A method for assisting positioning of a robot for inspecting the bottom of a vehicle

Abstract

The application discloses a kind of car bottom inspection robot auxiliary positioning methods, it is related to the technical field of robot auxiliary positioning, first, the data feature of specified component of the car bottom data after noise reduction processing is compared with being constructed, determine that car bottom data includes specified component data;Then the position adjustment signal of specified component is returned to car bottom inspection robot, and according to the mapping relationship of car bottom data and the space coordinates provided by car bottom inspection robot, the space coordinates of the specified component located in car bottom are calculated;Finally, the position adjustment offset of car bottom inspection robot is determined according to the space coordinates of the position where car bottom inspection robot itself is located and the space coordinates of the specified component located in car bottom;The application does not depend on the identification of fixed landmark mark, is flexible and convenient, and the maintenance cost is relatively low, and is not affected by external environment, is more suitable for the actual situation of subway car bottom inspection business.

Description

Technical Field

[0001] This invention relates to the technical field of robot-assisted positioning, and specifically to a method for assisting positioning of a vehicle undercarriage inspection robot. Background Technology

[0002] The statements in this section are provided only as background information in connection with this disclosure and may not constitute prior art.

[0003] After arriving at the subway car undercarriage inspection tunnel according to the planned path, the subway car undercarriage inspection robot travels along the predetermined inspection direction; after passing the bogie and through passage, the robot needs to confirm the coordinates of the work position again through auxiliary positioning.

[0004] Currently, existing assisted positioning technologies are mainly based on road signs. This technology involves installing several road signs at characteristic points in the environment where the AGV robot is located. The AGV robot uses sensors to measure the road signs and calculate its own pose. This method relies on road signs in the environment, and the actual positioning accuracy depends mainly on the accurate identification of the road signs and the accuracy and speed of extracting environmental position information.

[0005] The disadvantages of the existing technology are that the installation and maintenance costs are high, and the sensor detection has blind spots when the external environment changes dynamically. Secondly, in the scenario of subway car undercarriage inspection, the installation of road sign devices under the car is not in line with the actual business due to the limitation of vehicle space. Summary of the Invention

[0006] The purpose of this invention is to address the problems of high installation and maintenance costs, blind spots in sensor detection when the external environment changes dynamically, and the impracticality of installing road marking devices under vehicles due to space limitations. This invention provides a method for assisting the positioning of under-vehicle inspection robots. By analyzing data from a laser rangefinder and / or images from a line scan camera, the method sends a position adjustment signal to the inspection robot and provides the position adjustment offset, thereby enabling the robot to find the optimal working position and solving the aforementioned problems.

[0007] The technical solution of the present invention is as follows:

[0008] A method for assisting in the positioning of a vehicle undercarriage inspection robot includes:

[0009] The noise-reduced undercarriage data is compared with the data features of the specified components to determine that the undercarriage data includes the specified component data.

[0010] The system returns a position adjustment signal for a designated component to the undercarriage inspection robot, and calculates the spatial coordinates of the designated component located under the vehicle based on the mapping relationship between the undercarriage data and the spatial coordinates provided by the undercarriage inspection robot.

[0011] The position adjustment offset of the undercarriage inspection robot is determined based on the spatial coordinates of the robot's own location and the spatial coordinates of the designated components located under the vehicle.

[0012] Furthermore, the undercarriage data includes: undercarriage waveform data acquired by a laser sensor; the noise reduction processing includes: performing noise reduction processing on the undercarriage waveform data using a clustering algorithm or a filtering algorithm;

[0013] or / and

[0014] The vehicle undercarriage data includes: vehicle undercarriage image data acquired by the line scanning module; the noise reduction processing includes: using a filtering algorithm to perform noise reduction processing on the vehicle undercarriage image data acquired by the line scanning module.

[0015] Furthermore, the data features of the specified component are constructed using a direct method;

[0016] The direct method construction includes:

[0017] Construct a feature vector B for a specified component; the feature vector B is a waveform feature vector B1 and / or an image feature vector B2.

