Method, device and apparatus for correcting positioning deviation

Abstract

The present application discloses a positioning offset correction method, which is applied to a device, wherein the device is driven by a navigation and positioning system, and the method includes: acquiring a target image, and identifying a target pattern of the target image through a recognition model; when the target pattern is identified, recording the first state and the first positioning position of the device; when the device returns to the first positioning position through the navigation and positioning system, recording the second state and the second positioning position; adjusting the device from the second state to the first state, and recording the third positioning position; determining the positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correcting the navigation route of the navigation and positioning system according to the positioning offset. In the above manner, the device recognizes the target pattern and adjusts its own state according to the target pattern. When the device state is equal to the initial state, if the positioning position of the navigation and positioning system is different, the positioning offset of the navigation and positioning system such as the Beidou navigation system and the GPS positioning system is automatically corrected, so that the trajectory is more accurate during automatic navigation.

Description

Technical Field

[0001] The present invention relates to the field of satellite navigation positioning, and in particular to a positioning offset correction method, device and equipment. Background Art

[0002] The BeiDou Navigation Satellite System (BDS), or Global Positioning System (GPS), is a commonly used navigation and positioning system that provides global location information such as longitude, latitude, and altitude. Both the BeiDou and GPS systems have demonstrated excellent performance in areas such as autonomous driving, vehicle monitoring, and precision agriculture. Due to the importance and widespread application of positioning functions, higher standards and requirements have been set for the precise positioning of the BeiDou and GPS systems.

[0003] Due to the delay effect of the atmospheric ionosphere, satellite clock error, satellite orbit error and the multipath effect caused by urban buildings, the positioning of the navigation and positioning system may be inaccurate, causing some user navigation trajectory points to deviate far from the actual position. For these offset points, their positions need to be corrected to make the navigation trajectory closer to the user's trajectory.

[0004] There are various methods for correcting positioning deviations in the existing technology, but most of them rely on setting up additional sensors to obtain more external information, and through multi-sensor combined positioning, the information obtained by the multiple sensors is used to calculate the deviation value through probability statistics and other methods to correct the navigation deviation. However, the above methods rely on sensor data acquisition. When the sensor sensitivity decreases, the positioning correction effect will deteriorate. Therefore, during the invention process, the inventors found it necessary to propose a positioning correction method to solve the positioning offset problem of the navigation positioning system during automatic navigation. Summary of the Invention

[0005] This application provides a positioning offset correction method, device, and apparatus. By using image recognition methods, the positioning offset of navigation and positioning systems such as the Beidou navigation system and the GPS positioning system is automatically corrected, thereby making the trajectory more accurate during automatic navigation.

[0006] The embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a positioning offset correction method, which is applied to a device that is driven by a navigation and positioning system. The method includes:

[0008] Acquire a target image, and identify a target pattern in the target image using a recognition model;

[0009] When the target pattern is recognized, recording a first state and a first positioning position of the device;

[0010] When the device returns to the first positioning position through the navigation and positioning system, recording the second state and the second positioning position;

[0011] Adjusting the device from the second state to the first state, and recording a third positioning position;

[0012] The positioning offset of the navigation and positioning system is determined according to the second positioning position and the third positioning position, and the navigation route of the navigation and positioning system is corrected according to the positioning offset.

[0013] In some embodiments, identifying a target pattern in a target image by using a recognition model includes:

[0014] Preprocess the target image;

[0015] Input the preprocessed target image into the recognition model;

[0016] The recognition model extracts features from the preprocessed target image through convolution and pooling to obtain first data;

[0017] The first data is input into the fully connected layer to determine whether the target image contains the target pattern.

[0018] In some embodiments, determining a positioning offset of a navigation positioning system according to the second positioning position and the third positioning position, and correcting a navigation route of the navigation positioning system according to the positioning offset includes:

[0019] The positioning offset of the navigation positioning system is determined according to the difference between the second positioning position and the third positioning position, the offset vector is determined according to the positioning offset, and the navigation route of the navigation positioning system is moved to a size and direction equal to the offset vector to correct the navigation route of the navigation positioning system.

[0020] In some embodiments, before acquiring the target image and identifying the target pattern in the target image using the recognition model, the method further includes:

[0021] obtaining samples of several target patterns;

[0022] Assign a corresponding label to each sample and mark the pixel position of the target pattern in each sample;

[0023] According to the deep learning algorithm, the recognition model is trained using samples of the target pattern and the label corresponding to each sample to obtain a trained recognition model.

[0024] In some embodiments, the first state includes a first distance and a first orientation between the device and the target pattern and a first posture of the device, and the first positioning position includes a first coordinate of the device under a navigation and positioning system.

[0025] In some embodiments, when the device returns to the first location through the navigation and positioning system, recording the second state and the second location includes:

[0026] When the device returns to the first positioning position through the navigation and positioning system, the target pattern is recognized again through the recognition model;

[0027] When the target pattern is recognized, the second state and the second positioning position are recorded. The second state includes the second distance and second orientation of the device from the target pattern and the second posture of the device. The second positioning position includes the second coordinates of the device under the navigation and positioning system.

