CN-121994131-A - Tunnel detection vehicle positioning system based on absolute value encoder

CN121994131ACN 121994131 ACN121994131 ACN 121994131ACN-121994131-A

Abstract

The invention discloses a tunnel detection vehicle positioning system based on an absolute value encoder, which relates to the field of intelligent maintenance of subway tunnels and comprises a ring number identification module, an absolute value encoder module, a slip detection module, a main control module, a reference positioning information determining module, a precise positioning information determining module and a final positioning information determining module, wherein the ring number identification module is used for acquiring ring number images on a subway tunnel, identifying ring numbers and uploading the ring numbers to the main control module, the absolute value encoder module is used for acquiring and uploading encoder readings to the main control module, the slip detection module is used for acquiring detection vehicle motion data and detection vehicle driving wheel motion data, detecting whether slip occurs at each moment by using a pre-trained slip detection model to obtain slip data and transmitting the slip data to the main control module, and the main control module is used for determining reference positioning information of a detection vehicle based on position information corresponding to the ring numbers, determining the precise positioning information according to the reference positioning information and the encoder readings, and compensating the precise positioning information according to the slip data to obtain the final positioning information. According to the invention, through cooperation of multiple modules, the positioning precision of the tunnel detection vehicle is improved, and a support is provided for the accurate maintenance of hidden danger of the tunnel.