[0018] Furthermore, the data features of the specified components are constructed through deep learning modeling;

[0019] The deep learning model construction includes:

[0020] Noise reduction processing is performed on a large amount of data corresponding to a specified component, wherein the data is waveform data and / or image data;

[0021] Based on the denoised data, a discrimination model for a specified component is constructed using a deep learning model; the discrimination model is a waveform discrimination model and / or an image discrimination model.

[0022] Furthermore, the comparison is performed using a direct comparison method;

[0023] The direct comparison method includes:

[0024] Extract feature vector A from the vehicle underside data; feature vector A is waveform feature vector A1 and / or image feature vector A2; feature vector A and feature vector B are of the same type.

[0025] Calculate the cosine similarity between feature vector A and feature vector B;

[0026] If the cosine similarity is greater than or equal to the threshold s0, the undercarriage data includes the specified component data; if the cosine similarity is less than the threshold s0, the undercarriage data does not include the specified component data, and new undercarriage data needs to be obtained.

[0027] Furthermore, the formula for calculating the cosine similarity is as follows:

[0028]

[0029] Furthermore, the comparison is performed using a deep learning discriminant method;

[0030] The deep learning discriminant comparison includes:

[0031] The discriminant model is used to discriminate and predict the undercarriage data, and the probability P that the undercarriage data includes the data of specified parts is output.

[0032] If the probability P is greater than or equal to the threshold P0, the undercarriage data includes the specified component data; if the probability P is less than the threshold P0, the undercarriage data does not include the specified component data, and new undercarriage data needs to be obtained.

[0033] Furthermore, the mapping relationship includes:

[0034] The undercarriage inspection robot provides K spatial coordinates corresponding to the horizontal coordinate of the undercarriage data in the real working environment per unit time.

[0035] Let the vector formed by K spatial coordinates be:

[0036] vp = [p0, p1, ..., p] K ]

[0037] Among them, p0 and p K These are the spatial coordinates corresponding to the x-coordinates of the starting and ending points of the vehicle underside data, respectively.

[0038] The index of the kth spatial coordinate in the vehicle undercarriage data is:

[0039]

[0040] Where N is the number of horizontal coordinates in the vehicle undercarriage data acquired per unit time;

[0041] The vector formed by the indices of the vehicle undercarriage data corresponding to the K spatial coordinates is:

[0042] n(vp)=[n(p0),n(p1)……n(p K )).

[0043] Furthermore, when the undercarriage data includes undercarriage waveform data;

[0044] The mapping relationship includes:

[0045] The undercarriage inspection robot provides K spatial coordinates in the real working environment corresponding to the horizontal coordinates of the data points of the undercarriage waveform data per unit time;

[0046] Let the vector formed by K spatial coordinates be:

[0047] vp = [p0, p1, ..., p] K ]

[0048] Among them, p0 and p K These are the spatial coordinates corresponding to the x-coordinates of the first and last data points of the vehicle underside waveform data, respectively.

[0049] The index of the kth spatial coordinate in the vehicle underside waveform data is:

[0050]

[0051] Where N is the number of data points in the waveform data of the vehicle underside acquired per unit time;

[0052] The vector formed by the indices of the waveform data under the vehicle corresponding to the K spatial coordinates is:

[0053] n(vp)=[n(p0),n(p1)……n(p K )]

[0054] The calculation of the spatial coordinates of a specified component located under the vehicle includes:

[0055] Waveform breakpoints in the vehicle underside waveform data are obtained using clustering algorithms, Jenks' natural breakpoint algorithm, or kernel density estimation algorithm; these breakpoints constitute multiple waveforms.

[0056] Calculate the range of each waveform and compare it with the threshold range; if the range falls within the threshold range, determine that the waveform corresponding to the range is the waveform of the specified component located under the vehicle, and extract the waveform data of the waveform.

[0057] The waveform data is fitted using a polynomial function, and the minimum value of the polynomial function is obtained using the gradient descent algorithm. The x-coordinate of the corresponding data point in the vehicle underside waveform data is then obtained.

