Method for realizing enhanced high-precision navigation of mobile phone by using recorder

Abstract

The invention belongs to the technical field of vehicle-mounted navigation, and particularly relates to a method for realizing enhanced high-precision navigation of a mobile phone by using a recorder, which comprises the following steps of: searching equipment in a set range by mobile phone navigation software, establishing connection, and determining a data transmission protocol between related connection parts after the connection is established; the video acquisition end compresses and encodes acquired video data and then transmits a compressed and encoded video stream to the mobile phone end; mobile phone navigation software adopts a neural network deep learning visual model algorithm to carry out lane line detection and analysis; and the mobile phone navigation software performs lane-level navigation route planning and more detailed voice broadcasting by a navigation engine according to the acquired information. According to the technical scheme, the vehicle-mounted automobile data recorder is used as a camera of mobile phone navigation software, and a visual algorithm is combined with mobile phone positioning information, so that the accuracy of lane-level navigation during mobile phone navigation is improved, and the accuracy of voice broadcast during mobile phone navigation is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of vehicle-mounted navigation, and in particular relates to a method for realizing enhanced high-precision navigation of a mobile phone by utilizing a recorder. Background Art

[0002] Current ordinary mobile phone navigation can only use the Beidou or GPS system built into the mobile phone for positioning, and cannot achieve high-precision navigation.

[0003] Some high-end mobile phones also support Beidou dual-frequency positioning, and even support Beidou dual-frequency + GPS multi-mode positioning. Amap or Baidu Maps have also launched lane-level navigation services for these high-end mobile phones. However, due to the small size and thinness of mobile phones, the performance of dual-frequency Beidou and GPS multi-mode antennas is far behind that of professional positioning equipment. In addition, mobile phone navigation is generally fixed in the car with a mobile phone holder. The film on the window and the electronic equipment in the car will cause attenuation or interference to the mobile phone receiving Beidou and GPS satellite signals, thereby reducing the positioning ability of the mobile phone. At the same time, when driving, due to changes in the road environment, the surrounding high-rise buildings will block or reflect Beidou and GPS satellite signals, which will further reduce the positioning ability of the mobile phone. Limited by the above factors, the lane-level navigation currently implemented on high-end mobile phones is based on the real-time trajectory of the mobile phone positioning and the positioning information of the surrounding vehicles, and the lane information position estimated by the big data model (pseudo lane-level navigation) is not 100% accurate. Reflect the real lane information position of the vehicle currently navigated by the mobile phone. It is impossible to achieve true lane-level navigation. Therefore, in the current mobile phone navigation, when encountering complex intersections, since the map does not know which lane the vehicle being navigated by the mobile phone is currently in, it is unable to provide more detailed road planning and more detailed voice broadcasts. As a result, the current use of high-end mobile phone navigation also often results in people taking the wrong road at complex intersections.

[0004] The problem is even more serious when using ordinary mobile phones that do not support Beidou dual-frequency or multi-mode positioning capabilities for navigation. Not only does it not support pseudo-lane-level navigation, but the mobile phone's poor positioning capability and the influence of the in-car environment and external high-rise buildings often lead to positioning errors during the navigation process (the vehicle using mobile navigation is positioned in the opposite lane, or offset to other roads), causing the map to re-plan the wrong navigation route, seriously affecting the user experience of mobile navigation and affecting normal use. Summary of the invention

[0005] The purpose of the present invention is to provide a method for realizing enhanced high-precision navigation of a mobile phone by using a recorder. The driving recorder in the car is used to transmit the real-time video stream collected by the driving recorder to the mobile phone through the WIFI direct technology. After the mobile phone map obtains the real-time video stream of the driving recorder, the current lane is identified in real time through the neural network deep learning visual processing model. The map navigation engine combines the mobile phone positioning information and the identified current lane information to make a more detailed navigation route planning, so as to solve the problem in the above-mentioned background technology that the mobile phone has weak positioning ability in the car, is easily affected by the environment inside and outside the car, and the positioning is inaccurate, and high-precision mobile phone navigation cannot be realized.