Inventors

DANG XIFENG
WU JINCHENG
LI KUAN
WANG BIN
GONG SHUAISHUAI
DONG XIAOLONG
SI RUI
FANG XINHUA
LV YUAN

Assignees

中铁一局集团城市轨道交通工程有限公司
中铁一局集团有限公司

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The tunnel detection vehicle positioning system based on the absolute value encoder is characterized by comprising a ring number identification module, an absolute value encoder module, a slip detection module and a main control module; The ring number identification module is used for acquiring ring number images on the subway tunnel, identifying the ring numbers and uploading the ring numbers to the main control module; The absolute value encoder module is used for collecting and uploading encoder readings to the main control module; The skid detection module is used for acquiring detected vehicle motion data and detected vehicle driving wheel motion data, detecting whether skid occurs at each moment by using a pre-trained skid detection model to obtain skid data, and transmitting the skid data to the main control module; the main control module is used for determining reference positioning information of the detection vehicle based on the position information corresponding to the ring number, determining accurate positioning information according to the reference positioning information and the reading of the encoder, and compensating the accurate positioning information according to the slipping data to obtain final positioning information.
2. The tunnel inspection vehicle positioning system based on the absolute value encoder according to claim 1, wherein the ring number identification module comprises an image acquisition unit, an image preprocessing unit and a ring number identification unit; The image acquisition unit adopts a color temperature adjustable light supplementing component and an HDR camera, is arranged at the position 1.2-1.5m in front of the head of the detection vehicle, and the camera lens faces the tunnel segment ring number area; the image preprocessing unit is used for carrying out graying, gaussian filtering denoising and Retinex image enhancement processing on the original ring number image acquired by the HDR camera, so that image interference caused by dust adhesion and uneven illumination is eliminated; the ring number recognition unit receives the preprocessed image, recognizes the ring number text information by using a built-in lightweight deep learning recognition model, and uploads the ring number text information to the main control module.
3. The absolute value encoder-based tunnel inspection vehicle positioning system of claim 2, wherein the lightweight deep learning identification model is developed and trained based on YOLOv-Nano networks, the improvement comprising: Optimizing YOLOv-Nano backbone network, replacing C2f module with C2f-Lite module, reducing Bottleneck number of each C2f-Lite module from 6 to 4, replacing standard convolution by depth separable convolution, and reducing backbone network parameter number; embedding ECA attention mechanism layers in FPN and PAN structures of YOLOv-Nano necks, carrying out global average pooling on feature graphs to obtain channel statistics, adjusting dimensions through 1X 1 convolution, activating output channel weights through Sigmoid, multiplying and outputting the channel weights with original feature graphs, and enhancing ring number character feature extraction capacity; the training comprises: Collecting subway tunnel segment ring number images covering dust coverage, greasy dirt adhesion, character fading and edge blurring under different illumination conditions, marking ring number text information corresponding to each image after preprocessing, constructing a plurality of training samples, dividing the training samples into training sets and test sets, adopting an Adam optimizer, performing iterative training by using the training sets, performing testing by using the test sets, and completing training when the recognition accuracy meets a set threshold.
4. The tunnel detection vehicle positioning system based on the absolute value encoder is characterized in that the absolute value encoder module comprises an encoder body, a speed reducer, a coding disc and a photoelectric identification unit, wherein a detection vehicle driving wheel main shaft is in driving connection with the main shaft of the encoder body through the speed reducer, the coding disc is fixed on the encoder main shaft and is driven by the encoder main shaft to rotate, code channels arranged according to a specific rule are arranged on the coding disc, each code channel corresponds to one bit of binary numbers, and the photoelectric identification unit outputs a group of binary readings through the light transmission or reflection condition of the identification code channels and transmits the binary readings to the main control module in a PN communication mode through a shielding cable.
5. The system for positioning a tunnel inspection vehicle based on an absolute value encoder according to claim 4, wherein the code channels arranged according to a specific rule are arranged so that the binary readings correspond to the absolute positions of the encoding disks one by one, and the binary readings at adjacent positions have only one bit change, so that error codes caused by simultaneous change of multiple bits during position switching are avoided, and the specific encoding method of the code channels arranged according to the specific rule is as follows: Step 1, equally dividing the coding disc into A plurality of sector areas, each of which is sequentially set with an original binary number Setting the corresponding codes after each original binary number is converted as ; Step 2, determining the highest bit of the code, and letting ; Step 3, according to the formula Sequentially calculating the rest bits of the code from the high bit to the low bit; Step 4, pressing To the point of The sequence arrangement of the sector area code channels is converted to obtain codes, and the light transmission and reflection conditions of each sector area code channel are determined according to the codes; Wherein, the Each bit of the binary number from high to low is represented separately, Each bit of the code from high to low is represented separately, , Representing an exclusive or operation.
6. The tunnel inspection vehicle positioning system based on an absolute value encoder according to claim 1, wherein the pre-trained slip detection model is constructed and trained based on a CNN-LSTM architecture, and specifically comprises: setting experimental tracks with different working conditions, and synchronously collecting detection vehicle motion data, detection vehicle driving wheel motion data and real slip labels; Carrying out data cleaning and normalization processing on the collected original data to obtain a plurality of training samples, and dividing the training samples into a training set and a testing set; The method comprises the steps of constructing a slip detection model comprising a two-channel CNN feature extraction layer, a feature fusion layer and an LSTM network, adopting an Adam optimizer, adopting a cross entropy loss function to perform model training by adopting a cross entropy loss function, testing by utilizing the training set, and completing training when the recognition accuracy meets a set threshold.
7. The tunnel inspection vehicle positioning system based on the absolute value encoder according to claim 6, wherein the different working conditions comprise three working conditions of starting, running and stopping of the inspection vehicle on oil pollution, accumulated water or dewing tracks respectively; collecting the motion data of the detection vehicle comprises the steps of collecting real-time speed, acceleration, engine output power and engine torque of the detection vehicle along a time sequence at a sampling frequency of 10 Hz; The acquisition of the motion data of the driving wheel of the detection vehicle comprises the acquisition of the real-time rotation speed and the acceleration of the driving wheel along a time sequence at the sampling frequency of 10 Hz; the real slip label collection comprises the steps of marking based on deviation between theoretical displacement of an absolute value encoder module and actual displacement of a detection vehicle, marking a slip state as 1, marking a non-slip state as 0, and recording a quantized value during slip.
8. The tunnel detection vehicle positioning system based on the absolute value encoder, which is characterized in that the two-channel CNN characteristic extraction layer comprises a first channel and a second channel, wherein the first channel takes preprocessed driving wheel rotating speed and acceleration time sequence data as input, and extracts driving wheel rotating speed change rate characteristics and driving wheel acceleration fluctuation characteristics through 3 layers of rolling layers and 2 layers of maximum pooling layers, and the second channel takes preprocessed detection vehicle speed and acceleration time sequence data as input, and extracts detection vehicle speed stability characteristics and acceleration change characteristics through 3 layers of rolling layers and 2 layers of maximum pooling layers; The feature fusion layer is used for constructing a driving wheel feature matrix and a detection vehicle feature matrix according to the features extracted by the first channel and the second channel, calculating the similarity of two matrix elements, generating weight through a softmax function, and finally calculating and generating a fusion matrix according to the weight; The LSTM network outputs the skid state and the quantized value during skid through the time sequence dependency relation of the input gate, the forgetting gate and the output gate capturing data based on the generated fusion matrix.
9. The absolute value encoder-based tunnel inspection vehicle positioning system of claim 8, wherein the drive wheel feature matrix And the feature matrix of the detection vehicle Expressed as: ; ; Wherein, the And Respectively represent the first The wheel rotation speed change rate and acceleration fluctuation characteristics of the driving wheel at each moment; And Respectively represent the first Detecting the speed stability and acceleration change characteristics of the vehicle at each moment; the calculation formula of the similarity is as follows: wherein, the method comprises the steps of, Representing similarity matrix First, the Line 1 The column elements are arranged in a row, And Representing the first and second drive wheel feature matrices and the second test vehicle feature matrix, respectively Line 1 The column elements are arranged in a row, ; The fusion matrix Is expressed as: wherein, the method comprises the steps of, 、 A first weight and a second weight, respectively, generated by a softmax function.
10. The system for locating a vehicle for detecting a tunnel based on an absolute value encoder according to claim 4, wherein said determining the reference locating information of the vehicle for detecting a tunnel comprises determining the absolute value coordinates of the ring number according to a pre-established mapping table of the ring number and the absolute position coordinates of the subway tunnel segment Determining the relative distance between the HDR camera mounting position and the detection locomotive Calculating the reference positioning information ; The method for determining the accurate positioning information comprises the steps of determining the circumference of the driving wheel of the detection vehicle Reduction ratio of the speed reducer Determining the difference value of the absolute position of the coding disc from the moment of the reference positioning information to the current moment by the ring number identification module last time Calculating the accumulated displacement And summing with the reference positioning information to obtain the accurate positioning information ; The formula for obtaining the final positioning information is as follows: wherein, the method comprises the steps of, In order to obtain the final positioning information, And detecting the sum of quantized values of the slip at each moment for the slip detection model.