[0058] Find the x-coordinate The relative position of the vehicle underside waveform data sequence number corresponding to K spatial coordinates;

[0059] Let the x-coordinate be... The position is at n(p) k ) to n(p k+1 If the x-axis is between ), then the x-axis is... The corresponding spatial coordinates are:

[0060]

[0061] Furthermore, when the undercarriage data includes undercarriage image data;

[0062] The mapping relationship includes:

[0063] The undercarriage inspection robot provides K spatial coordinates corresponding to the horizontal coordinate of the undercarriage image data in the real working environment per unit time.

[0064] Let the vector formed by K spatial coordinates be:

[0065] vp = [p0, p1, ..., p] K ]

[0066] Among them, p0 and p K These are the spatial coordinates corresponding to the leftmost and rightmost horizontal coordinates of the vehicle underside image data, respectively.

[0067] The index of the kth spatial coordinate in the vehicle undercarriage image data is:

[0068]

[0069] Where N is the number of horizontal coordinates in the undercarriage image data acquired per unit time;

[0070] The vector formed by the indices of the K spatial coordinates corresponding to the vehicle undercarriage image data is:

[0071] n(vp)=[n(p0),n(p1)……n(p K )]

[0072] The calculation of the spatial coordinates of a specified component located under the vehicle's underside based on the underside data includes:

[0073] Image segmentation algorithms are used to extract image regions of specified components from under-vehicle image data.

[0074] Obtain the x-coordinate of the center point of a specified component image region in the vehicle underside image data.

[0075] Find the x-coordinate The relative position of the vehicle undercarriage image data sequence number corresponding to K spatial coordinates;

[0076] Let the x-coordinate be... The position is at n(p) k ) to n(p k+1 If the x-axis is between ), then the x-axis is... The corresponding spatial coordinates are:

[0077]

[0078] Furthermore, determining the position adjustment offset of the undercarriage inspection robot includes:

[0079]

[0080] Where Δp is the position adjustment offset of the undercarriage inspection robot, and p(A) is the spatial coordinate of the robot's own center obtained by the undercarriage inspection robot;

[0081] When Δp is less than 0, the undercarriage inspection robot needs to retreat a distance of |Δp|.

[0082] When Δp is greater than 0, the undercarriage inspection robot needs to move forward a distance of |Δp|.

[0083] When Δp = 0, the undercarriage inspection robot does not need to move.

[0084] Compared with existing technologies, the advantages of this invention are:

[0085] A method for assisting in the positioning of a vehicle undercarriage inspection robot includes: comparing noise-reduced vehicle undercarriage data with the data features of designated components to determine that the vehicle undercarriage data includes the data of designated components; returning a position adjustment signal of the designated components to the vehicle undercarriage inspection robot, and calculating the spatial coordinates of the designated components located under the vehicle undercarriage based on the mapping relationship between the vehicle undercarriage data and the spatial coordinates provided by the vehicle undercarriage inspection robot; determining the position adjustment offset of the vehicle undercarriage inspection robot based on the spatial coordinates of the robot's own position and the spatial coordinates of the designated components located under the vehicle undercarriage; and sending a position adjustment signal to the inspection robot and providing the position adjustment offset by analyzing data from a laser rangefinder and / or images from a line scan camera, thereby enabling the inspection robot to find the optimal working position. This method is flexible and convenient, as it does not rely on the recognition of fixed road signs. Furthermore, it can simultaneously identify multiple vehicle undercarriage components and output their spatial coordinates. Moreover, since it relies only on a laser distance sensor and a line scan camera, the maintenance cost is relatively low. And when the external environment changes dynamically, it does not cause significant interference to the solution, making it suitable for the actual situation of subway vehicle undercarriage inspection operations. Attached Figure Description

[0086] Figure 1 A flowchart of an auxiliary positioning method for a vehicle under-vehicle inspection robot;

[0087] Figure 2 A schematic diagram illustrating laser sensor data acquisition;

[0088] Figure 3 A schematic diagram of the waveform data of the vehicle underside acquired by the laser sensor;

[0089] Figure 4 This is a schematic diagram illustrating data acquisition by the line scan module.

