RSS ranging parameter calibration and measurement method based on visual positioning

Abstract

The invention discloses an RSS distance measurement parameter calibration and measurement method based on visual positioning, and the method comprises the steps: calculating a path loss parameter eta through a visual distance measurement module and a data processing module in combination with an RSS model, and obtaining an average value through integrating the measurement results of each time to optimize the path loss; the distance between the measurement points is calculated by estimating the signal intensity, it is guaranteed that the signal intensity of each measurement point in n times of measurement has enough difference, and it is guaranteed that the path loss value measured each time is more representative; and meanwhile, multiple rounds of measurement are carried out by adopting an iterative substitution method, threshold errors are compared, and the accuracy of a final result is ensured. The method is low in application cost, wide in application scene, small in error and convenient to operate.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a RSS ranging parameter calibration and measurement method based on visual positioning. Background Art

[0002] With the development of wireless local area networks, signal strength measurement is crucial for applications such as routing, data collection and analysis, and navigation. Accurately measured signal strength can help networks and related applications achieve higher work efficiency. At present, the ranging algorithms in wireless positioning technology can be roughly divided into two types: distance vector-based and non-distance vector-based. The ranging algorithm based on the distance vector has high measurement accuracy, but the hardware requirements are high. Among the ranging algorithms based on the distance vector, TOA (Time of Arrival), TDOA (Time Difference of Arrival), AOA (Angle of Arrival), RSSI (Received Signal Strength Indicator) and the like are all typical ranging technologies. Among them, RSSI ranging technology determines the distance between the signal point and the receiving point according to the strength of the received signal, and then performs positioning calculations based on the corresponding data. Thanks to the fact that the wireless signal strength value supports measurement on most wireless devices, it does not require additional hardware equipment. Therefore, although the ranging accuracy of RSSI ranging technology is lower than that of other technologies, it is still widely used and convenient. Therefore, RSSI ranging technology not only does not require the addition of additional hardware equipment, but also guarantees the measurement accuracy to the greatest extent.

[0003] In recent years, RSSI ranging technology has received extensive attention and research, but it also faces a key problem, that is, the impact of the actual environment on the propagation of wireless signals cannot be ignored. Due to different path loss model parameters in different actual environments, even the wireless signal strength values ​​of the same distance vector have significant differences, which ultimately makes the positioning result have a large error. Path loss, also known as propagation loss, refers to the loss caused by the propagation of radio waves in space. It is caused by the radiation diffusion of the transmission power and the propagation characteristics of the channel, and reflects the change in the mean value of the received signal power in a macroscopic range.

[0004] Due to the existence of path loss, the final measurement result will have a large error. Therefore, how to accurately measure the path loss model parameters in the ranging process has become a crucial issue in this technology. Only by knowing the accurate path loss parameters can the adjustment of parameters and paths in the subsequent measurement process be facilitated. However, the traditional manual measurement method cannot adapt well to the actual environment, and large measurements and calculations will also consume a lot of manpower and time. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide an RSS ranging parameter calibration and measurement method based on visual positioning, which adopts a binocular camera and a signal receiver, combines visual ranging with a neural network and an intelligent vehicle control system, and can effectively improve the ranging accuracy and the accuracy of the intelligent vehicle's travel direction. At the same time, iterative replacement and multiple rounds of measurement are adopted in the measurement process to estimate the operation of representative measurement points, thereby ensuring the reliability and accuracy of the desired path loss.

[0006] Technical solution: A method for calibrating and measuring RSS distance measurement parameters based on visual positioning, wherein the RSS distance measurement parameter calibration and measurement system includes a visual distance measurement module, a signal measurement module, a motion module and a data processing module; the visual distance measurement module uses a neural network model and combines a camera to identify a signal generating device, measures the distance to the signal generating device, and assists the motion module in correcting the accuracy of each movement; the signal measurement module measures the signal strength emitted by the current position signal generating device through a signal receiving device; the motion module combines the set number of times n and the measurement point distance d n The difference between the average path loss measured in the previous and next two rounds Control the movement and stop of the smart car; the data processing module is used for image processing and data calculation involved in the processing flow, and finally according to the reference distance d 0 and signal strength G(d 0 ) Calculate the distance d between each measuring point of wheel m mn and the path loss η at the measurement point mn Then, take the mean value; the steps are as follows:

