor 111 to stop moving, and at the same time controls the camera 121 to stop shooting images. The calibration control software may then control the camera 121 to move back to the initial position. In the moving process, if the calibration device determines that the camera 121 does not return to the initial position according to the information returned by the motor 111, the calibration control software sends a control instruction to control the motor 111 to drive the camera 121 to return to the initial position, so as to complete the calibration operation of one round. It should be noted that, the calibration control software may control the motor 111 to drive the camera 121 to perform calibration operation for one or more rounds according to the calibration configuration data for different shooting specifications (including resolution and frame rate) of the camera 121.
For example, after calibration is completed, the calibration control software may invoke a dynamic positioning performance evaluation algorithm to evaluate the positioning calculation speed and positioning error data of the camera 121 under the conditions of different resolutions, frame rates, different sizes of the tag 131, and the like, and output an optimal solution meeting the performance requirement.
Still another aspect of the present application provides a dynamic positioning performance evaluation system (not shown). The dynamic positioning performance evaluation system may include a dynamic positioning performance evaluation apparatus 100 and a robot arm (not shown) that may complete a predetermined action in response to an input action instruction. With continued reference to fig. 2 and 3, an optimal combination scheme of the resolution, frame rate, and size of the tag 131 of the camera 121 may be selected according to the dynamic positioning performance evaluation scheme output from the dynamic positioning performance evaluation apparatus 100, and the robot arm may complete its predetermined action under the optimal combination scheme condition.
The above description is only illustrative of the embodiments of the application and of the technical principles applied. It will be appreciated by those skilled in the art that the scope of the application is not limited to the specific combination of the above technical features, but also encompasses other technical solutions which may be formed by any combination of the above technical features or their equivalents without departing from the technical concept. Such as the above-mentioned features and the technical features disclosed in the present application (but not limited to) having similar functions are replaced with each other.
Claims (15)
1. A dynamic positioning performance evaluation method, characterized in that the method comprises:
setting a driving mechanism and a visual module, wherein the driving mechanism is configured to control the movement of the visual module so as to obtain position information data of the visual module corresponding to a first time stamp through feedback information calculation of the driving mechanism;
setting a calibration module, and acquiring an image of the calibration module under a preset condition through the vision module to obtain positioning information data corresponding to the image and a positioning time stamp, wherein the vision module and the calibration module move relatively;
comparing the first time stamp with the positioning time stamp by adopting a dynamic displacement compensation algorithm to calculate and obtain various positioning errors;
analyzing the weight of preset parameters by adopting a dynamic positioning performance evaluation algorithm to obtain corresponding evaluation values, wherein the preset parameters comprise a plurality of positioning errors, and the dynamic positioning performance evaluation algorithm is established based on an expert system and a gray clustering evaluation model; and
and designing and training a BP neural network model, and inputting the evaluation value into the trained BP neural network model to output a dynamic positioning performance evaluation scheme.
2. The method of claim 1, wherein the vision module comprises a camera;
the driving mechanism comprises a motor and a driving belt, the motor drives the driving belt to move, and the camera is positioned on the driving belt and moves together with the driving belt; and
the calibration module comprises a label, and the positioning error comprises a label positioning error, wherein the label and the camera move relatively in the process of the joint movement of the camera and the driving belt.
3. The method of claim 2, wherein the step of setting the preset condition comprises:
providing a plurality of said labels having different sizes; and
the shooting specifications of the camera are configured to different resolutions and/or different frame rates to acquire images of the tag of different sizes under the shooting specifications.
4. The method of claim 2, wherein the plurality of positioning errors comprises:
a first positioning error in a first direction, a second positioning error in a second direction, and a third positioning error in a third direction, wherein the first direction, the second direction, and the third direction are perpendicular to each other; and
The step of obtaining the first positioning error comprises:
obtaining the movement speed, the movement acceleration and the first timestamp of the camera at a first position through the feedback information of the driving mechanism, and calculating to obtain a first position parameter of the camera;
inputting the first position parameter, the first timestamp, the movement speed and the movement acceleration;
inputting the positioning time stamp, wherein the first position is the position with the minimum time deviation from the label corresponding to the positioning time stamp; and
calculating to obtain the difference between the positioning time stamp and the first time stamp, calculating to obtain the second position parameter of the camera corresponding to the positioning time stamp,
the first positioning error is obtained based on the difference value calculation of the second position parameter and the positioning information data corresponding to the positioning time stamp.