[0006] To achieve the above objectives, this application is implemented through the following technical solutions:

[0007] A method for realizing enhanced high-precision navigation of a mobile phone using a recorder comprises the following steps:

[0008] S1. The mobile navigation software searches for Wi-Fi Direct devices within the set range through the Wi-Fi P2P API and establishes a connection;

[0009] S2, after establishing a connection with the Wi-Fi Direct device, determining a data transmission protocol between related connection parts;

[0010] S3, the video acquisition end compresses and encodes the collected video data to obtain a sub-video stream, and transmits the sub-video stream to the mobile phone end based on the data transmission protocol;

[0011] S4. The mobile navigation software uses a neural network deep learning visual model algorithm to detect and analyze the lane line of the input sub-video stream to obtain the lane line information of the vehicle's current position;

[0012] S5. The mobile phone navigation software performs lane-level navigation route planning and more detailed voice broadcasting by the navigation engine of the mobile phone navigation software according to the lane line information of the current position of the vehicle obtained in step S4.

[0013] Furthermore, in step S1, the Wi-Fi Direct devices within the set range include at least a driving recorder.

[0014] Furthermore, in step S1, the mobile navigation software searches for Wi-Fi Direct devices within a set range through the Wi-Fi P2P API, and the specific steps for establishing a connection are:

[0015] S11. The mobile navigation software uses the Wi-Fi P2P API to actively discover or passively discover Wi-Fi Direct devices within a set range and establish a connection.

[0016] S12. After the connection is established, the connected device selects the Group Owner;

[0017] S13, Group Owner automatically assigns IP addresses to form a network connection;

[0018] S14. After the network is formed, the mobile phone directly communicates with the Wi-Fi Direct device through the assigned IP address.

[0019] Furthermore, in step S2, after the Wi-Fi Direct connection is established, data transmission adopts the UDP protocol, and a sequence number is added to each data packet in the UDP transmission protocol.

[0020] Further, in step S3, the video acquisition end compresses and encodes the collected video data to obtain a sub-video stream, and the specific steps of transmitting the sub-video stream to the mobile phone end based on the data transmission protocol are as follows:

[0021] S31, the video acquisition end uses the original video stream for the video data stored in the T card of the video acquisition end, and uses the sub-video stream after compression and encoding for the video data to be transmitted to the mobile phone end;

[0022] S32, using the H.265 video encoding format for the sub-video stream to be transmitted to the mobile phone;

[0023] S33, splitting the sub-video stream to be transmitted to the mobile phone into small data packets of a set value for transmission.

[0024] Furthermore, a delay optimization technical solution is also included, and the specific steps are:

[0025] S34, frame rate control, controlling the frame rate of the sub-stream transmitted to the mobile phone to be lower than the frame rate of the main stream transmitted to the T card;

[0026] S35, resolution control, controlling the video resolution of the sub-stream transmitted to the mobile phone to be lower than the video resolution of the main stream transmitted to the T card;

[0027] S36, through QoS control, priority is given to ensuring the bandwidth of sub-stream video data transmission;

[0028] S37. On the mobile phone side, transmission and decoding of sub-stream data are performed in parallel.

[0029] Furthermore, in step S4, the mobile phone navigation software uses a neural network deep learning visual model algorithm to detect and analyze lane lines on the input sub-video stream, and the specific steps of obtaining lane line information of the current position of the vehicle are as follows:

[0030] S41, performing image decoding, image distortion correction, and image ROI extraction on the video data of the sub-video stream to obtain the pre-processed video data of the sub-video stream;

[0031] S42, inputting the preprocessed video data of the sub-video stream into the LaneNet or YOLO-Lane model to extract lane line features and detect lane line coordinates, and calculating the offset of the vehicle relative to the lane center;

[0032] S43. Obtain lane line information of the vehicle's current position according to the vehicle offset information in step S42.