[0007] S1, the visual ranging module installed on the smart car performs image semantic segmentation, target recognition and visual ranging;

[0008] S2, before starting the measurement, the speed of the smart car and the path loss error threshold T d , number of measurements per round n, reference distance d 0 , Baseline RSS change value G λ (Δd) and the typical value of path loss in this environment η ′ , sent to the data processing module;

[0009] S3, the camera in the visual ranging module recognizes and faces the target signal generating device through image semantic segmentation;

[0010] S4, specifies a certain direction with the signal device as the center, denoted as λ;

[0011] S5, move the smart car to a distance d from the signal device 0 At this point, measure the signal strength here as the reference distance signal strength G λ (d 0 );

[0012] S6, the first round of the first measurement begins. The data processing module calculates the path loss according to the RSS model and calculates the distance d between the first measurement point and the signal generating device. 11 ;

[0013] S7, move the smart car to d 11 At the first measurement point, the actual signal strength G is measured. λ (d 11 ), the data processing module calculates the path loss parameter η 11 and save;

[0014] S8, estimating signal strength according to the second measurement point Calculate the corresponding distance d of the second measurement point 12 ; According to the estimated signal strength Calculate the distance d corresponding to the third measurement point 13 , and so on, measure the path loss at the subsequent n-3 measurement points;

[0015] S9, after all n measurement points in the first round are measured, the data processing module averages the n path loss values ​​to obtain the average path loss measured in the first round

[0016] S10, in the second round of measurement, the path loss typical value η ′ Replaced by the average path loss of the first round Calculate the new measurement point position d 21 to d 2n ;

[0017] S11, repeat the operations of step S8 and step S9, and the data processing module calculates the average path loss parameter measured in the second round

[0018] S12, calculate the difference in average path loss between the previous and next rounds If its absolute value is less than the path loss error threshold T d , then end the movement and measurement; if it is greater than T d , then the next round of measurement is performed again, and steps S6 to S9 are repeated until the following equation is satisfied:

[0019]

[0020] Where m is the current measurement round number, and m-1 is the previous measurement round number; is the average path loss measured in the mth round;

[0021] The average path loss measured in this round at the end of the measurement The final measurement result.

[0022] Further, in step S4, a straight line is drawn between the λ direction and the signal device as a path, and the front of the smart car is adjusted to be in the same direction as the binocular camera to ensure that the car body is parallel to the λ direction.

[0023] Further, in step S6, the distance d between the first measuring point and the signal generating device is calculated. 11 , the expression is as follows:

[0024]

[0025] in, Estimate the signal strength value for the first measurement point, G λ (d 0 ) is the signal strength at the reference distance, G λ (Δd) is the baseline RSS change value; η ′ is the typical value of path loss in this environment.

[0026] Further, in step S8, the signal strength of each measurement point is estimated The expression is as follows:

[0027]

[0028] Where n is the number of measurements, G λ (d 0 ) is the reference distance signal strength.

[0029] Compared with the prior art, the present invention has the following significant effects:

[0030] 1. The present invention combines visual ranging, neural network and other technologies to ensure that the camera can accurately locate the target signal generating device and complete ranging, greatly improving the accuracy of ranging and measurement data; and the present invention has the advantages of low application cost, wide application scenarios, high accuracy, small error, and easy operation;

[0031] 2. In the present invention, the position of each measurement point is calculated using an estimated signal strength that is sufficiently different from the reference distance signal strength, ensuring that the signal strength between each measurement point is sufficiently different, avoiding the difference being too small to cause the path loss to be approximate and lack representativeness; at the same time, an iterative substitution method is used to perform multiple rounds of measurements and compare the path loss threshold error to ensure the accuracy of the final result. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the application scenario of the present invention;

[0033] Figure 2 The overall structure diagram of the RSS ranging parameter calibration and measurement system is shown in FIG.