[0028] In some embodiments, adjusting the device from the second state to the first state and recording the third location comprises:

[0029] adjusting the device from the second state to the first state so that the second distance, the second orientation, and the second posture of the device are the same as the first distance, the first orientation, and the first posture;

[0030] Record the third positioning position of the device under the navigation and positioning system, where the third positioning position includes the third coordinates of the device under the navigation and positioning system.

[0031] In a second aspect, an embodiment of the present invention provides a positioning offset correction device, which is applied to a device that travels via a navigation and positioning system. The positioning offset correction device includes:

[0032] A deep learning unit, configured to acquire a target image and identify a target pattern in the target image through a recognition model;

[0033] a first recording unit, configured to record a first state and a first positioning position of the device when a target pattern is recognized;

[0034] a second recording unit, configured to record the second state and the second positioning position when the device returns to the first positioning position through the navigation and positioning system;

[0035] a third recording unit, configured to adjust the device from the second state to the first state and record a third positioning position;

[0036] The correction unit is used to determine the positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correct the navigation route of the navigation and positioning system according to the positioning offset.

[0037] In a third aspect, an embodiment of the present invention provides a device, including:

[0038] at least one processor; and,

[0039] a memory communicatively connected to at least one processor; wherein,

[0040] The memory stores instructions that can be executed by at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform the positioning offset correction method according to the first aspect.

[0041] In a fourth aspect, an embodiment of the present invention provides a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a device to execute the positioning offset correction method as described in the first aspect.

[0042] The beneficial effects of the embodiments of the present invention are as follows: Different from the prior art, the embodiments of the present invention provide a positioning offset correction method, which is applied to a device that travels via a navigation and positioning system. The method includes: acquiring a target image and identifying a target pattern in the target image through a recognition model; when the target pattern is identified, recording a first state and a first positioning position of the device; when the device returns to the first positioning position through the navigation and positioning system, recording a second state and a second positioning position; adjusting the device from the second state to the first state, recording a third positioning position; determining a positioning offset of the navigation and positioning system based on the second and third positioning positions, and correcting the navigation route of the navigation and positioning system based on the positioning offset. The device recognizes the target pattern and adjusts its own state according to the target pattern. When the device state is equal to the initial state, if the positioning position of the navigation and positioning system is different, the positioning offset of navigation and positioning systems such as the Beidou navigation system and the GPS positioning system is automatically corrected, making the trajectory more accurate during automatic navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] One or more embodiments are exemplarily illustrated by pictures in the corresponding drawings. These exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0044] Figure 1 This is a schematic diagram of an application scenario of a positioning offset correction method provided by an embodiment of the present invention;

[0045] Figure 2 1 is a flow chart of a positioning offset correction method provided by an embodiment of the present invention;

[0046] Figure 3 1 is a flow chart of an image recognition process of a positioning offset correction method provided by an embodiment of the present invention;

[0047] Figure 4 1 is a schematic diagram of a training process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0048] Figure 5 This is another schematic diagram of a training process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0049] Figure 6 1 is a schematic diagram of a recognition process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0050] Figure 7 1 is a schematic diagram of a posture adjustment process of a positioning offset correction method provided by an embodiment of the present invention;

[0051] Figure 8 1 is a schematic diagram of a trajectory correction process of a positioning offset correction method provided by an embodiment of the present invention;

[0052] Figure 9 1 is a schematic structural diagram of a positioning offset correction device provided by an embodiment of the present invention;

[0053] Figure 10 It is a structural diagram of a device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0055] See also Figure 1 , Figure 1 This is a schematic diagram of an application scenario of a positioning offset correction method provided by an embodiment of the present invention;

[0056] like Figure 1 As shown, the application scenario includes a device 10 , a target pattern 20 and a satellite 30 .

[0057] The navigation and positioning system provides navigation and positioning services to the device 10 via the satellite 30 . The device 10 travels via the navigation and positioning system and can identify the target pattern 20 .

[0058] Among them, the device 10 can be any device that can travel along the route planned by the navigation and positioning system, such as an unmanned ship, an autonomous car, a drone, etc.; the target pattern 20 can be any pattern, such as a certain shape, a certain animal picture, a certain logo, etc., and the target pattern 20 can be fixed at a certain position for the device 10 to identify; the satellite 30 is a satellite of the navigation and positioning system, and the navigation and positioning system includes the Beidou navigation system and the GPS positioning system, etc. The navigation and positioning system obtains the positioning information of the device 10 through the satellite 30, such as the latitude and longitude of the device 10, and provides navigation services for the device 10.