5. The method of claim 4, wherein the step of obtaining the second positioning error and the third positioning error comprises:
measuring to obtain a first positioning measurement value and a second positioning measurement value of the camera at the first position;
obtaining a first positioning calculation value and a second positioning calculation value based on a self-recognition visual marker positioning algorithm; and
Comparing the first positioning calculation value with the first positioning measurement value to obtain the second positioning error, comparing the second positioning calculation value with the second positioning measurement value to obtain the third positioning error,
the first positioning measurement value is a measurement value along the second direction, the second positioning measurement value is a measurement value along the third direction, the first positioning calculation value is a calculation value along the second direction, and the second positioning calculation value is a calculation value along the third direction.
6. The method of claim 2, wherein the preset parameters further comprise a location calculation speed average, and the step of analyzing weights of the preset parameters to obtain corresponding evaluation values comprises:
analyzing the weight of the positioning calculation speed average value by adopting the dynamic positioning performance evaluation algorithm so as to obtain a corresponding evaluation value of the positioning calculation speed average value; and
and analyzing the weights of a plurality of positioning errors by adopting the dynamic positioning performance evaluation algorithm so as to obtain corresponding evaluation values of each positioning error.
7. The method of claim 4, wherein the first positioning error comprises a first error mean, a first error standard deviation, a first maximum error value, a first minimum error value, a first error absolute value maximum, a first error absolute value minimum;
The second positioning error comprises a second error mean value, a second error standard deviation, a second maximum error value, a second minimum error value, a second error absolute value maximum value and a second error absolute value minimum value; and
the third positioning error comprises a third error mean value, a third error standard deviation, a third maximum error value, a third minimum error value, a third error absolute value maximum value and a third error absolute value minimum value.
8. The method of claim 1 or 6, wherein the dynamic positioning performance evaluation algorithm based on the expert system and the gray cluster evaluation model build comprises:
dividing the influence degree of the performance index of the preset parameter on the dynamic positioning performance evaluation scheme into a plurality of grades, and setting a corresponding score for each grade;
determining a sample matrix of performance index scores of the experts in the expert system on the preset parameters;
determining gray class and constructing a probability function, and calculating a gray evaluation matrix; and
and outputting a gray cluster analysis result.
9. The method of claim 1 or 6, wherein the BP neural network model comprises an input layer, a hidden layer, and an output layer, and wherein the step of designing and training the BP neural network model comprises:
Setting an initial weight parameter and a critical value of the BP neural network model, wherein the initial weight parameter and the critical value are random values;
inputting the preset parameters and the corresponding evaluation values thereof to calculate to obtain an intermediate output value;
the intermediate output value is processed through an excitation function to obtain an output value of the hidden layer;
the output value of the hidden layer is input to the output layer after being calculated and processed, and the calculated evaluation value is output after being processed by the excitation function so as to complete forward propagation;
adopting a gradient descent method to adjust weights and thresholds of the input layer, the hidden layer and the output layer so as to finish back propagation; and
and reciprocally performing the forward propagation and the backward propagation until the error is less than a preset expected value.
10. The method of claim 3, wherein the camera captures the label image more than one pass under the predetermined condition,
the camera moves back and forth once along the direction of the driving belt from the initial position and returns to the initial position for one round.
11. A dynamic positioning performance evaluation apparatus, characterized in that the apparatus comprises:
A calibration module;
the visual module is used for acquiring the image of the calibration module under the preset condition to obtain positioning information data corresponding to the image and the positioning time stamp, wherein the visual module and the calibration module move relatively;
the driving mechanism is configured to control the movement of the vision module so as to calculate position information data of the vision module corresponding to the first time stamp through feedback information of the driving mechanism;
the first algorithm module is used for comparing the first timestamp with the positioning timestamp based on a dynamic displacement compensation algorithm so as to calculate and obtain various positioning errors;
the second algorithm module is used for analyzing the weight of preset parameters based on a dynamic positioning performance evaluation algorithm to obtain corresponding evaluation values, wherein the preset parameters comprise a plurality of positioning errors, and the dynamic positioning performance evaluation algorithm is established based on an expert system and a gray clustering evaluation model; and
and the trained BP neural network model is used for processing the evaluation value and outputting a dynamic positioning performance evaluation scheme.