[0033] Furthermore, in step S41, firstly, the video data of the sub-video stream after the image decoding is subjected to image distortion correction, and the ROI of the image after the distortion correction is cropped to select the part of the image close to the ground;

[0034] The distortion correction model formula is as follows:

[0035] , After centralization, , where x and y are the original horizontal and vertical coordinates of the pixel in the image, respectively, and r is the radial distance from the point to the center of the image: , k1, k2, k3 are radial distortion coefficients, xoriginal and yoriginal are the coordinates of the origin in the image, and xcorrected and ycorrected are the corrected pixel coordinates.

[0036] Furthermore, in step S42, calculating the offset of the vehicle relative to the lane center includes the following steps:

[0037] S421. Detect lane line coordinates B by using the YOLO-Lane Bounding Box method:

[0038] Detection box form: B = {x, y, w, h, c}, where x and y are the coordinates of the center point of the detection box, w and h are the width and height of the detection box, and c is the category confidence;

[0039] S422. Calculate the offset Δd of the center of the vehicle relative to the center of the lane according to the lane line coordinates. The offset formula is: Δd=dcar-dlane center, where dcar is the horizontal coordinate of the center of the vehicle in the image, and dlane center is the horizontal coordinate of the midpoint of the left and right lane lines.

[0040] Furthermore, step S43 also includes performing sliding average on the lane line coordinate detection result in step S42 to smooth the lane line, and smoothing the lane line includes the following steps:

[0041] S431. Use a sliding average formula to calculate a smooth lane line. The sliding average formula is:

[0042] xt=α·xt + (1-α)·xt-1, where xt is the coordinate of the center point of the detection box at time t, and α is the smoothing factor, with a typical value of 0.6-0.8;

[0043] S432: Based on the obtained smooth lane line, the Kalman filter is used to smooth the position estimate.

[0044] The state equation is: k+1 =F k X k +W k ,

[0045] The observation equation is: Z k =H k X k +V k , where F k is the state transfer matrix, H k is the observation matrix, X k is the system state, Z k is the observed value, W k and V k is the noise term.

[0046] The beneficial effects of the present invention are:

[0047] The use of the patented technology of this invention can not only enable high-end mobile phones to achieve true lane-level navigation, but also enable ordinary mobile phones that do not support Beidou dual-frequency and multi-mode positioning technology to achieve lane-level navigation, greatly improving the user experience of mobile phone navigation, and solving the problem of positioning errors caused by the weak positioning ability of the mobile phone when using mobile phone navigation, which is easily affected by external environment interference and leads to wrong positioning, resulting in wrong route planning by the mobile phone navigation software and taking the wrong road. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The present invention is a block diagram of using a recorder to implement enhanced high-precision navigation on a mobile phone. DETAILED DESCRIPTION

[0049] The technical solution of the present invention is described in detail below through examples. The following examples are merely exemplary and can only be used to explain and illustrate the technical solution of the present invention, but cannot be construed as limiting the technical solution of the present invention.

[0050] This technical solution uses the driving recorder that is already installed on most vehicles to equip the mobile phone navigation software with a dedicated camera for driving, allowing the mobile phone navigation software to make truly high-precision lane-level navigation planning and more detailed voice broadcasts through the camera image (vision) combined with the mobile phone positioning information.

[0051] At present, vehicles are generally equipped with driving recorders, which record the driving images in front of the car in real time while driving. At the same time, due to the timely update of map data and fast version upgrade of mobile navigation software, most car owners are still accustomed to using mobile navigation. This technical solution is to solve the problem that the mobile phone has weak positioning ability in the car, is easily affected by the environment inside and outside the car, and cannot achieve high-precision mobile navigation. Using the driving recorder in the car, the real-time video stream collected by the driving recorder is transmitted to the mobile phone through WIFI direct technology (while it does not affect the mobile phone to access the Internet through the 4G / 5G network, and does not affect the normal navigation function of the mobile phone). After the mobile map software obtains the real-time video stream of the driving recorder, it uses the neural network deep learning visual processing model to identify the current lane in real time (it can also be further combined with YOLO and SSD models for vehicle and pedestrian detection) (the traditional model recognition algorithm for using the camera on the car to identify the current lane is currently relatively mature, and this technical solution uses the neural network deep learning visual algorithm to accurately identify the current lane). The map navigation engine combines the phone's positioning information and the identified current lane information (and can also be further combined with surrounding vehicle and pedestrian information) to make a more detailed navigation route plan (for example, you need to exit the main road onto the secondary road 500 meters ahead. The navigation engine determines based on the current vehicle speed that it will take 20 seconds to reach the exit to the secondary road. The current vehicle is still in the lane farthest from the secondary road, and the traffic volume in the adjacent lanes is still relatively large. It will take a relatively long time to turn out. At this time, a very urgent and clear prompt is required (flashing alarm on the phone screen and detailed voice broadcast) to avoid missing the intersection to turn to the secondary road).