[0059] See also Figure 2 , Figure 2 1 is a flow chart of a positioning offset correction method provided by an embodiment of the present invention;

[0060] like Figure 2 As shown, the positioning offset correction method is applied to a device that is driven by a navigation positioning system. The method includes:

[0061] Step S110: Acquire a target image and identify a target pattern in the target image using a recognition model;

[0062] For details, please refer to Figure 3 , Figure 3 1 is a flow chart of an image recognition process of a positioning offset correction method provided by an embodiment of the present invention;

[0063] like Figure 3 As shown, the image recognition process includes:

[0064] Step S111: acquiring a target image and preprocessing the target image;

[0065] Specifically, the device can obtain the target image through the camera, and preprocess the target image after obtaining it. The preprocessing methods include binarization and grayscale. Grayscale includes component method, maximum method, average method and weighted average method. For example, the component method takes the brightness of the three RGB components in the target image as the grayscale values ​​of the three grayscale images, and finally selects one of the three grayscale images according to application needs. Binarization includes global binarization, local binarization and local adaptive binarization. For example, global binarization sets a global threshold T and uses T to divide the image data into two parts: a pixel group greater than or equal to T and a pixel group less than T. The pixel value of the pixel group greater than or equal to T is set to white, that is, 255 (or black), and the pixel value of the pixel group less than T is set to black, that is, 0 (or white). The range of T is between [0,255].

[0066] Step S112: inputting the pre-processed target image into the recognition model;

[0067] Specifically, the recognition model is a deep learning model. After the pre-processed target image is input into the recognition model, the recognition model can determine whether the target image contains the target pattern. Therefore, before step S112, the recognition model needs to be trained.

[0068] Please refer to Figure 4 , Figure 4 1 is a schematic diagram of a training process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0069] like Figure 4 As shown in Figure 2, the training process of the recognition model includes:

[0070] Step S1121: obtaining samples of several target patterns;

[0071] Specifically, the target pattern is first selected, and then the image containing the target pattern is placed in different locations, and the target pattern is photographed from multiple directions and angles using a camera or other tool to obtain several samples of the target pattern. To ensure the training effect of the recognition model, the target pattern can also be placed in different backgrounds, different scenes, and different locations for photographing. Through a large number of different samples, the generalization ability of the recognition model is improved, so that the recognition model can still recognize the target pattern through the recognition model even in rainy, foggy, and other situations where the photographed image is blurry.

[0072] Step S1122: label each sample accordingly and mark the pixel position of the target pattern in each sample;

[0073] Specifically, after obtaining a sample of the target pattern, use an image annotation software tool (such as LabelStudio, CVAT, etc.) to frame the target pattern with a rectangle in the sample of the target pattern. For example, use four points (x0, y0), (x1, y1), (x2, y2), and (x3, y3) to mark the position of the target pattern in the sample. The matrix composed of these four pixel points is the pixel position of the target pattern in the sample. Then mark the type of the sample (for example, mark it as a cat, car, ship, etc.) and generate an XML file, which records the coordinate positions of the four vertices of the matrix.

[0074] Step S1123: Divide the samples into a test set, a training set, and a validation set;

[0075] Specifically, the samples are divided into a test set, a training set, and a validation set according to a certain ratio, for example, 50% is a test set, 25% is a training set, and the remaining 25% is a validation set.

[0076] Step S1124: According to the deep learning algorithm, the recognition model is trained using the training set samples of the target pattern and the label corresponding to each sample to obtain a trained recognition model.

[0077] Specifically, deep learning algorithms include convolutional neural networks (CNN), recurrent neural networks (RNN), recursive neural networks (RNN), etc. In the embodiment of the present invention, convolutional neural networks are preferred.

[0078] If the deep learning algorithm is a convolutional neural network, the convolutional neural network is trained using the test set samples and the labels corresponding to each sample. The weight matrix of the convolutional neural network is optimized through the output results of each sample to obtain the trained recognition model.

[0079] In an embodiment of the present invention, to ensure that the recognition accuracy of the recognition model is sufficiently high, a preset threshold (e.g., 95%) may be set. If the recognition accuracy of a sample reaches the preset threshold, it indicates that the trained recognition model's recognition capability meets expectations, and the trained recognition model is used as the trained recognition model to recognize the target image. Otherwise, the recognition model is trained again until the recognition accuracy of the recognition model reaches the preset threshold.

[0080] Furthermore, a convolutional neural network can include multiple convolutional layers, multiple pooling layers, and at least one fully connected layer. The convolutional layer extracts sample features by convolving filters (convolution kernels). The pooling layer compresses the features extracted by the convolutional layer to reduce the amount of data. Finally, the data is input into the fully connected layer for classification, resulting in a classification result. Initially, the weight parameters of the weight matrix can be randomly generated, and the weight matrix can be optimized during training through backpropagation.

[0081] Step S1125: Use the test set samples to test the recognition model and adjust the hyperparameters of the recognition model;

[0082] Specifically, in convolutional neural networks, hyperparameters include: the size of the convolution kernel, the size of the pooling kernel, the number of layers of the convolutional neural network, the loss function, etc. If the recognition accuracy of the test set samples is not ideal, the hyperparameters of the convolutional neural network are adjusted until the recognition accuracy of the recognition model reaches the preset threshold.

