A method and device for intelligent trajectory following vehicle based on image recognition

Abstract

The present application provides a method and device for intelligent trajectory following based on image recognition, which relates to the field of image recognition technology, including: receiving positioning timing information from a target vehicle; when the positioning timing information has a null time zone, obtaining the adjacent positions of the starting time and the adjacent positions of the end time of the null time zone; downloading road image information from a road network database; receiving road image information of the null time zone from the target vehicle; performing similarity analysis on the road image information and the null time zone road image information to obtain the null time zone embedded position timing information; after compensating the positioning timing information according to the null time zone embedded position timing information, the target vehicle motion trajectory is obtained. The present application can solve the technical problem in the prior art that the loss or inaccuracy of positioning information affects the acquisition of the real-time position of the vehicle and the accurate tracking of the motion trajectory, realize trajectory compensation and recovery during signal loss, and improve the technical effect of the continuity and accuracy of the vehicle trajectory.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a method and device for intelligent trajectory following based on image recognition. Background Art

[0002] With the development of intelligent transportation systems, vehicle positioning and path tracking technologies play an important role in improving traffic management, road safety, and navigation systems. Existing vehicle positioning systems are usually based on the Global Positioning System (GPS), Beidou Navigation System or other wireless communication technologies. They collect vehicle location information in real time, generate vehicle movement trajectories, and provide drivers with real-time navigation and traffic information. In addition, vehicle positioning technology is also widely used in autonomous driving, logistics transportation, smart parking and other fields. However, although these technologies have made significant progress in many aspects, they still face some technical difficulties in actual application, especially in the case of weak signals or signal loss, traditional positioning systems are difficult to deal with effectively.

[0003] At present, GPS and other positioning technologies have obvious blind spots in complex urban environments. For example, when a vehicle is traveling in a tunnel, underground garage, or urban block with dense high-rise buildings, the signal is easily blocked or reflected, resulting in loss or inaccurate positioning information. These situations usually cause the system to be unable to accurately obtain the real-time position of the vehicle, which in turn affects the normal operation of the navigation system and the tracking of the vehicle's movement trajectory. In addition, the emergency response mechanism of existing positioning technologies when the signal is lost is relatively weak, and there is a lack of effective compensation and recovery strategies. When the vehicle's positioning signal is interrupted, the existing technology cannot accurately infer the actual position of the vehicle, resulting in a long period of null value trajectory, which in turn affects the reliability and accuracy of the overall system.

[0004] To sum up, there are technical problems in the prior art that, due to the obstruction or reflection of positioning signals in complex urban environments, positioning information is lost or inaccurate, which further affects the acquisition of the vehicle's real-time position and the accurate tracking of the motion trajectory, resulting in the inability to accurately estimate the vehicle's position, thereby generating a long period of null-value trajectory, affecting the reliability and accuracy of trajectory following. Summary of the invention

[0005] The purpose of this application is to provide a method and device for intelligent trajectory following based on image recognition, so as to solve the technical problem in the prior art that due to the obstruction or reflection of the positioning signal in a complex urban environment, the positioning information is lost or inaccurate, which further affects the acquisition of the real-time position of the vehicle and the accurate tracking of the motion trajectory, resulting in the inability to accurately infer the vehicle position, thereby generating a long period of empty trajectory, affecting the reliability and accuracy of the trajectory following.

[0006] In view of the above problems, the present application provides a method and device for intelligent trajectory following based on image recognition.

[0007] In the first aspect, the present application provides a method for intelligent trajectory following based on image recognition, which is implemented by an intelligent trajectory following device based on image recognition, including: receiving positioning timing information from a target vehicle; when the positioning timing information has a null time zone, obtaining the adjacent positions of the starting time and the adjacent positions of the end time of the null time zone; downloading road image information from a road network database based on the adjacent positions of the starting time and the adjacent positions of the end time; receiving the road image information of the null time zone from the target vehicle; performing similarity analysis on the road image information and the road image information of the null time zone to obtain the null time zone embedded position timing information; after compensating the positioning timing information according to the null time zone embedded position timing information, obtaining the target vehicle movement trajectory.