[0034] Figure 3 A flowchart of the neural network training of the visual measurement module of the present invention;

[0035] Figure 4 It is a flow chart of the distance measurement operation of the visual measurement module of the present invention;

[0036] Figure 5 is an architecture diagram of the data processing module of the present invention;

[0037] Figure 6 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0039] The present invention provides a method for calibrating and measuring RSS ranging parameters based on visual positioning, and its application scenarios are as follows: Figure 1 The overall architecture of the RSS ranging parameter calibration and measurement system in the present invention is shown in FIG. Figure 2 As shown, it mainly includes a visual ranging module, a signal measurement module, a motion module and a data processing module. The visual ranging module includes a camera, a servo and a trained neural network model. When working, the servo rotates and the camera captures real-time images and stores them, and combines the neural network model to perform image semantic segmentation and target recognition. After the camera faces and recognizes the signal generating device, the servo stops rotating and measures the distance to the signal generating device, while assisting the motion module to correct the accuracy of each movement. The signal measurement module measures the signal strength emitted by the signal generating device at the current position through the signal receiving device. The motion module combines the set number of times n and the measurement point distance d n The difference between the average path loss measured in the previous and next two rounds Control the movement and stop of the smart car; the data processing module is used for image processing and data calculation involved in the process, and finally according to the reference distance d 0 and signal strength G(d 0 ) Calculate the distance d between each measuring point of wheel m mn and the path loss η at the measurement point mn Then, take the mean.

[0040] The key to the visual ranging module is to use the training of the neural network model to achieve image semantic segmentation, target recognition and ranging. The training process and ranging process are as follows: Figure 3 As shown, the implementation steps are as follows:

[0041] Step A1, collect typical signal generating equipment images as a data set and perform preprocessing (calibration, normalization and standardization, etc.);

[0042] Step A2: Build a convolutional neural network, including convolutional layers, pooling layers, and fully connected layers. The convolutional layer is used to extract image features, the pooling layer is used to reduce the dimension of the feature map, and the fully connected layer is used for the final classification.

[0043] Step A3: After compiling the neural network model and defining the loss function and evaluation indicators, split the data set into a training set and a validation set, perform model training and adjustment, and evaluate the model performance through the validation set.

[0044] like Figure 4 As shown in the figure, after loading the trained neural network model in the data processing module, the servo rotates when the car starts to move, and the camera takes real-time images and stores them in the data processing module. The data processing module extracts features from the image in combination with the neural network model, and converts the feature map into the original size through convolution and deconvolution operations for image semantic segmentation, thereby realizing target recognition, and then the image edge processing is used to detect the edge information of the object, that is, the outline of the object. The focal length is calculated using the image outline coordinates and width, and finally the distance between the camera and the target can be estimated through the focal length.

[0045] The main body of the signal measurement module is the signal receiving device (wireless network card, zigbee chip, etc.). When obtaining the signal strength, the electromagnetic waves measured by the signal receiver are converted into electrical signals. At the same time, its measurement time is ensured to match the time when the smart car moves to the measurement point to avoid missed measurements, false measurements, etc. (affecting data accuracy and continuity). The measured signal strength value is sent to the data processing module for storage for subsequent calculations.

[0046] The data processing module of the present invention is as follows Figure 5 As shown, it is used to process the real-time pictures taken by the camera in the visual ranging module, perform feature extraction, semantic segmentation and parallel calculation on the pictures, and then store the data in the memory for the central processor to perform convolution and focal length and distance calculation. At the same time, after receiving the signal strength and distance information measured by the signal measurement module and the visual ranging module, the central processor calculates the position information of each measurement point and the path loss value, and stores them in the memory for calculating the subsequent mean value; finally, the central processor cooperates with the motion module to compare the difference between the path loss mean value and the path loss error threshold value in each round of measurement to determine whether to continue the next round of movement.

[0047] like Figure 6 Shown is a general flow chart of the present invention, comprising the following steps:

[0048] Step 1: The visual ranging module installed on the smart car first uses the trained neural network model to combine the images taken by the camera to achieve semantic segmentation, target recognition and visual ranging, thereby achieving accurate recognition of the device;

[0049] The present invention uses the RSS model to calculate the path loss, and the formula is as follows:

[0050]

[0051] Among them, G λ (d) is the received signal strength at distance d in the λ direction, G λ (d 0 ) is the distance d in the λ direction 0 The received signal strength at d 0 is the reference distance and η is the path loss.