[0083] Step S1126: Using the validation set samples to evaluate the recognition model;

[0084] Use the recognition model optimized by the training set and test set to identify the validation set samples and calculate the recognition accuracy. If the recognition accuracy of the recognition model reaches the preset threshold (for example, 95%), it proves that its recognition effect meets the requirements and its generalization ability meets the standards, and it can be used by the device to identify the target pattern in the target image.

[0085] Figure 5 This is another schematic diagram of a training process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0086] like Figure 5 As shown, the training process includes:

[0087] Acquire a data set, which is a sample of a plurality of target patterns. The specific steps are the same as step S1121 and will not be repeated here.

[0088] Perform feature engineering, which is to label each sample with a corresponding label and mark the pixel position of the target pattern in each sample. The specific steps are the same as step S1122 and will not be repeated here;

[0089] The samples are divided into a test set, a training set, and a validation set. The specific steps are the same as step S1123 and will not be repeated here;

[0090] According to the deep learning algorithm, the recognition model is trained using the training set samples of the target pattern and the labels corresponding to each sample to obtain a trained recognition model;

[0091] Specifically, deep learning algorithms include convolutional neural networks (CNN), recurrent neural networks (RNN), recursive neural networks (RNN), etc. In the embodiment of the present invention, convolutional neural networks are preferred.

[0092] If the deep learning algorithm is a convolutional neural network, the convolutional neural network is trained using the test set samples and the labels corresponding to each sample. The weight matrix of the convolutional neural network is optimized through the output results of each sample, and the training is continuously iterated. After the iteration is completed, the recognition model is predicted using the test set to obtain the prediction result. If the prediction result of the model meets the requirements, that is, the recognition ability of the recognition model meets expectations, it can be verified using the validation set.

[0093] Furthermore, a preset threshold (for example, 95%) can be set. If the recognition accuracy predicted by the model reaches the preset threshold, it means that the recognition ability of the trained recognition model meets expectations and can be verified using the validation set. Otherwise, the recognition model is trained again until the recognition accuracy of the recognition model reaches the preset threshold. When the recognition model is tested using the test set samples, the hyperparameters of the recognition model can be adjusted. Specifically, in the convolutional neural network, the hyperparameters include: the size of the convolution kernel, the size of the pooling kernel, the number of layers of each layer of the convolutional neural network, the loss function, etc. If the recognition accuracy of the test set samples is not ideal, the hyperparameters of the convolutional neural network can be adjusted until the recognition accuracy of the recognition model reaches the preset threshold.

[0094] Finally, the recognition model is evaluated using the validation set samples. The validation set samples are recognized using the recognition model optimized by the training set and test set, and the recognition accuracy is calculated. If the recognition accuracy of the recognition model reaches the preset threshold (for example, 95%), it proves that its recognition effect meets the requirements and its generalization ability meets the standards, and it can be used by the device to recognize the target pattern in the target image.

[0095] Among them, the iteration of the convolutional neural network refers to updating the weight matrix. Initially, the various weight parameters of the weight matrix can be randomly generated, and the weight matrix can be optimized through back propagation during the training process.

[0096] In this embodiment of the present invention, if all training sets are trained once, it is called an Epoch (one generation of training); using a small part of the training set to perform a backpropagation on the model weights to update the parameters is called a Batch (a batch of data); the number of Batches required to complete an Epoch is called an Iteration (one training).

[0097] If you want to train 50,000 pictures, the number of pictures to be trained in each epoch is 50,000;

[0098] If the batch size is set to 256,

[0099] The training set has a batch size of 50000 / 256=195+1=196;

[0100] The number of batches to be completed in each epoch is 196;

[0101] Each Epoch has 196 Iterations.

[0102] The number of model weight updates sent in each epoch is 196;

[0103] After 10 generations of training, the number of model weight updates is 196*10;

[0104] Different generations of recognition models use the same training set. For example, the first and tenth generations use the same 50,000 images, but the weight matrices of the models are completely different.

[0105] Step S113: The recognition model extracts features from the preprocessed target image through convolution and pooling to obtain first data;

[0106] Since this step involves sending the collected target image into the trained recognition model and the process of automatic recognition by the recognition model is not much different from that of conventional convolutional neural networks, only a brief introduction is given.

[0107] Specifically, convolution is performed with different convolution kernels and the target image to extract different features of the target image, and then the features extracted by the convolution layer are compressed by the pooling method. The pooling methods include average pooling, maximum pooling, overlapping pooling, etc. Different pooling kernels and different pooling methods can be selected for pooling, and finally the first data is obtained. The first data includes n features of the target image. If each feature is represented by an i*j matrix, the size of the first data is i*j*n.