[0008] In the second aspect, the present application also provides an image recognition-based intelligent trajectory following device for executing an image recognition-based intelligent trajectory following method as described in the first aspect, including: a positioning timing information receiving module, the positioning timing information receiving module is used to receive positioning timing information from a target vehicle; a null value time zone adjacent position obtaining module, the null value time zone adjacent position obtaining module is used to obtain the starting time adjacent position and the end time adjacent position of the null value time zone when the positioning timing information has a null value time zone; a road image information downloading module, the road image information downloading module is used to obtain the starting time adjacent position and the end time adjacent position of the null value time zone according to the starting time adjacent position A positioning module is provided adjacent to the end time, and the road image information is downloaded from the road network database; a null-value time zone road image information receiving module is used to receive the null-value time zone road image information from the target vehicle; a similarity analysis module is used to perform similarity analysis on the road image information and the null-value time zone road image information to obtain the null-value time zone embedded position timing information; a positioning timing information compensation module is used to compensate the positioning timing information according to the null-value time zone embedded position timing information to obtain the target vehicle motion trajectory.

[0009] The technical solution provided in the present application has at least the following technical effects or advantages: by receiving positioning timing information from a target vehicle; when the positioning timing information has a null time zone, obtaining the adjacent positions of the starting time and the adjacent positions of the end time of the null time zone; downloading road image information from a road network database according to the adjacent positions of the starting time and the adjacent positions of the end time; receiving the road image information of the null time zone from the target vehicle; performing similarity analysis on the road image information and the road image information of the null time zone to obtain the timing information of the embedded position of the null time zone; after compensating the positioning timing information according to the timing information of the embedded position of the null time zone, obtaining the motion trajectory of the target vehicle, that is, by realizing trajectory compensation and recovery during signal loss, the continuity and accuracy of the vehicle trajectory are improved, and the technical effect of accurately inferring and recovering the vehicle motion trajectory even in a signal blind spot is achieved.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0012] Figure 1 A flowchart of a method for intelligent trajectory following based on image recognition is provided for this application;

[0013] Figure 2 This is a structural schematic diagram of a device for intelligent trajectory following based on image recognition in the present application.

[0014] Explanation of the reference numerals: positioning timing information receiving module 11, null-value time zone adjacent position obtaining module 12, road image information downloading module 13, null-value time zone road image information receiving module 14, similarity analysis module 15, positioning timing information compensation module 16. DETAILED DESCRIPTION

[0015] This application provides a method and device for intelligent trajectory following based on image recognition, which solves the technical problem in the prior art that the blocking or reflection of the positioning signal in a complex urban environment causes the positioning information to be lost or inaccurate, further affecting the acquisition of the vehicle's real-time position and the accurate tracking of the motion trajectory, resulting in the inability to accurately infer the vehicle's position, thereby generating a long period of null value trajectory, affecting the reliability and accuracy of the trajectory following. The method realizes trajectory compensation and recovery during signal loss, improves the continuity and accuracy of the vehicle trajectory, and achieves the technical effect of accurately inferring and recovering the vehicle's motion trajectory even in the signal blind area.

[0016] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0017] For example, please refer to the attached Figure 1 The present application provides a method for intelligent trajectory following a vehicle based on image recognition, which is applied to a device for intelligent trajectory following a vehicle based on image recognition, and specifically includes:

[0018] Step 1: Receive positioning timing information from the target vehicle.

[0019] Specifically, the target vehicle is the vehicle to be followed. Through the positioning device on the vehicle, the position data of the target vehicle is obtained in real time, and the positioning timing information is received, including a continuous time series data stream formed by the latitude and longitude, speed, timestamp and other information of the vehicle. The positioning timing information represents the position change of the vehicle at different times, forming a time series data structure for subsequent analysis of the vehicle's movement trajectory. Through the positioning timing information, the vehicle's movement path can be tracked, and then trajectory analysis can be performed to help achieve real-time positioning of the vehicle.

[0020] Step 2: When the positioning timing information has a null time zone, obtain the adjacent positions of the start time and the adjacent positions of the end time of the null time zone.