[0052] Step 2: Before starting the measurement, the artificially set smart car speed and path loss error threshold value T can be set as needed. d , number of measurements per round n, reference distance d 0 (generally 1 meter), benchmark RSS change value G λ (Δd) and the typical value of path loss in this environment η ′ Send to the data processing module.

[0053] Step 3: The camera of the smart car rotates the servo, and the image captured by the camera in real time is combined with the model to complete the image semantic segmentation until the target signal generating device is recognized and facing.

[0054] Step 4: Specify a direction with the signal device as the center, denoted as λ. A straight line is drawn between the λ direction and the signal device as the path. The car adjusts the front of the car to be consistent with the direction of the binocular camera, and the car body is parallel to the λ direction.

[0055] Step 5: Make the smart car move to the distance signal device d according to the set speed 0 The signal strength at this point is measured (generally 1 meter), which is used as the reference distance signal strength G λ (d 0 );

[0056] Step 6: Start the first round of the first measurement. The data processing module uses formula (2) to estimate the signal strength value at the first measurement point. Subtract the signal strength G at the reference distance λ (d 0 ) is equal to the benchmark RSS change value G λ (Δd), calculate the corresponding distance d 11 It is the distance between the first measurement point and the signal generating device, that is:

[0057]

[0058] The path loss calculated here is the typical value of the path loss in this environment η ′ .

[0059] Step 7: The smart car moves to d 11 At the first measurement point, the actual signal strength G is measured. λ (d 11 ), the data processing module calculates the path loss parameter η according to formula (3) 11 and save it.

[0060]

[0061] Step 8: Estimating the signal strength at the second measurement point Substitute into formula (2) and calculate the corresponding distance d of the second measurement point 12 ; The estimated signal strength at the third measurement point is G λ (d 0 )+G λ (Δd)-2G λ (Δd)+3G λ Substitute (Δd) into formula (2) to calculate the corresponding distance d of the third measurement point 13 ; And so on, measure the path loss at the next n-3 measurement points. Therefore, the estimated signal strength at each measurement point is And the number of measurements n, the benchmark RSS change value G λ (Δd) and reference distance signal strength G λ (d 0 ) are as follows:

[0062]

[0063] Step 9: After the first round of n measurement points are measured, η 1n The post-data processing module averages the n path loss values ​​to obtain the average path loss measured in the first round.

[0064] Step 10: In the second round of measurement, the path loss parameter is iteratively replaced when calculating the distance of n measurement points, that is, in the second round of measurement, the path loss typical value η in formula (2) is replaced by ′ Replaced by the average path loss of the first round Calculate the new measurement point position d 21 to d 2n .

[0065] Step 11, repeat the operations of steps 8 and 9, and the data processing module calculates the average path loss parameter measured in the second round

[0066] Step 12: Calculate the difference in average path loss between the previous and next rounds

[0067] If its absolute value is less than the path loss error threshold T d , then the movement and measurement are ended;

[0068] If greater than T d , then the next round of measurement is performed again, and the operations of steps 6 to 9 are repeated until formula (5) is satisfied:

[0069]

[0070] Where m is the current measurement round number, and m-1 is the previous measurement round number; is the average path loss measured in the mth round.

[0071] Step 13, the average path loss measured in this round at the end of the measurement The final measurement result.

[0072] Specifically, the present invention combines visual ranging with neural networks and smart car control systems to effectively improve ranging accuracy and the accuracy of the vehicle's travel direction. Among them, optimizing the neural network model and interactive recognition using typical device appearance training is the key to improving device recognition accuracy. The measurement results of the visual ranging method are less affected by the propagation medium and communication noise, and relatively accurate distance data can be obtained in various measurement situations. The neural network model is used to enhance the adaptability of the visual ranging system in multiple scenarios, improve measurement accuracy and reduce the possibility of manual debugging. The ranging data is fed back to the smart car control system to ensure the accuracy of each movement distance of the car during its travel.

[0073] It is worth noting that the estimated signal strength is used to infer the distance of the measurement point in order to ensure that the signal strength of each measurement point during n measurements is sufficiently different, so as to ensure that the path loss value measured each time is more representative; at the same time, the iterative substitution method is used to perform multiple rounds of measurements and compare the threshold error to ensure the accuracy of the final result.