12. The apparatus of claim 11, wherein the vision module comprises a camera;
The driving mechanism comprises a motor and a driving belt, the motor drives the driving belt to move, and the camera is positioned on the driving belt and moves together with the driving belt; and
the calibration module comprises a label, and the positioning error comprises a label positioning error, wherein the label and the camera move relatively in the process of the joint movement of the camera and the driving belt.
13. The apparatus of claim 12, wherein the preset conditions include:
the number of the labels is a plurality of, and the labels are respectively of different sizes; and
the shooting specifications of the cameras are configured to be of different resolutions and/or different frame rates, wherein the cameras acquire images of the tag under the shooting specifications.
14. The apparatus of claim 11, wherein the predetermined parameters further comprise a position calculation speed average.
15. A dynamic positioning performance evaluation system, the system comprising:
the dynamic positioning performance evaluation apparatus according to any one of claims 11 to 14; and
and the mechanical arm is used for completing the preset action, wherein the preset action is completed under the condition of the dynamic positioning performance evaluation scheme output by the dynamic positioning performance evaluation equipment.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310863675.8A CN116625409B (en) | 2023-07-14 | 2023-07-14 | Dynamic positioning performance evaluation method, device and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310863675.8A CN116625409B (en) | 2023-07-14 | 2023-07-14 | Dynamic positioning performance evaluation method, device and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116625409A true CN116625409A (en) | 2023-08-22 |
CN116625409B CN116625409B (en) | 2023-10-20 |
Family
ID=87592409
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310863675.8A Active CN116625409B (en) | 2023-07-14 | 2023-07-14 | Dynamic positioning performance evaluation method, device and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116625409B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105091911A (en) * | 2015-09-07 | 2015-11-25 | 中国人民解放军信息工程大学 | Detection system and method for dynamic positioning precision of POS (point of sale) system |
JP2017020922A (en) * | 2015-07-13 | 2017-01-26 | 株式会社日立システムズ | Work action support navigation system and method, and computer program for work action support navigation, memory medium storing program for work action support navigation, self-propelled robot loaded with work action support navigation system, and intelligent helmet used with work action support navigation system |
CN106960456A (en) * | 2017-03-28 | 2017-07-18 | 长沙全度影像科技有限公司 | A kind of method that fisheye camera calibration algorithm is evaluated |
CN110136125A (en) * | 2019-05-17 | 2019-08-16 | 北京深醒科技有限公司 | One kind replicating mobile counterfeiting detection method based on the matched image of level characteristics point |
US20200380683A1 (en) * | 2018-02-28 | 2020-12-03 | Canon Kabushiki Kaisha | Image processing apparatus, image processing method, and storage medium |
CN113091768A (en) * | 2021-03-12 | 2021-07-09 | 南京理工大学 | MIMU integral dynamic intelligent calibration compensation method |
CN114777771A (en) * | 2022-04-13 | 2022-07-22 | 西安电子科技大学 | Outdoor unmanned vehicle combined navigation positioning method |
CN115272923A (en) * | 2022-07-22 | 2022-11-01 | 华中科技大学同济医学院附属协和医院 | Intelligent identification method and system based on big data platform |
CN116311594A (en) * | 2023-05-11 | 2023-06-23 | 中国人民解放军海军工程大学 | Ship subsystem state analysis method, device and storage medium |
-
2023
- 2023-07-14 CN CN202310863675.8A patent/CN116625409B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017020922A (en) * | 2015-07-13 | 2017-01-26 | 株式会社日立システムズ | Work action support navigation system and method, and computer program for work action support navigation, memory medium storing program for work action support navigation, self-propelled robot loaded with work action support navigation system, and intelligent helmet used with work action support navigation system |
CN105091911A (en) * | 2015-09-07 | 2015-11-25 | 中国人民解放军信息工程大学 | Detection system and method for dynamic positioning precision of POS (point of sale) system |