[0021] Specifically, when the positioning time series information has a null time zone, it means that the positioning data of the target vehicle is missing for a period of time, and the specific location information cannot be obtained. In order to fill the null time zone, the coordinates of the adjacent positions of the starting time and the ending time of the null time zone are determined. The adjacent position at the starting time refers to the vehicle position at the time point before the null time zone, while the adjacent position at the ending time refers to the vehicle position at the time point after the null time zone, which is then used to infer the movement trajectory of the vehicle in the null time zone, so as to perform reasonable trajectory filling and inference.

[0022] Step 3: Downloading road image information from a road network database according to the adjacent positions at the starting time and the adjacent positions at the end time.

[0023] Specifically, the road image information of the adjacent positions at the start time and the adjacent positions at the end time is downloaded from the road network database. The road image information represents the road environment on the vehicle's driving path and is used to restore the specific road scenes that the vehicle passes through, such as streets, intersections, road signs, etc., to help conduct a more detailed analysis of the driving trajectory and traffic environment.

[0024] Step 4: Receive the null-time zone road image information from the target vehicle.

[0025] Specifically, road image information is received from the target vehicle's sensor or other data source when the target vehicle fails to provide valid positioning data within a specific time period. The null time zone refers to the situation where the target vehicle cannot obtain accurate location data within a certain period of time due to reasons such as signal loss and equipment failure. Therefore, the lack of location data is compensated by receiving road image information in the null time zone. The road image information in the null time zone includes environmental images around the vehicle, such as traffic signs, pedestrians or other obstacles on the road ahead. The road image information in the null time zone can be used to obtain the driving environment of the target vehicle in the null time zone, which helps to infer the driving trajectory of the target vehicle or judge the driving conditions during the null time period.

[0026] Step 5: Perform similarity analysis on the road image information and the road image information of the null-value time zone to obtain the null-value time zone embedding position time series information.

[0027] Specifically, similarity analysis is performed on road image information and road image information in the null time zone, which means that the similarity of the two types of image data needs to be compared to evaluate their consistency in content. Road image information is usually the image obtained by the vehicle when there is a positioning signal. The road image information in the null time zone refers to the image data obtained by the camera or other sensors during the period when the positioning signal is lost. The original positioning timing information may be blank or missing, resulting in the inability to accurately track the movement trajectory of the vehicle. The purpose of similarity analysis is to find the similarities between road image information and road image information in the null time zone, such as the shape of the road ahead, traffic signs, and the distribution of vehicles and pedestrians, and to confirm the visual similarity of the images by calculating the similarity. By comparing and matching the analysis results, the specific position and timing of the vehicle in the null time zone are determined, the embedded position timing information of the null time zone is obtained, and the driving trajectory of the vehicle in the null time period is inferred.

[0028] Step 6: After compensating the positioning timing information according to the embedded position timing information of the null time zone, the movement trajectory of the target vehicle is obtained.

[0029] Specifically, during the period when the vehicle positioning signal is lost, the null value time zone is embedded in the position time series information and inserted into the originally missing time series data to fill the originally missing positioning data, thereby restoring the complete positioning trajectory and obtaining the target vehicle motion trajectory. The target vehicle motion trajectory can fully and accurately display the vehicle's driving trajectory, and even in the case of signal loss, the vehicle's motion state can be restored.

[0030] The method of intelligent trajectory following based on image recognition is applied to an intelligent trajectory following device based on image recognition, which can realize trajectory compensation and recovery during signal loss, improve the continuity and accuracy of the vehicle trajectory, and achieve the technical effect of accurately inferring and recovering the vehicle movement trajectory even in the signal blind area.

[0031] Furthermore, the present application also includes: uploading the adjacent positions at the starting time and the adjacent positions at the end time to the road network database, downloading the first connected road image until the Nth connected road image; the first connected road image includes a left image in the first traveling direction, a right image in the first traveling direction, and a front image in the first traveling direction; until the Nth connected road image includes a left image in the Nth traveling direction, a right image in the Nth traveling direction, and a front image in the Nth traveling direction.