[0052] like Figure 1 As shown, the implementation steps of this technical solution are: establishment and management of Wi-Fi Direct---data transmission and optimization---video stream transmission optimization---delay optimization solution---security considerations---mobile navigation software uses a neural network deep learning model algorithm for lane line detection and analysis---the navigation engine of the mobile navigation software performs lane-level navigation route planning and more detailed voice broadcasting.

[0053] S1. The mobile navigation software searches for Wi-Fi Direct devices within a set range through the Wi-Fi P2P API and establishes a connection. In the technical solution of the present application, in order to ensure accurate navigation, the search range of the Wi-Fi P2P API is limited to the space of the vehicle. It is not recommended to expand the search range to neighboring vehicles, because if the search range is too large, the data transmission will be affected, resulting in reduced accuracy, and there will be certain interference with the data transmission.

[0054] The process of discovering and connecting Wi-Fi Direct devices works as follows:

[0055] When the mobile phone navigation software is started, the Wi-Fi Direct device can be discovered and connected automatically. Alternatively, when the mobile phone navigation software is started, the Wi-Fi Direct device can be discovered and connected automatically, but the user can make a choice through inquiry. In this way, the user can decide whether to start the high-precision navigation method of the present technical solution according to the conditions of the driving road.

[0056] In this application, the Wi-Fi Direct devices within the set range include at least a driving recorder. In this technical solution, the real-time road status of the vehicle during driving is collected by the camera of the driving recorder, and the collected video is transmitted to the mobile phone in a specific way.

[0057] S2, after establishing a connection with the Wi-Fi Direct device, determining a data transmission protocol between related connection parts;

[0058] In this application, the mobile navigation software searches for nearby Wi-Fi Direct devices through the Wi-Fi P2P API. There are two specific discovery methods, one is active device discovery and the other is passive device discovery. For active device discovery, the Wi-Fi P2P API broadcasts a discovery request and waits for responses from other devices; for passive device discovery, the Wi-Fi P2P API listens to discovery requests broadcast by nearby devices.

[0059] In this embodiment, a mobile phone and a driving recorder are taken as examples for description. After the mobile phone and the driving recorder are connected, the Group Owner (GO) is selected through role negotiation.

[0060] In this application, WPS (Wi-Fi Protected Setup) is used for security authentication, including two security authentication methods: one is PIN mode, that is, confirmation by PIN code, and the other is PBC (Push Button Configuration) mode, which triggers authentication by pressing a button.

[0061] After passing the security authentication, the Group Owner automatically assigns an IP address (usually in the 192.168.xx range) to form a network connection. After the network is formed, the mobile phone communicates directly with the Wi-Fi Direct device through the assigned IP address.

[0062] After the Wi-Fi Direct connection is established, data transmission can be achieved through the following protocols: UDP Socket. Since navigation has high requirements for real-time data and it is necessary to reduce the delay of real-time video transmission as much as possible, the UDP protocol is used in this embodiment. Since UDP cannot guarantee the order and integrity of data packets, it is necessary to add a sequence number to each data packet in the transmission protocol.

[0063] S3. The video acquisition end compresses and encodes the collected video data to obtain a sub-video stream, and transmits the compressed and encoded sub-video stream to the mobile phone end.