[0090] Figure 5 A schematic diagram of the scanning area for the line scanning module to scan the subway car underpass passageway;

[0091] Figure 6 A schematic diagram of noise reduction for waveform data under the vehicle body;

[0092] Figure 7 A schematic diagram illustrating the noise reduction process for undercarriage image data;

[0093] Figure 8 Flowchart for obtaining the spatial coordinates of specified components based on undercarriage waveform data;

[0094] Figure 9 This is a schematic diagram of the waveform breakpoint obtained in Example 1;

[0095] Figure 10 This is a schematic diagram of the waveform corresponding to the specified component obtained in Example 1;

[0096] Figure 11 Flowchart for obtaining spatial coordinates of specified components based on undercarriage image data;

[0097] Figure 12 A schematic diagram illustrating the calculation of the offset for adjusting the position of the undercarriage inspection robot. Detailed Implementation

[0098] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0099] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0100] Example 1

[0101] Currently, existing assisted positioning technologies are mainly based on road signs. This technology involves installing several road signs at characteristic points in the environment where the AGV robot is located. The AGV robot uses sensors to measure the road signs and calculate its own pose. This method relies on road signs in the environment, and the actual positioning accuracy depends mainly on the accurate identification of the road signs and the accuracy and speed of extracting environmental position information.

[0102] The disadvantages of the existing technology are that the installation and maintenance costs are high, and the sensor detection has blind spots when the external environment changes dynamically. Secondly, in the scenario of subway car undercarriage inspection, the installation of road sign devices under the car is not in line with the actual business due to the limitation of vehicle space.

[0103] In this embodiment, it should be noted that the undercarriage inspection robot consists of a working robotic arm, an inspection trolley, and various sensors.

[0104] This embodiment addresses the aforementioned problems by proposing an auxiliary positioning method for a vehicle under-vehicle inspection robot. By analyzing data from a laser rangefinder and / or images from a line scan camera, a position adjustment signal is sent to the inspection robot, along with the position adjustment offset, thereby enabling the inspection robot to find the optimal working position.

[0105] Please see Figure 1 A method for assisting in the positioning of a vehicle undercarriage inspection robot, specifically including:

[0106] The noise-reduced undercarriage data is compared with the data features of the specified components to determine whether the undercarriage data includes the specified component data; that is, to determine whether the undercarriage inspection robot has reached the working position corresponding to the specified component located under the vehicle.

[0107] The system returns a position adjustment signal for a specified component to the undercarriage inspection robot, and calculates the spatial coordinates of the specified component located under the vehicle based on the mapping relationship between the undercarriage data and the spatial coordinates provided by the undercarriage inspection robot; the adjustment signal is, for example, the return value 1 for the through passage, the return value 2 for the bogie wheel axle, etc.

[0108] The position adjustment offset of the undercarriage inspection robot is determined based on the spatial coordinates of its own location and the spatial coordinates of the designated component located under the vehicle. Preferably, the spatial coordinates of the undercarriage inspection robot's own location are directly provided by the inspection trolley, that is, the inspection trolley provides the spatial coordinates of its own center location.

[0109] In this embodiment, specifically, the vehicle underside data includes: waveform data of the vehicle underside acquired by a laser sensor, such as... Figure 3As shown; the noise reduction process includes: using clustering algorithms or filtering algorithms to perform noise reduction processing on the vehicle underside waveform data; the noise reduction processing for the vehicle underside waveform data is as follows: Figure 6 As shown;

[0110] or / and

[0111] The vehicle undercarriage data includes: vehicle undercarriage image data acquired by the line scanning module; the noise reduction processing includes: performing noise reduction processing on the vehicle undercarriage image data acquired by the line scanning module using a filtering algorithm; preferably, a Gaussian filtering algorithm is used for noise reduction processing; taking the vehicle undercarriage image data as an example, the noise reduction processing for the vehicle undercarriage image data is as follows: Figure 7 As shown.

[0112] Specifically, whether to choose undercarriage waveform data or undercarriage image data as the undercarriage data depends mainly on the type of the specified component; for example, undercarriage waveform data is used for bogie axles, while undercarriage image data is used for through passages.

[0113] like Figure 2 As shown, the laser beam of the laser sensor is perpendicular to the upper surface of the inspection trolley, and the laser point is located on the bottom surface of the subway car. During the movement of the inspection trolley, the bottom of the subway car forms a laser point movement trajectory, and the laser sensor acquires the corresponding bottom waveform data.