[0032] Specifically, the adjacent positions at the start time and the adjacent positions at the end time are uploaded to the road network database for subsequent analysis. The road network database contains information about roads, traffic, etc., and can update the road information passed by the vehicle through the uploaded vehicle positioning data.

[0033] Next, download the road image data related to vehicle travel from the road network database, including the first connected road image to the Nth connected road image. The road image data is matched with the road connection information in the road network and is used to visualize the road conditions passed by the vehicle. Wherein, N is an integer greater than or equal to 1. The first connected road to the Nth connected road are roads obtained by random extraction without replacement in the road network database.

[0034] The first connected road image includes a first traveling direction left image, a first traveling direction right image, and a first traveling direction front image, indicating that on the first connected road, image data in three different directions are acquired and provided. The first traveling direction left image shows the road environment on the left side of the vehicle's traveling direction, the first traveling direction right image shows the road environment on the right side of the vehicle's traveling direction, and the first traveling direction front image shows the road environment in front of the vehicle's traveling direction.

[0035] Until the Nth connected road image includes the Nth traveling direction left image, the Nth traveling direction right image and the Nth traveling direction front image, that is, when the vehicle travels to the Nth connected road, the road image related to the Nth connected road continues to be downloaded. Each road image includes views in three directions, thereby obtaining the driving route and environment of the target vehicle.

[0036] Furthermore, the present application also includes: the null time zone road image information includes a road image in front of the vehicle, a road image on the left side of the vehicle, and a road image on the right side of the vehicle.

[0037] Specifically, when null values ​​appear in the positioning time series information, the driving environment of the vehicle in the null time zone is supplemented or inferred by providing road images in different directions. The road image information in the null time zone includes the road image in front of the vehicle, the road image on the left side of the vehicle, and the road image on the right side of the vehicle. The road image in front of the vehicle shows the road environment in front of the vehicle, such as possible traffic lights and pedestrian crossings; the road image on the left side of the vehicle shows the road conditions on the left side of the vehicle, which may be roadside buildings or parking lots; and the road image on the right side of the vehicle shows the road environment on the right side, which may include sidewalks or other road facilities. The road image information in the null time zone plays a supplementary role in processing the null time zone, making the trajectory data more complete and accurate.

[0038] Furthermore, the present application also includes: obtaining a first null-value moment image of road image information in a null-value time zone; obtaining window size information of the first null-value moment image; randomly cutting the road image information according to the window size information to obtain a window cut image; comparing the similarity between the window cut image and the first null-value moment image, and when the similarity is greater than or equal to a similarity threshold, setting the acquisition position of the window cut image to the embedding position of the first null-value moment image, and adding the null-value time zone embedding position timing information.

[0039] Specifically, a first null-value moment image is obtained by randomly extracting from the null-value time zone road image information. The first null-value moment image is image information of any null-value time zone in the null-value time zone road image information.

[0040] The window size information of the image at the first null value moment is obtained. The window size information refers to the image size information, including the image width, height and other parameter information, which is used for subsequent image processing and analysis. The road image information is randomly selected and cut according to the window size information to obtain multiple window cut images corresponding to multiple image sizes.

[0041] Exemplarily, the window size information may also refer to the time length information of the time window. The road image information is randomly cut according to the time length to obtain a plurality of window cut images of the time length corresponding to the window size information.

[0042] Compare the similarity between the items in the window cut image and the items in the first null value moment image. When the similarity is greater than or equal to the similarity threshold, it means that the window cut image and the first null value moment image are highly similar. It can be considered that the window cut image comes from the same area of ​​the first null value moment image. Then, the acquisition position of the window cut image is set to the embedded position of the first null value moment image, and the embedded position timing information of the null value time zone is added, that is, updated to the position timing information of the null value time zone, and the trajectory data of the null value period is restored. Among them, the similarity threshold is obtained by a technician in this field according to the actual situation.

[0043] Furthermore, the present application also includes: extracting the first detection object type set, the first detection object size feature set and the first detection object distribution position set of the window cut image; extracting the second detection object type set, the second detection object size feature set and the second detection object distribution position set of the first null value moment image; based on the first detection object distribution position set and the second detection object distribution position set, analyzing the proportion of detection objects in the same position whose types are the same as those of the second detection object type set and whose sizes are the same as those of the first detection object size feature set and the second detection object size feature set, to obtain the similarity.