[0108] Step S114: inputting the first data into the fully connected layer to determine whether the target image contains the target pattern;

[0109] Since this step involves sending the collected target image into the trained recognition model and the process of automatic recognition by the recognition model is not much different from that of conventional convolutional neural networks, only a brief introduction is given.

[0110] Specifically, the first data is input into the fully connected layer, and the scores of the target image belonging to each category are calculated in the fully connected layer according to the weight matrix of the recognition model. Then, the probability of the target image belonging to each category is calculated according to the normalization method. If the probability of belonging to the target pattern is the largest, it is determined that the target image contains the target pattern, and the classification result is obtained.

[0111] Furthermore, open source deep learning software such as Yolo-fastest can be used to obtain several samples of the target pattern, label each sample accordingly, and mark the pixel position of the target pattern in each sample. Finally, the processed target pattern samples are input into software such as Yolo-fastest, which can automatically train and obtain a trained convolutional neural network model.

[0112] Figure 6 1 is a schematic diagram of a recognition process of a recognition model of a positioning offset correction method provided by an embodiment of the present invention;

[0113] like Figure 6As shown in the figure, the recognition model includes: an input layer (Input), two convolutional layers (conv1 and conv2), two pooling layers (pool1 and pool2), two fully connected layers (FC1 and FC2) and an output layer (Output).

[0114] The convolution layer convolves the sample with the filter (convolution kernel) to extract the features of the sample. The pooling layer compresses the features extracted by the convolution layer to reduce the amount of data. Finally, the data is input into the fully connected layer for classification to obtain the scores of the target image belonging to each category. The output layer normalizes the scores of the target image belonging to each category obtained by the fully connected layer to obtain the probability value of the target image belonging to each category. If the probability value of the target image containing the target pattern is the largest (for example, the probability of containing the target image is 90%, and the probability of belonging to other categories is 10%), it is determined that the target image contains the target pattern.

[0115] Step S120: When the target pattern is recognized, the first state and the first positioning position of the device are recorded;

[0116] Specifically, before the device automatically drives through the navigation and positioning system, it will first obtain a target image. When the target pattern is recognized in the target image, the first state and first positioning position of the device are recorded; wherein, the first state includes the distance and orientation of the device from the target image and the posture of the device (such as the device heading, pitch angle, etc.), and the first positioning position is the coordinate of the device under the navigation and positioning system, for example, the first positioning position is (116°E, 40°N).

[0117] Step S130: When the device returns to the first positioning position through the navigation and positioning system, the second state and the second positioning position are recorded;

[0118] Specifically, the device can automatically travel along the navigation route planned by the navigation and positioning system. When the device returns to the first positioning position according to the navigation route, the device acquires the target image again and recognizes the target pattern through the recognition model. When the target pattern is recognized, the second state and the second positioning position are recorded. The second state includes the second distance and the second orientation of the device from the target pattern and the second posture of the device. The second positioning position includes the second coordinates of the device under the navigation and positioning system. For example, the second coordinates are (116°E, 40°N). The first positioning position is equal to the second positioning position. However, when the navigation and positioning system produces a positioning deviation due to external factors, although the first positioning position of the device is different from the second positioning position, The two positioning positions are equal, but the geographical locations when the device records the first positioning position and the second positioning position are different. For example, the first positioning position and the second positioning position are both (116°E, 40°N). When the first positioning position is recorded, the device is 5 cm away from the target pattern, and the target pattern is directly north of the device. When the second positioning position is recorded, the device is 20 cm away from the target pattern, and the target pattern is northwest of the device. Due to the positioning deviation of the navigation and positioning system, the geographical location of the device is different when recording twice (that is, the first state and the second state are different), but the first positioning position and the second positioning position are the same. The difference between the first state and the second state of the device is the deviation of the navigation and positioning system.

[0119] After acquiring the target image, the device can preprocess it. Preprocessing methods include binarization and grayscale conversion, which have been described in the previous steps and will not be repeated here. The device can also automatically detect in real time whether the preprocessed target image contains the target image using open source deep learning software such as Yolo-fastest.

[0120] Step S140: adjusting the device from the second state to the first state, and recording the third positioning position;

[0121] Specifically, see Figure 7 , Figure 7 1 is a schematic diagram of a posture adjustment process of a positioning offset correction method provided by an embodiment of the present invention;

[0122] like Figure 7 As shown, the schematic diagram of the state adjustment process includes a device 51 and a target pattern 52. The first posture recorded by the device 51 is Z0, and the second posture recorded is Z1.

[0123] It is not difficult to see that the first posture and the second posture of the device are different. Therefore, through auxiliary equipment such as the 9-axis sensor, the second posture Z1 of the device is continuously adjusted until it is consistent with the first posture Z0. For example, if the device 51 is an unmanned ship, the heading is adjusted by rotating the device when the heading deviates, and the center of gravity is controlled to move forward / backward when the pitch angle deviates, so that the first posture Z0 is adjusted to be equal to the second posture Z1.