[0114] like Figure 4 As shown, the scanning direction of the line scanning module is perpendicular to the upper surface of the inspection trolley, and the scanning area is located on the underside surface of the subway car; during the movement of the inspection trolley, the scanning area of ​​the line scanning module is a rectangular area; taking the recognition of the subway car underpass as an example, the scanning area is as follows: Figure 5 As shown.

[0115] In this embodiment, the data features of the specified component are specifically constructed using a direct method;

[0116] The direct method construction includes: constructing a feature vector B for a specified component; the feature vector B is a waveform feature vector B1 and / or an image feature vector B2.

[0117] That is, by using prior knowledge, a waveform feature vector B1 is constructed for a specified component (such as bogie wheel axle), such as the distance between adjacent inflection points, amplitude and other features;

[0118] Using prior knowledge, construct an image feature vector B2 or an image feature matrix B3 for the image data of a specified component (such as a through passage). For example, a gray-level co-occurrence matrix that reflects texture features. When it is an image feature matrix, it is also necessary to convert the image feature matrix into an image feature vector. Therefore, it is necessary to convert the image feature matrix B3 into an image feature vector B3.

[0119] In this embodiment, the comparison is specifically performed using a direct comparison method;

[0120] The direct comparison method includes:

[0121] Extract feature vector A from the vehicle underside data; feature vector A is waveform feature vector A1 and / or image feature vector A2; feature vector A and feature vector B are of the same type; that is, if the vehicle underside data is waveform data, then feature vector B is waveform feature vector B1, and feature vector A is waveform feature vector A1; furthermore, feature vector A can be extracted using the same extraction method as feature vector B; when the extracted feature matrix is ​​image feature matrix A3, it is also necessary to convert image feature matrix A3 into image feature vector A3;

[0122] Calculate the cosine similarity between feature vector A and feature vector B;

[0123] If the cosine similarity is greater than or equal to the threshold s0, then the undercarriage data includes the specified component data. When the undercarriage data is waveform data, it means that the waveform data includes the waveform data of the specified component, indicating that the inspection robot has initially reached the working position. When the undercarriage data includes undercarriage image data, it means that the image data includes the image data of the specified component, indicating that the inspection robot has initially reached the working position. If the cosine similarity is less than the threshold s0, then the undercarriage data does not include the specified component data, and new undercarriage data needs to be acquired.

[0124] The formula for calculating cosine similarity is as follows:

[0125]

[0126] Let the eigenvectors be A = (a1, a2, ..., a3). p B = (b1, b2, ..., b) p ),

[0127] The cosine similarity between eigenvector A and eigenvector B is:

[0128]

[0129] If the extracted features are image feature matrix B3 and image feature matrix A3, then A3 and A3 need to be reconstructed into vectors first; then the cosine similarity is calculated.

[0130] Let the image feature matrices A3 and B3 be:

[0131]

[0132] First, reconstruct the image feature matrices A3 and B3 into vectors. The reconstructed vectors are as follows:

[0133] A3 = (a1, a2, ..., a p ,ap+1 ,a p+2 …,a 2p ,a 2p+1 ,…,a 3p )

[0134] B3 = (b1, b2, ..., b p ,b p+1 ,b p+2 …,b 2p ,b 2p+1 ,…,b 3p )

[0135] The cosine similarity between image feature vectors A3 and B3 is:

[0136]

[0137] In this embodiment, specifically, when the vehicle undercarriage data includes vehicle undercarriage waveform data; please refer to [link to relevant documentation]. Figure 10 ;

[0138] The mapping relationship includes:

[0139] The undercarriage inspection robot provides K spatial coordinates (one-dimensional coordinates) corresponding to the abscissa of the data points of the undercarriage waveform data per unit time in the real working environment; that is, the inspection vehicle provides the spatial coordinates (one-dimensional coordinates) corresponding to the undercarriage waveform data in the real working environment. Assuming that in this embodiment, the frequency of spatial coordinate acquisition is K = 50, meaning the inspection vehicle provides the spatial coordinates corresponding to the abscissa of 50 data points in the undercarriage waveform data per second, and the number of data points collected per second in the undercarriage waveform data is N = 1000;

[0140] Let the vector formed by K spatial coordinates be:

[0141] vp = [p0, p1, ..., p] K ]

[0142] Among them, p0 and p K These are the spatial coordinates corresponding to the x-coordinates of the first and last data points of the vehicle underside waveform data, respectively.