[0044] Specifically, the first detected object type set, the first detected object size feature set and the first detected object distribution position set of the window clipping image are extracted. The first detected object type set refers to a set of object types randomly identified in the image, for example, including "pedestrians" and "cars"; the first detected object size feature set is the size features of the objects corresponding to the first detected object type set, such as width and height; the first detected object distribution position set refers to the position of the objects corresponding to the first detected object type set in the image, which is represented by a coordinate system to help determine the spatial position of each object.

[0045] The image at the first null value moment may contain more than one type of object and may include multiple types of objects. A second detection object type set excluding the first detection object type set is randomly extracted from the image at the first null value moment, and then the second detection object size feature set and the second detection object distribution position set corresponding to the second detection object type set and the second detection object type set are obtained according to the method for obtaining the first detection object type set, the first detection object size feature set and the first detection object distribution position set.

[0046] Based on the first detection object distribution position set and the second detection object distribution position set, analyze the proportion of detection objects with the same type as the first detection object type set and the second detection object type set, and the same size as the first detection object size feature set and the second detection object size feature set, that is, check the items with the same type and the same size feature as the first detection object type set and the second detection object type set, and obtain the similarity by calculating the similarity value between the items with the same type and the same size feature and the first detection object type set and the second detection object type set.

[0047] Furthermore, the present application also includes: obtaining an object classification channel; inputting the window cut image into the object classification channel to obtain the first detected object type set, wherein the first detected object type set has the first detected object distribution position set in the window cut image; and calculating the first detected object size feature set of the first detected object type set based on the first detected object distribution position set.

[0048] Specifically, an object classification channel is obtained. The object classification channel refers to an algorithm or model used to classify images or data, and can identify different types of objects in an image through a trained neural network, machine learning model, or other algorithm. For example, the object classification channel can be used to identify different objects such as cars, pedestrians, and buildings in an image.

[0049] The window cut image is input into the object classification channel for analysis. The object classification channel processes the input window cut image, identifies the object type therein, and outputs a type set containing the identified objects to obtain a first detected object type set, wherein the first detected object type set has a first detected object distribution position set in the window cut image. The first detected object distribution position set refers to the position of the objects corresponding to the first detected object type set in the image, and is used to help obtain the specific coordinates of the objects in the image.

[0050] According to the first detected object distribution position set, the first detected object size feature set of the first detected object type set is calculated. The first detected object size feature set includes information such as the width, height, and aspect ratio of the first detected object in the image, which is used to quantify the size of the object. For example, if the first detected object type set is identified as a car, and the position of the first detected object type set in the image is detected to be a rectangular area, the width and height of the rectangular area are calculated, and the size feature set is output.

[0051] Furthermore, the present application also includes: when the target vehicle motion trajectory has an intersection with the prohibited driving area, the intersection trajectory is rendered and marked to obtain an updated target vehicle motion trajectory.

[0052] Specifically, the target vehicle motion trajectory refers to the driving path of the target vehicle depicted according to the positioning information within a certain period of time, and the prohibited driving area refers to the area that the vehicle cannot enter according to factors such as traffic rules and urban planning. When the target vehicle motion trajectory intersects with the prohibited driving area, it is necessary to check whether the target vehicle's driving route overlaps or intersects with the defined prohibited driving area. For example, the area beside a road is marked as a prohibited entry area due to construction. If the target vehicle's driving trajectory intersects with the prohibited driving area, it is identified that the target vehicle has driving behavior in the prohibited driving area. Then, the intersection trajectory is rendered and marked, and the part where the target vehicle intersects with the prohibited driving area is visually marked or highlighted. For example, the intersection trajectory is displayed by different colors, icons or other methods, so that this part of the path is significantly visible on the map or image, and then the updated target vehicle motion trajectory is obtained, and a new, updated target vehicle motion trajectory is generated, which contains all the information of the vehicle's driving, and also specifically marks the part that intersects with the prohibited driving area, so as to help the driver or management system clearly understand whether the vehicle violates the driving regulations, and provide a basis for subsequent traffic management or decision-making.