[0074] In addition, when calculating the distance corresponding to the measuring point, the present invention combines formula (1) and formula (4). When the number of measurements n is an odd number, the estimated signal strength is greater than the reference distance signal strength, and the calculated measuring point distance is greater than the reference distance; when n is an even number, the estimated signal strength is less than the reference distance signal strength, and the calculated measuring point distance is less than the reference distance. Therefore, in fact, each linear movement of the car in the λ direction is alternating forward and backward.

[0075] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for calibrating and measuring RSS ranging parameters based on visual positioning, wherein: The RSS distance measurement parameter calibration and measurement system includes a visual distance measurement module, a signal measurement module, a motion module and a data processing module; the visual distance measurement module uses a neural network model and combines the camera to identify the signal generating device, measures the distance to the signal generating device, and assists the motion module in correcting the accuracy of each movement; the signal measurement module measures the signal strength emitted by the current position signal generating device through the signal receiving device; The motion module combines the setting times n and the measuring point distance d n The difference between the average path loss measured in the previous and next two rounds Control the movement and stop of the smart car; the data processing module is used for image processing and data calculation involved in the processing flow, and finally calculates the distance d of each measurement point of the m wheel according to the reference distance d0 and the signal strength G(d0) mn and the path loss η at the measurement point mn After that, take the average value; it is characterized by comprising the following steps: S1, the visual ranging module installed on the smart car performs image semantic segmentation, target recognition and visual ranging; S2, before starting the measurement, the speed of the smart car and the path loss error threshold T d , number of measurements per round n, reference distance d0, benchmark RSS change value G λ (Δd) and the typical value of path loss in this environment η ′ , sent to the data processing module; S3, the camera in the visual ranging module recognizes and faces the target signal generating device through image semantic segmentation; S4, specifies a certain direction with the signal device as the center, denoted as λ; S5, move the smart car to the distance signal device d0, measure the signal strength here as the reference distance signal strength G λ (d0); S6, the first round of the first measurement begins. The data processing module calculates the path loss according to the RSS model and calculates the distance d between the first measurement point and the signal generating device. 11 ; S7, move the smart car to d 11 At the first measurement point, the actual signal strength G is measured λ (d 11 ), the data processing module calculates the path loss parameter η 11 and save; S8, estimating signal strength according to the second measurement point Calculate the corresponding distance d of the second measurement point 12 ; According to the estimated signal strength Calculate the distance d corresponding to the third measurement point 13 , and so on, measure the path loss at the subsequent n-3 measurement points; S9, after all n measurement points in the first round are measured, the data processing module averages the n path loss values ​​to obtain the average path loss measured in the first round S10, in the second round of measurement, the path loss typical value η ′ Replaced by the average path loss of the first round Calculate the new measurement point position d 21 to d 2n ; S11, repeat the operations of step S8 and step S9, and the data processing module calculates the average path loss parameter measured in the second round S12, calculate the difference in average path loss between the previous and next rounds If its absolute value is less than the path loss error threshold T d , then end the movement and measurement; if it is greater than T d , then the next round of measurement is performed again, and steps S6 to S9 are repeated until the following equation is satisfied: Where m is the current measurement round number, and m-1 is the previous measurement round number; is the average path loss measured in the mth round; The average path loss measured in this round at the end of the measurement The final measurement result.

2. According to the RSS ranging parameter calibration and measurement method based on visual positioning according to claim 1, it is characterized in that: In step S4, a straight line is drawn between the λ direction and the signal device as a path, and the front of the smart car is adjusted to be in the same direction as the binocular camera to ensure that the car body is parallel to the λ direction.

3. The RSS ranging parameter calibration and measurement method based on visual positioning according to claim 1 is characterized in that: In step S6, the distance d between the first measuring point and the signal generating device is calculated. 11 , the expression is as follows: in, Estimate the signal strength value for the first measurement point, G λ (d0) is the signal strength at the reference distance, G λ (Δd) is the baseline RSS change value; η ′ is the typical value of path loss in this environment.

4. The RSS ranging parameter calibration and measurement method based on visual positioning according to claim 3 is characterized in that: In step S8, the signal strength of each measurement point is estimated The expression is as follows: Where n is the number of measurements, G λ (d0) is the reference distance signal strength.

Images