[0143] The index of the kth spatial coordinate in the vehicle underside waveform data is:

[0144]

[0145] Where N is the number of data points in the waveform data of the vehicle underside acquired per unit time;

[0146] The vector formed by the indices of the waveform data under the vehicle corresponding to the K spatial coordinates is:

[0147] n(vp)=[n(p0),n(p1)……n(pK )]

[0148] Please see Figure 8 The calculation of the spatial coordinates of a specified component located under the vehicle includes:

[0149] Waveform breakpoints in the vehicle underside waveform data are obtained using clustering algorithms, Jenks' natural breakpoint algorithm, or kernel density estimation algorithm; the waveform breakpoints constitute multiple waveforms; preferably, two adjacent waveform breakpoints constitute one waveform.

[0150] Figure 9 An example of waveform breakpoints is given, where a, b, c, and d are four waveform breakpoints, and these four waveform breakpoints form three waveform segments ab, bc, and cd.

[0151] Calculate the range of each waveform and compare it with the threshold range; if the range falls within the threshold range, determine that the waveform corresponding to the range is the waveform of the specified component located under the vehicle, and extract the waveform data of the waveform.

[0152] like Figure 10 As shown, the range Δab of waveform ab is equal to the maximum value minus the minimum value between points a and b, the range Δbc of waveform bc is equal to the maximum value minus the minimum value between points b and c, and the range Δcd of waveform cd is equal to the maximum value minus the minimum value between points c and d. Suppose that when the range Δ is greater than or equal to Δ1 and less than or equal to Δ2, the corresponding waveform is a specified component; then waveforms ab and cd are the waveforms corresponding to the specified component.

[0153] The waveform data is fitted using a polynomial function, and the minimum value of the polynomial function is obtained using the gradient descent algorithm. The x-coordinate of the corresponding data point in the vehicle underside waveform data is then obtained.

[0154] Find the x-coordinate The relative position of the vehicle underside waveform data sequence number corresponding to K spatial coordinates;

[0155] Let the x-coordinate be... The position is at n(p) k ) to n(p k+1 If the x-axis is between ), then the x-axis is... The corresponding spatial coordinates are:

[0156]

[0157] In this embodiment, specifically, when the vehicle undercarriage data includes vehicle undercarriage image data;

[0158] The mapping relationship includes:

[0159] The undercarriage inspection robot provides K spatial coordinates (in one dimension) corresponding to the abscissa of the undercarriage image data in the real working environment per unit time; that is, the inspection vehicle provides the spatial coordinates (in one dimension) corresponding to the undercarriage image data in the real working environment. Assuming that in this embodiment, the frequency of spatial coordinate acquisition is K = 50, meaning the inspection vehicle provides the spatial coordinates corresponding to the abscissa of 50 undercarriage image data per second, and the number of abscissas collected per second is N = 1000;

[0160] Let the vector formed by K spatial coordinates be:

[0161] vp = [p0, p1, ..., p] K ]

[0162] Among them, p0 and p K These are the spatial coordinates corresponding to the leftmost and rightmost horizontal coordinates of the vehicle underside image data, respectively.

[0163] The index of the kth spatial coordinate in the vehicle undercarriage image data is:

[0164]

[0165] Where N is the number of horizontal coordinates in the undercarriage image data acquired per unit time;

[0166] The vector formed by the indices of the K spatial coordinates corresponding to the vehicle undercarriage image data is:

[0167] n(vp)=[n(p0),n(p1)……n(p K )];

[0168] Please see Figure 11 The step of calculating the spatial coordinates of a specified component located under the vehicle based on the vehicle underside data includes:

[0169] Image segmentation algorithms are used to extract image regions of specified components from under-vehicle image data.

[0170] Obtain the x-coordinate of the center point of a specified component image region in the vehicle underside image data.