[0064] The amount of real-time video data collected by the camera of the driving recorder is very large, and most of these video data have no direct relationship with the technical solution of the present application. If all these video data are transmitted to the mobile phone, it will not only occupy a large bandwidth, but also increase the processing volume of the mobile phone. Therefore, in order to reduce the occupation of the video data network bandwidth and improve the data transmission efficiency, in the technical solution of the present application, the driving recorder adopts a dual-stream working mode, that is, the video data collected by the camera to be saved in the T card of the driving recorder adopts the original video stream, that is, the video collected by the camera of the driving recorder in the prior art is stored in the original video stream of the driving recorder, and the video stream to be transmitted to the mobile phone adopts the compressed sub-video stream. The specific steps are:

[0065] The H.265 (HEVC) video encoding format is used for the sub-video stream to be transmitted to the mobile phone. In the technical solution of this application, in the video images captured by the camera of the driving recorder, only the information about the lane has an effect on the navigation engine, while the other parts of the video have no effect on the navigation engine, so there is no need for data transmission. In addition, in order to reduce the bandwidth occupied when transmitting the sub-video stream data, the sub-video stream to be transmitted to the mobile phone is split into small data packets of a set value for transmission to avoid network congestion caused by excessive data packets. As for the size of the data packet, those skilled in the art can set it according to the actual situation.

[0066] This application also includes a delay optimization solution, including the following steps:

[0067] 1. In order not to affect the frame rate of the original video stream saved to the T card of the driving recorder, it is necessary to reduce the frame rate of the sub-video stream transmitted to the mobile phone, for example, from 30 FPS to 15-20 FPS to reduce the amount of data.

[0068] 2. Without affecting the original video resolution saved in the dash cam's T card, reduce the resolution of the sub-video stream transmitted to the mobile phone, for example, adjust from 1080p / 2K / 4K to 540p or 480p to ensure smooth transmission.

[0069] 3. Through QoS (Quality of Service) control, the bandwidth of sub-video data transmission is prioritized.

[0070] 4. In this technical solution, the delay of data processing is reduced by executing the transmission and decoding of the sub-video stream video data in parallel.

[0071] This application also includes data security processing. Wi-Fi Direct uses WPA2 encryption to ensure data security. It also ensures that the connected dashcam device is a trusted device. The device security verification is performed through other verification mechanisms in the PIN. In the data transmission layer (UDP Socket), SSL / TLS is used to encrypt data.

[0072] S4. Mobile navigation software uses a neural network deep learning visual model algorithm to detect and analyze lane lines. The specific steps are as follows:

[0073] S41, image preprocessing and model input, the sub-video stream video data is input into the LaneNet or YOLO-Lane model after image decoding, image distortion correction, and image ROI extraction. In this application, the YOLO-Lane model is used as an example for explanation:

[0074] The camera of the vehicle dashcam usually has fisheye or wide-angle distortion. First, the decoded sub-video stream data must be subjected to image distortion correction. The distortion correction model formula is as follows:

[0075] ,

[0076] Where x and y are the original horizontal and vertical coordinates (uncorrected coordinates) of the pixel in the image, respectively. These coordinates are relative coordinates based on the center of the image and are usually centered:

[0077] , where xoriginal and yoriginal are the coordinates of the origin in the image, xcenter and ycenter are the coordinates of the center of the image, and r is the radial distance from the point to the center of the image: , k1, k2, k3 are radial distortion coefficients, which are used to describe the degree of lens distortion. These parameters are obtained through camera calibration. xcorrected and ycorrected are the corrected pixel coordinates. Crop the ROI of the image after distortion correction and select the part of the image close to the ground (such as the lower half) to reduce the interference and calculation amount of irrelevant areas.

[0078] S42, input the preprocessed sub-video stream video data into the LaneNet or YOLO-Lane model to extract lane line features and detect lane line coordinates, and calculate the offset of the vehicle relative to the lane center. The specific steps are:

[0079] 1. Use the YOLO-Lane model for lane line detection and use the YOLO-Lane Bounding Box method: The detection box form is: B = {x, y, w, h, c}, where: x and y are the coordinates of the center point of the detection box, w and h are the width and height of the detection box, and c is the category confidence (the confidence of the lane line).