[0124] Furthermore, after Z1=Z0, it also includes adjusting the position of the device 51 from the target pattern 52. The position of the device 51 includes the distance and orientation of the device 51 from the target pattern 52. If the first position of the device 51 when recording the first posture Z0 is S0, and the second position when recording the second posture is S1, the second position S1 of the device is continuously adjusted through auxiliary equipment such as distance sensing sensors until it is the same as the first position S0. Among them, if there is a deviation in the distance between the device 51 and the target pattern 52, the device is moved forward or backward. If the orientation of the device 51 from the target pattern 52 is different, the device is moved to the orientation recorded in the first state to ensure that S1 is consistent with S0.

[0125] When the device's posture Z1 = Z0 and position S0 = S1, the device's second state is equal to the device's first state. At this time, the device 51 is already in the same geographical location as when the first state was recorded. However, if there is a positioning deviation in the navigation and positioning system, then when the geographical locations are the same, the coordinates under the navigation and positioning system will be unequal. Therefore, if there is a positioning deviation in the navigation and positioning system, after the device's second state = the first state, the recorded third positioning position is not equal to the first positioning position.

[0126] In one embodiment of the present invention, the device first uses GPS navigation to return to a rough initial position, and then adjusts to the precise initial position through target pattern recognition. Specifically, during the actual operation of the device, it first uses Beidou / GPS to return to the starting position S1 (the longitude and latitude of S0 and S1 are consistent in the Beidou / GPS system, but due to the offset, the actual geographical locations of S0 and S1 are different). For example, S0 is when the device is 5 cm directly opposite the target pattern. Due to the offset, S1 may be when the device is 10 cm directly opposite the target pattern.

[0127] Then, perform a state adjustment. This involves returning the device to a position 5 cm directly opposite the target pattern. During this process, the device's posture and location will be adjusted. Ensure that the adjusted location and posture are identical to the initial state.

[0128] Step S150: determining a positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correcting a navigation route of the navigation and positioning system according to the positioning offset;

[0129] Specifically, determining a positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correcting a navigation route of the navigation and positioning system according to the positioning offset includes:

[0130] The positioning offset of the navigation positioning system is determined according to the difference between the second positioning position and the third positioning position, the offset vector is determined according to the positioning offset, and the navigation route of the navigation positioning system is moved to a size and direction equal to the offset vector to correct the navigation route of the navigation positioning system.

[0131] Specifically, see Figure 8 , Figure 8 1 is a schematic diagram of a trajectory correction process of a positioning offset correction method provided by an embodiment of the present invention;

[0132] like Figure 8 As shown, the trajectory correction process includes the device 61 and the target pattern 62, the initial trajectory 63, the offset trajectory 64, the correction trajectory 65 and the offset 66. The correction trajectory 65 and the initial trajectory 63 are theoretically completely overlapped (but in practice there will always be errors, so the correction trajectory 65 and the initial trajectory 63 can only be basically overlapped, but Figure 8 The correction trajectory 65 and the initial trajectory 63 can be regarded as overlapping trajectories. The first position when the device records the first state is S0, the second position when the device records the second state is S1, and the third position when the device records the third state is S2. Figure 8 The device has been adjusted from the second state to the third state (the third state of the device is equal to the first state of the device).

[0133] After the status adjustment of the device is completed, the positioning offset between the two points is determined according to the second positioning position recorded at the second position S1 and the third positioning position recorded at the third position S2. For example, the second positioning position recorded at the second position S1 is (116°27′32"E, 40°N), and the third positioning position recorded at the third position S2 is (115°27′30"E, 40°N). The coordinate difference between the second positioning position and the third positioning position is (2"E, 0). The direction and size of the offset vector are determined according to the coordinate difference. The size of the offset vector needs to be converted into longitude and latitude and distance units. Finally, the trajectory planned by the navigation and positioning system is translated as a whole according to the offset vector, and the movement is equal to the size and direction of the offset vector to correct it. The navigation route of the navigation positioning system is to determine the offset vector based on the determined offset 66. The size of the offset vector is the distance obtained by converting the latitude and longitude and distance units of the second positioning position and the third positioning position (for example, 15m), and the direction is from the second position S1 in the second state to the third position S2 in the third state (for example, northwest). The offset trajectory 64 is moved a distance equal to the size and direction of the offset vector to obtain a correction trajectory 65. Because the first state of the device 61 is equal to the second state, the correction trajectory 65 coincides with the initial trajectory 63. Therefore, when the device 61 sets out again, its driving trajectory will be the same as the previous time. The positioning deviation of the navigation positioning system is corrected in the above manner to ensure that the device 61 travels along the desired route.

[0134] Furthermore, when the device moves from S1 to S2, its actual geographical location changes, and the positioning position of the navigation and positioning system will also change. However, this process is relatively short, and the positioning position of the navigation and positioning system can be ignored. Therefore, it is feasible to calculate the offset vector by latitude and longitude.