[0171] Find the x-coordinate The relative position of the vehicle undercarriage image data sequence number corresponding to K spatial coordinates;

[0172] Let the x-coordinate be... The position is at n(p) k ) to n(p k+1 If the x-axis is between ), then the x-axis is... The corresponding spatial coordinates are:

[0173]

[0174] In this embodiment, specifically, such as Figure 12 As shown, determining the position adjustment offset of the undercarriage inspection robot includes:

[0175]

[0176] Where Δp is the position adjustment offset of the undercarriage inspection robot, and p(A) is the spatial coordinate of the robot's own center obtained by the undercarriage inspection robot; that is, the spatial coordinate of the inspection vehicle's own center obtained by the inspection vehicle.

[0177] When Δp is less than 0, the undercarriage inspection robot needs to retreat a distance of |Δp|.

[0178] When Δp is greater than 0, the undercarriage inspection robot needs to move forward a distance of |Δp|.

[0179] When Δp = 0, the undercarriage inspection robot does not need to move.

[0180] Example 2

[0181] Example 2 is a further improvement on Example 1. The same components will not be described again here; please refer to [link / reference]. Figure 1-3 In this embodiment, the data features of the specified component are constructed through deep learning modeling;

[0182] The deep learning model construction includes:

[0183] Denoising processing is performed on a large amount of data corresponding to a specified component, wherein the data is waveform data and / or image data; that is, a large amount of existing waveform data and / or image data corresponding to the specified component should be collected first, and the noise reduction processing method can be directly adopted in the aforementioned manner.

[0184] Based on the denoised data, a discrimination model for a specified component is constructed using a deep learning model; the discrimination model is a waveform discrimination model and / or an image discrimination model; preferably, the deep learning model is, for example, a ResNet series network, a SENet network, a DCL network, etc.

[0185] In this embodiment, the comparison is performed using a deep learning discriminant method;

[0186] The deep learning discriminant comparison includes:

[0187] The discriminant model is used to discriminate and predict the undercarriage data, and the probability P of the undercarriage data including the specified component data is output; that is, the undercarriage data is predicted based on the waveform discriminant model or the image discriminant model, and the probability P of the undercarriage data including the waveform or image of the specified component is output.

[0188] If the probability P is greater than or equal to the threshold P0, the undercarriage data includes the specified component data; if the probability P is less than the threshold P0, the undercarriage data does not include the specified component data, and new undercarriage data needs to be obtained.

[0189] The embodiments described above merely illustrate specific implementation methods of this application, and while the descriptions are detailed and specific, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the technical solution of this application, and these modifications and improvements all fall within the scope of protection of this application.

[0190] This background section is provided to generally present the context of the invention. The work of the currently named inventors, the work to the extent described in this background section, and aspects of this section that did not constitute prior art at the time of application are neither expressly nor impliedly acknowledged as prior art to the invention.

Claims

1. A method for assisting positioning of a car bottom inspection robot, characterized in that, include: The noise-reduced undercarriage data is compared with the data features of the specified components to determine that the undercarriage data includes the specified component data. The system returns a position adjustment signal for a designated component to the undercarriage inspection robot, and calculates the spatial coordinates of the designated component located under the vehicle based on the mapping relationship between the undercarriage data and the spatial coordinates provided by the undercarriage inspection robot. The position adjustment offset of the undercarriage inspection robot is determined based on the spatial coordinates of the robot's own location and the spatial coordinates of the designated components located under the vehicle. The mapping relationship includes: The undercarriage inspection robot provides K spatial coordinates corresponding to the horizontal coordinate of the undercarriage data in the real working environment per unit time. Let the vector formed by K spatial coordinates be: wherein, and are the spatial coordinates corresponding to the horizontal coordinates of the start and end points of the car bottom data, respectively; The index of the kth spatial coordinate in the vehicle undercarriage data is: Where N is the number of horizontal coordinates in the vehicle undercarriage data acquired per unit time; The vector formed by the indices of the vehicle undercarriage data corresponding to the K spatial coordinates is: When the vehicle undercarriage data includes vehicle undercarriage waveform data; The calculation of the spatial coordinates of a specified component located under the vehicle includes: Waveform breakpoints in the vehicle underside waveform data are obtained using clustering algorithms, Jenks' natural breakpoint algorithm, or kernel density estimation algorithm; these breakpoints constitute multiple waveforms. Calculate the range of each waveform and compare it with the threshold range; if the range falls within the threshold range, determine that the waveform corresponding to the range is the waveform of the specified component located under the vehicle, and extract the waveform data of the waveform. Fitting the waveform data with a polynomial function, obtaining the minimum value of the polynomial function by using a gradient descent algorithm, and obtaining the abscissa of the corresponding data point in the bogie waveform data ; Finding the horizontal coordinate The relative position in the data sequence of the car bottom wave form corresponding to the K spatial coordinates; Let the x-coordinate be... The position is in arrive Between, the x-axis The corresponding spatial coordinates are: 。 2. The undercarriage inspection robot-assisted positioning method according to claim 1, characterized in that, The vehicle underside data includes: vehicle underside waveform data acquired by a laser sensor; the noise reduction processing includes: performing noise reduction processing on the vehicle underside waveform data using clustering algorithms or filtering algorithms; or / and The vehicle undercarriage data includes: vehicle undercarriage image data acquired by the line scanning module; the noise reduction processing includes: using a filtering algorithm to perform noise reduction processing on the vehicle undercarriage image data acquired by the line scanning module.