[0080] 2. After lane line detection, calculate the offset Δd of the vehicle center relative to the lane center. The offset formula is: Δd=dcar-dlane center, where dcar is the horizontal coordinate of the vehicle center in the image (usually the midpoint of the image width), and dlane center is the horizontal coordinate of the midpoint of the left and right lane lines.

[0081] S43, post-processing and dynamic adjustment, is achieved by smoothing lane lines and vehicle offset information, specifically the following steps:

[0082] 1. In order to reduce detection fluctuations, the detection results can be calculated using a sliding average. The sliding average formula is: xt=α·xt + (1-α)·xt-1, where xt is the coordinate of the center point of the detection box at time t, and α is the smoothing factor, with a typical value of 0.6-0.8.

[0083] 2. Based on the obtained smooth lane line, use Kalman filter to smooth the position estimate. Kalman filter is used to smooth the lane line coordinates or vehicle position offset:

[0084] The state equation is: k+1 =F k X k +W k ,

[0085] The observation equation is: Z k =H k X k +V k , where X kis the system state (such as the lane center coordinate), Z k is the observed value (such as the current lane line coordinate), W k and V k is the noise term, F k is the state transfer matrix, which describes the evolution process from the current state to the next state. k is the observation matrix, which describes how to derive the observable quantities from the current system state.

[0086] S5. The mobile phone navigation software performs lane-level navigation route planning and more detailed voice broadcasting by the navigation engine according to the information obtained in step S4.

[0087] After the mobile navigation software obtains the lane information of the current navigation vehicle based on the neural network deep learning visual model, combined with the current mobile phone positioning information (which includes the current vehicle speed information), the navigation engine makes real lane-level driving guidance and provides detailed lane-level voice broadcast prompts (such as voice guidance to enter the turning lane in advance when turning is required to avoid illegal turns on the straight road; when selecting lanes at complex intersections, it determines whether the lane where the current vehicle is located is the lane the navigation engine expects the vehicle to travel in, and makes correct and detailed guidance in advance (such as voice prompts: change to the right 2 lanes to enter XX road), etc.).

[0088] This technical solution uses the vehicle-mounted driving recorder as the camera of the mobile phone navigation software, and uses visual algorithms combined with the mobile phone positioning information to improve the accuracy of lane-level navigation during mobile phone navigation and improve the accuracy of voice broadcast during mobile phone navigation.

[0089] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will be appreciated by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A method for realizing enhanced high-precision navigation of a mobile phone using a recorder, characterized in that: The following steps are involved: S1. The mobile navigation software searches for Wi-Fi Direct devices within the set range through the Wi-Fi P2P API and establishes a connection; S2, after establishing a connection with the Wi-Fi Direct device, determining a data transmission protocol between related connection parts; S3, the video acquisition end compresses and encodes the collected video data to obtain a sub-video stream, and transmits the sub-video stream to the mobile phone end based on the data transmission protocol; S4. The mobile navigation software uses a neural network deep learning visual model algorithm to detect and analyze the lane line of the input sub-video stream to obtain the lane line information of the vehicle's current position; S5. The mobile phone navigation software performs lane-level navigation route planning and more detailed voice broadcasting by the navigation engine of the mobile phone navigation software according to the lane line information of the current position of the vehicle obtained in step S4.

2. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 1, characterized in that: In step S1, the Wi-Fi Direct devices within the set range at least include a driving recorder.

3. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 1, characterized in that: In step S1, the mobile navigation software searches for Wi-Fi Direct devices within a set range through the Wi-Fi P2P API and establishes a connection in the following specific steps: S11. The mobile navigation software uses the Wi-Fi P2P API to actively discover or passively discover Wi-Fi Direct devices within a set range and establish a connection. S12. After the connection is established, the connected device selects the Group Owner; S13, Group Owner automatically assigns IP addresses to form a network connection; S14. After the network is formed, the mobile phone directly communicates with the Wi-Fi Direct device through the assigned IP address.

4. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 1, characterized in that: In step S2, after the Wi-Fi Direct connection is established, data transmission adopts the UDP protocol, and a sequence number is added to each data packet in the UDP transmission protocol.

5. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 1, characterized in that: In step S3, the video acquisition end compresses and encodes the collected video data to obtain a sub-video stream, and the specific steps of transmitting the sub-video stream to the mobile phone end based on the data transmission protocol are as follows: S31, the video acquisition end uses the original video stream for the video data stored in the T card of the video acquisition end, and uses the sub-video stream after compression and encoding for the video data to be transmitted to the mobile phone end; S32, using the H.265 video encoding format for the sub-video stream to be transmitted to the mobile phone; S33, splitting the sub-video stream to be transmitted to the mobile phone into small data packets of a set value for transmission.

6. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 5, characterized in that: It also includes a delay optimization technical solution, the specific steps are: S34, frame rate control, controlling the frame rate of the sub-stream transmitted to the mobile phone to be lower than the frame rate of the main stream transmitted to the T card; S35, resolution control, controlling the video resolution of the sub-stream transmitted to the mobile phone to be lower than the video resolution of the main stream transmitted to the T card; S36, through QoS control, priority is given to ensuring the bandwidth of sub-stream video data transmission; S37. On the mobile phone side, transmission and decoding of sub-stream data are performed in parallel.

7. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 1, characterized in that: In step S4, the mobile phone navigation software uses a neural network deep learning visual model algorithm to detect and analyze the lane line of the input sub-video stream. The specific steps of obtaining the lane line information of the current position of the vehicle are as follows: S41, performing image decoding, image distortion correction, and image ROI extraction on the video data of the sub-video stream to obtain the pre-processed video data of the sub-video stream; S42, inputting the preprocessed video data of the sub-video stream into the LaneNet or YOLO-Lane model to extract lane line features and detect lane line coordinates, and calculating the offset of the vehicle relative to the lane center; S43. Obtain lane line information of the vehicle's current position according to the vehicle offset information in step S42.

8. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 7, characterized in that: In step S41, firstly, the video data of the sub-video stream after the image decoding is subjected to image distortion correction, and the image ROI after the distortion correction is cropped to select the part of the image close to the ground; The distortion correction model formula is as follows: ， After centralization, , where x and y are the original horizontal and vertical coordinates of the pixel in the image, respectively, and r is the radial distance from the point to the center of the image: , k1, k2, k3 are radial distortion coefficients, xoriginal and yoriginal are the coordinates of the origin in the image, and xcorrected and ycorrected are the corrected pixel coordinates.

9. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 7, characterized in that: In step S42, calculating the offset of the vehicle relative to the lane center includes the following steps: S421. Detect lane line coordinates B by using the YOLO-Lane Bounding Box method: Detection box form: B = {x, y, w, h, c}, where x and y are the coordinates of the center point of the detection box, w and h are the width and height of the detection box, and c is the category confidence; S422. Calculate the offset Δd of the center of the vehicle relative to the center of the lane according to the lane line coordinates. The offset formula is: Δd=dcar-dlane center, where dcar is the horizontal coordinate of the center of the vehicle in the image, and dlane center is the horizontal coordinate of the midpoint of the left and right lane lines.

10. The method for realizing enhanced high-precision navigation of a mobile phone by using a recorder according to claim 9, characterized in that: Step S43 also includes performing sliding average on the lane line coordinate detection result in step S42 to smooth the lane line. Smoothing the lane line includes the following steps: S431. Use a sliding average formula to calculate a smooth lane line. The sliding average formula is: xt=α·xt + (1-α)·xt-1, where xt is the coordinate of the center point of the detection box at time t, and α is the smoothing factor, with a typical value of 0.6-0.8; S432: Based on the obtained smooth lane line, the Kalman filter is used to smooth the position estimate. The state equation is: k+1 =F k X k +W k , The observation equation is: Z k =H k X k +V k , where F k is the state transfer matrix, H k is the observation matrix, X k is the system state, Z k is the observed value, W k and V k is the noise term.

Images