[0135] The device identifies the target pattern and adjusts its own state according to the target pattern. When the device state is equal to the initial state, if the positioning position of the navigation and positioning system is different, it automatically corrects the positioning offset of the Beidou navigation system, GPS positioning system and other navigation and positioning systems, making the trajectory more accurate during automatic navigation.

[0136] See also Figure 9 , Figure 9 1 is a schematic structural diagram of a positioning offset correction device provided by an embodiment of the present invention;

[0137] like Figure 9 As shown, the positioning offset correction device includes:

[0138] A deep learning unit 71 is configured to acquire a target image and identify a target pattern in the target image using a recognition model;

[0139] A first recording unit 72 is configured to record a first state and a first positioning position of the device when a target pattern is recognized;

[0140] A second recording unit 73 is configured to record the second state and the second positioning position when the device returns to the first positioning position through the navigation and positioning system;

[0141] a third recording unit 74, configured to adjust the device from the second state to the first state and record a third positioning position;

[0142] The correction unit 75 is configured to determine a positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correct a navigation route of the navigation and positioning system according to the positioning offset.

[0143] In this embodiment of the present invention, the deep learning unit 71 further includes:

[0144] Preprocess the target image;

[0145] Input the preprocessed target image into the recognition model;

[0146] The recognition model extracts features from the preprocessed target image through convolution and pooling to obtain first data;

[0147] The first data is input into the fully connected layer to determine whether the target image contains the target pattern.

[0148] In this embodiment of the present invention, the correction unit 75 further includes:

[0149] The positioning offset of the navigation positioning system is determined according to the difference between the second positioning position and the third positioning position, the offset vector is determined according to the positioning offset, and the navigation route of the navigation positioning system is moved to a size and direction equal to the offset vector to correct the navigation route of the navigation positioning system.

[0150] In an embodiment of the present invention, the positioning offset correction device further includes a training unit, which is specifically configured to:

[0151] obtaining samples of several target patterns;

[0152] Assign a corresponding label to each sample and mark the pixel position of the target pattern in each sample;

[0153] According to the deep learning algorithm, the recognition model is trained using samples of the target pattern and the label corresponding to each sample to obtain a trained recognition model.

[0154] In an embodiment of the present invention, the positioning offset correction device further includes an identification unit, which is specifically configured to:

[0155] When the device returns to the first positioning position through the navigation and positioning system, the target pattern is recognized again through the recognition model;

[0156] When the target pattern is recognized, the second state and the second positioning position are recorded. The second state includes the second distance and second orientation of the device from the target pattern and the second posture of the device. The second positioning position includes the second coordinates of the device under the navigation and positioning system.

[0157] In this embodiment of the present invention, the third recording unit 74 further includes:

[0158] adjusting the device from the second state to the first state so that the second distance, the second orientation, and the second posture of the device are the same as the first distance, the first orientation, and the first posture;

[0159] Record the third positioning position of the device under the navigation and positioning system, where the third positioning position includes the third coordinates of the device under the navigation and positioning system.

[0160] Since the device embodiment and the method embodiment are based on the same concept, the content of the device embodiment may refer to the method embodiment if the contents do not conflict with each other, and will not be repeated here.

[0161] The device identifies the target pattern and adjusts its own state according to the target pattern. When the device state is equal to the initial state, if the positioning position of the navigation and positioning system is different, the positioning offset of the navigation and positioning system such as the Beidou navigation system and the GPS positioning system will be automatically corrected to make the trajectory more accurate during automatic navigation.

[0162] See also Figure 10 , Figure 10 It is a structural diagram of a device provided by an embodiment of the present invention.

[0163] like Figure 10 As shown, the device 80 includes: one or more processors 81 and a memory 82, Figure 10 A processor 81 is taken as an example.

[0164] The processor 81 and the memory 82 may be connected via a bus or other means. Figure 10 The bus connection is taken as an example.

[0165] Processor 81 captures a target image and identifies a target pattern in the target image using a recognition model. When the target pattern is identified, the device records a first state and a first positioning position. When the device returns to the first positioning position via a navigation and positioning system, the device records a second state and a second positioning position. The device adjusts the device from the second state to the first state and records a third positioning position. The device determines a positioning offset for the navigation and positioning system based on the second and third positioning positions, and corrects the navigation route of the navigation and positioning system based on the positioning offset. The device recognizes the target pattern and adjusts its state based on the target pattern. When the device state is equal to the initial state, if the positioning position of the navigation and positioning system is different, the device automatically corrects the positioning offset of navigation and positioning systems such as the Beidou navigation system and the GPS positioning system, thereby ensuring a more accurate trajectory during automatic navigation.

[0166] Memory 82, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules for the positioning offset correction method in the embodiments of the present application. Processor 81 executes the non-volatile software programs, instructions, and modules stored in memory 82 to execute various functional applications and data processing of the device, thereby implementing the positioning offset correction method in the above-described method embodiment.