3. The undercarriage inspection robot-assisted positioning method according to claim 2, characterized in that, The data features of the specified components are constructed using a direct method; The direct method construction includes: Construct a feature vector B for a specified component; the feature vector B is a waveform feature vector B1 and / or an image feature vector B2.

4. The undercarriage inspection robot-assisted positioning method according to claim 2, characterized in that, The data features of the specified components are constructed through deep learning modeling. The deep learning model construction includes: Noise reduction processing is performed on a large amount of data corresponding to a specified component, wherein the data is waveform data and / or image data; Based on the denoised data, a discrimination model for a specified component is constructed using a deep learning model; the discrimination model is a waveform discrimination model and / or an image discrimination model.

5. The undercarriage inspection robot-assisted positioning method according to claim 3, characterized in that, The comparison was performed using a direct comparison method. The direct comparison method includes: Extract feature vector A from the vehicle underside data; feature vector A is waveform feature vector A1 and / or image feature vector A2; feature vector A and feature vector B are of the same type. Calculate the cosine similarity between feature vector A and feature vector B; If the cosine similarity is greater than or equal to the threshold s0, the undercarriage data includes the specified component data; if the cosine similarity is less than the threshold s0, the undercarriage data does not include the specified component data, and new undercarriage data needs to be obtained.

6. The undercarriage inspection robot-assisted positioning method according to claim 4, characterized in that, The comparison was performed using a deep learning discriminant method. The deep learning discriminant comparison includes: The discriminant model is used to discriminate and predict the undercarriage data, and the probability P that the undercarriage data includes the data of specified parts is output. If the probability P is greater than or equal to the threshold P0, the undercarriage data includes the specified component data; if the probability P is less than the threshold P0, the undercarriage data does not include the specified component data, and new undercarriage data needs to be obtained.

7. The assisted positioning method for a vehicle undercarriage inspection robot according to claim 1, characterized in that, When the vehicle undercarriage data includes undercarriage image data; Calculate the spatial coordinates of a specified component located under the vehicle based on the underbody data, including: Image segmentation algorithms are used to extract image regions of specified components from under-vehicle image data. Obtain the x-coordinate of the center point of a specified component image region in the vehicle underside image data. ; Find the x-coordinate The relative position of the vehicle undercarriage image data sequence number corresponding to K spatial coordinates; Let the x-coordinate be... The position is in arrive Between, the x-axis The corresponding spatial coordinates are: 。 8. The undercarriage inspection robot-assisted positioning method according to claim 7, characterized in that, The determination of the position adjustment offset of the undercarriage inspection robot includes: in, Adjust the offset for the undercarriage inspection robot. The spatial coordinates of the robot's center obtained for the undercarriage inspection robot; when When the value is less than 0, the undercarriage inspection robot needs to retreat. The distance; when When the value is greater than 0, the undercarriage inspection robot needs to move forward. The distance; when When the value is 0, the undercarriage inspection robot does not need to move.

Images