Description

Tunnel detection vehicle positioning system based on absolute value encoder Technical Field The invention relates to the technical field of intelligent maintenance of subway tunnels, in particular to a tunnel detection vehicle positioning system based on an absolute value encoder. Background Along with the continuous increase of subway operation mileage and operation duration, the detection tasks of hidden dangers such as subway tunnel cracks and water leakage are also continuously increased, and the subway tunnel maintenance market is also rapidly growing. The intelligent tunnel detection vehicle is used as maintenance core equipment, the positioning accuracy of the intelligent tunnel detection vehicle directly determines hidden danger maintenance accuracy, and the intelligent tunnel detection vehicle is an industry key requirement. Although the existing detection vehicle can judge hidden danger points, the problem of insufficient positioning accuracy of hidden danger points is caused by the fact that the positioning of part of tunnel detection vehicles depends on the ring number identification of tunnel segments, the illumination in a tunnel is complex, the ring number image blurring and error identification rate is high due to dust adhesion. The partial tunnel detection vehicle calculates and positions hidden trouble points by means of wheel revolution, oil stains, ponding, temperature difference condensation and the like exist in a tunnel track, torque fluctuation exists when the detection vehicle starts and stops, wheel slip is easy to occur, accumulated slip errors exist in calculated theoretical displacement, and the positioning position deviates from actual displacement. Therefore, how to improve the positioning accuracy of the tunnel inspection vehicle is a problem to be solved by those skilled in the art. Disclosure of Invention In view of the above, the present invention provides a tunnel inspection vehicle positioning system based on an absolute value encoder to solve the above-mentioned problems. In order to achieve the above purpose, the present invention adopts the following technical scheme: The invention discloses a tunnel detection vehicle positioning system based on an absolute value encoder, which comprises a ring number identification module, an absolute value encoder module, a slip detection module and a main control module, wherein the ring number identification module is used for identifying the ring number of a tunnel; The ring number identification module is used for acquiring ring number images on the subway tunnel, identifying the ring numbers and uploading the ring numbers to the main control module; The absolute value encoder module is used for collecting and uploading encoder readings to the main control module; The skid detection module is used for acquiring detected vehicle motion data and detected vehicle driving wheel motion data, detecting whether skid occurs at each moment by using a pre-trained skid detection model to obtain skid data, and transmitting the skid data to the main control module; the main control module is used for determining reference positioning information of the detection vehicle based on the position information corresponding to the ring number, determining accurate positioning information according to the reference positioning information and the reading of the encoder, and compensating the accurate positioning information according to the slipping data to obtain final positioning information. Further, the ring number identification module comprises an image acquisition unit, an image preprocessing unit and a ring number identification unit; The image acquisition unit adopts a color temperature adjustable light supplementing component and an HDR camera, is arranged at the position 1.2-1.5m in front of the head of the detection vehicle, and the camera lens faces the tunnel segment ring number area; the image preprocessing unit is used for carrying out graying, gaussian filtering denoising and Retinex image enhancement processing on the original ring number image acquired by the HDR camera, so that image interference caused by dust adhesion and uneven illumination is eliminated; the ring number recognition unit receives the preprocessed image, recognizes the ring number text information by using a built-in lightweight deep learning recognition model, and uploads the ring number text information to the main control module. Further, the lightweight deep learning recognition model is improved and trained based on YOLOv-Nano networks, and the improvement comprises: Optimizing YOLOv-Nano backbone network, replacing C2f module with C2f-Lite module, reducing Bottleneck number of each C2f-Lite module from 6 to 4, replacing standard convolution by depth separable convolution, and reducing backbone network parameter number; embedding ECA attention mechanism layers in FPN and PAN structures of YOLOv-Nano necks, carrying out global average pooling on feature grap