[0167] The memory 82 may include a program storage area and a data storage area. The program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the device. In addition, the memory 82 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state memory device. In some embodiments, the memory 82 may optionally include a memory remotely located relative to the processor 81, and these remote memories may be connected to the controller via a network. Examples of the aforementioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0168] One or more modules are stored in the memory 82, and when executed by one or more processors 81, perform the positioning offset correction method in any of the above method embodiments, for example, perform the above described Figure 2 Method steps S110 to S150.

[0169] It should be noted that the above-mentioned product can execute the method provided in the embodiment of this application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of this application.

[0170] The embodiment of the present application provides a non-volatile computer-readable storage medium, which stores computer-executable instructions. The computer-executable instructions are executed by one or more processors, such as Figure 10 A processor 81 in the embodiment can enable the one or more processors to execute the positioning offset correction method in any of the above method embodiments, and execute the above described Figure 2 Method steps S110 to S150.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Under the concept of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the present invention as described above. For the sake of simplicity, they are not provided in detail. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention. Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, or of course by hardware. Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by computer programs that instruct the relevant hardware to complete the program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

Claims

1. A positioning offset correction method, characterized in that: Applied to a device, the device being driven by a navigation and positioning system, the method comprising: Acquire a target image, and identify a target pattern in the target image using a recognition model; When the target pattern is recognized, a first state and a first positioning position of the device are recorded, wherein the first state includes a first distance and a first orientation of the device from the target pattern and a first posture of the device; When the device returns to the first positioning position through the navigation and positioning system, recording a second state and a second positioning position of the device, wherein the second state includes a second distance and a second orientation of the device from the target pattern and a second posture of the device; adjusting the device from the second state to the first state, and recording a third positioning position of the device; A positioning offset of the navigation and positioning system is determined according to the second positioning position and the third positioning position, and a navigation route of the navigation and positioning system is corrected according to the positioning offset.

2. The method according to claim 1, characterized in that The identifying the target pattern of the target image by using a recognition model includes: Preprocessing the target image; Inputting the preprocessed target image into the recognition model; The recognition model extracts features from the preprocessed target image by convolution and pooling to obtain first data; The first data is input into a fully connected layer to determine whether the target image contains the target pattern.

3. The method according to claim 1, characterized in that Determining a positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correcting a navigation route of the navigation and positioning system according to the positioning offset, includes: The positioning offset of the navigation and positioning system is determined according to the difference between the second positioning position and the third positioning position, an offset vector is determined according to the positioning offset, and the navigation route of the navigation and positioning system is moved to a size and direction equal to the offset vector to correct the navigation route of the navigation and positioning system.

4. The method according to claim 1, wherein Before acquiring the target image and identifying the target pattern in the target image by using the recognition model, the method further includes: obtaining a plurality of samples of the target pattern; Label each sample accordingly, and mark the pixel position of the target pattern in each sample; According to a deep learning algorithm, the recognition model is trained using samples of the target pattern and labels corresponding to each of the samples to obtain the trained recognition model.

5. The method according to any one of claims 1 to 4, characterized in that The first positioning position includes the first coordinates of the device under the navigation positioning system.

6. The method according to claim 5, characterized in that When the device returns to the first positioning position through the navigation and positioning system, recording the second state and the second positioning position includes: When the device returns to the first positioning position through the navigation and positioning system, recognizing the target pattern again through the recognition model; When the target pattern is recognized, the second state and the second positioning position are recorded, where the second positioning position includes the second coordinates of the device under the navigation and positioning system.

7. The method according to claim 6, characterized in that The step of adjusting the device from the second state to the first state and recording a third positioning position includes: adjusting the device from the second state to the first state so that the second distance, the second orientation, and the second posture of the device are the same as the first distance, the first orientation, and the first posture; A third positioning position of the device under the navigation and positioning system is recorded, where the third positioning position includes a third coordinate of the device under the navigation and positioning system.

8. A positioning offset correction device, characterized in that: Applied to equipment that travels via a navigation and positioning system, the positioning offset correction device comprises: A deep learning unit, configured to acquire a target image and identify a target pattern in the target image using a recognition model; a first recording unit, configured to record a first state and a first positioning position of the device when the target pattern is recognized, wherein the first state includes a first distance and a first orientation of the device from the target pattern and a first posture of the device; a second recording unit, configured to record a second state and a second positioning position of the device when the device returns to the first positioning position through the navigation and positioning system, wherein the second state includes a second distance and a second orientation of the device from the target pattern and a second posture of the device; a third recording unit, configured to adjust the device from the second state to the first state and record a third positioning position of the device; A correction unit is used to determine a positioning offset of the navigation and positioning system according to the second positioning position and the third positioning position, and correct a navigation route of the navigation and positioning system according to the positioning offset.

9. A device, characterized in that include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the positioning offset correction method according to any one of claims 1 to 7.

10. A non-volatile computer-readable storage medium, characterized in that: The non-volatile computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable the device to execute the positioning offset correction method according to any one of claims 1 to 7.

Images