CN106960456A (en) * | 2017-03-28 | 2017-07-18 | 长沙全度影像科技有限公司 | A kind of method that fisheye camera calibration algorithm is evaluated |
US20200380683A1 (en) * | 2018-02-28 | 2020-12-03 | Canon Kabushiki Kaisha | Image processing apparatus, image processing method, and storage medium |
CN110136125A (en) * | 2019-05-17 | 2019-08-16 | 北京深醒科技有限公司 | One kind replicating mobile counterfeiting detection method based on the matched image of level characteristics point |
CN113091768A (en) * | 2021-03-12 | 2021-07-09 | 南京理工大学 | MIMU integral dynamic intelligent calibration compensation method |
CN114777771A (en) * | 2022-04-13 | 2022-07-22 | 西安电子科技大学 | Outdoor unmanned vehicle combined navigation positioning method |
CN115272923A (en) * | 2022-07-22 | 2022-11-01 | 华中科技大学同济医学院附属协和医院 | Intelligent identification method and system based on big data platform |
CN116311594A (en) * | 2023-05-11 | 2023-06-23 | 中国人民解放军海军工程大学 | Ship subsystem state analysis method, device and storage medium |
Non-Patent Citations (2)
Title |
---|
吕栋等: "基于MEA-BP神经网络的卫星钟差预报", 《测绘学报》, no. 08 * |
翁吴彬: "北斗卫星导航***动态定位精度测试研究", 《计算机产品与流通》, no. 08 * |
Also Published As
Publication number | Publication date |
---|---|
CN116625409B (en) | 2023-10-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10867167B2 (en) | Collaborative deep network model method for pedestrian detection | |
CN113129341B (en) | Landing tracking control method and system based on light-weight twin network and unmanned aerial vehicle | |
CN112836640B (en) | Single-camera multi-target pedestrian tracking method | |
Siskind et al. | A maximum-likelihood approach to visual event classification | |
CN108748149B (en) | Non-calibration mechanical arm grabbing method based on deep learning in complex environment | |
CN109102547A (en) | Robot based on object identification deep learning model grabs position and orientation estimation method | |
CN108090896B (en) | Wood board flatness detection and machine learning method and device and electronic equipment | |
CN105069799A (en) | Angular point positioning method and apparatus | |
CN101523412A (en) | Face-based image clustering | |
CN112621766B (en) | Flying shooting image error prediction method based on machine learning | |
CN109389156B (en) | Training method and device of image positioning model and image positioning method | |
CN110045740A (en) | A kind of Mobile Robot Real-time Motion planing method based on human behavior simulation | |
CN112947458B (en) | Robot accurate grabbing method based on multi-mode information and computer readable medium | |
CN113129284B (en) | Appearance detection method based on 5G cloud edge cooperation and implementation system | |
CN109927033A (en) | A kind of target object dynamic adaptation method applied to conveyer belt sorting | |
CN114677323A (en) | Semantic vision SLAM positioning method based on target detection in indoor dynamic scene | |
CN110633738A (en) | Rapid classification method for industrial part images | |
CN112927264A (en) | Unmanned aerial vehicle tracking shooting system and RGBD tracking method thereof | |
CN111582395A (en) | Product quality classification system based on convolutional neural network | |
CN109544584B (en) | Method and system for realizing inspection image stabilization precision measurement | |
CN108280516B (en) | Optimization method for mutual-pulsation intelligent evolution among multiple groups of convolutional neural networks | |
CN116625409B (en) | Dynamic positioning performance evaluation method, device and system | |
CN115984319A (en) | Unmanned aerial vehicle target tracking method based on improved YoLov7 and Deepsort | |
CN113436293B (en) | Intelligent captured image generation method based on condition generation type countermeasure network | |
CN113949826A (en) | Unmanned aerial vehicle cluster cooperative reconnaissance method and system under limited communication bandwidth condition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CP03 | Change of name, title or address |
Address after: 301, 3rd Floor, Commercial and Garage, No. 17 Zhichun Road, Haidian District, Beijing, 100080 Patentee after: Xiangke Intelligent Technology (Beijing) Co.,Ltd. Country or region after: China Address before: A09, 9th Floor, Jinqiu International Building, No. 6 Zhichun Road, Haidian District, Beijing, 100080 Patentee before: Xiangke Intelligent Technology (Beijing) Co.,Ltd. Country or region before: China |
|
CP03 | Change of name, title or address |