[0053] In summary, the method of intelligent trajectory following based on image recognition provided by the present application has the following technical effects: by receiving positioning timing information from the target vehicle; when the positioning timing information has a null time zone, obtaining the adjacent positions of the starting time and the adjacent positions of the end time of the null time zone; downloading road image information from the road network database according to the adjacent positions of the starting time and the adjacent positions of the end time; receiving the road image information of the null time zone from the target vehicle; performing similarity analysis on the road image information and the road image information of the null time zone to obtain the timing information of the embedded position of the null time zone; after compensating the positioning timing information according to the timing information of the embedded position of the null time zone, obtaining the motion trajectory of the target vehicle, that is, by realizing trajectory compensation and recovery during signal loss, the continuity and accuracy of the vehicle trajectory are improved, so as to achieve the technical effect of accurately inferring and recovering the vehicle motion trajectory even in the signal blind spot.

[0054] Embodiment 2: Based on the same invention concept as the method for following a vehicle with intelligent trajectory based on image recognition in the above embodiment, the present application also provides a device for following a vehicle with intelligent trajectory based on image recognition, please refer to the attached Figure 2 , including: a positioning timing information receiving module 11, the positioning timing information receiving module 11 is used to receive positioning timing information from a target vehicle; a null value time zone adjacent position obtaining module 12, the null value time zone adjacent position obtaining module 12 is used to obtain the starting time adjacent position and the end time adjacent position of the null value time zone when the positioning timing information has a null value time zone; a road image information downloading module 13, the road image information downloading module 13 is used to download road image information from a road network database according to the starting time adjacent position and the end time adjacent position; null value A time zone road image information receiving module 14, the null-value time zone road image information receiving module 14 is used to receive null-value time zone road image information from the target vehicle; a similarity analysis module 15, the similarity analysis module 15 is used to perform similarity analysis on the road image information and the null-value time zone road image information to obtain the null-value time zone embedded position timing information; a positioning timing information compensation module 16, the positioning timing information compensation module 16 is used to compensate the positioning timing information according to the null-value time zone embedded position timing information to obtain the target vehicle movement trajectory.

[0055] Furthermore, the device for intelligent trajectory following based on image recognition is also used to: upload the adjacent positions at the starting time and the adjacent positions at the end time to the road network database, download the first connected road image until the Nth connected road image; the first connected road image includes a left image in the first traveling direction, a right image in the first traveling direction, and a front image in the first traveling direction; until the Nth connected road image includes a left image in the Nth traveling direction, a right image in the Nth traveling direction, and a front image in the Nth traveling direction.

[0056] Furthermore, the intelligent trajectory following device based on image recognition is also used for: the road image information in the null time zone includes an image of the road in front of the vehicle, an image of the road on the left side of the vehicle, and an image of the road on the right side of the vehicle.

[0057] Furthermore, the image recognition-based intelligent trajectory following device is also used to: obtain a first null-value moment image of the null-value time zone road image information; obtain window size information of the first null-value moment image; randomly cut the road image information according to the window size information to obtain a window cut image; compare the similarity between the window cut image and the first null-value moment image, and when the similarity is greater than or equal to a similarity threshold, set the acquisition position of the window cut image to the embedding position of the first null-value moment image, and add the null-value time zone embedding position timing information.

[0058] Furthermore, the intelligent trajectory following device based on image recognition is also used to: extract the first detected object type set, the first detected object size feature set and the first detected object distribution position set of the window cut image; extract the second detected object type set, the second detected object size feature set and the second detected object distribution position set of the first null value moment image; based on the first detected object distribution position set and the second detected object distribution position set, analyze the proportion of detected objects in the same position that have the same type as the first detected object type set and the second detected object type set, and the same size as the first detected object size feature set and the second detected object size feature set, to obtain the similarity.

[0059] Furthermore, the intelligent trajectory following device based on image recognition is also used to: obtain an item classification channel; input the window cut image into the item classification channel to obtain the first detected object type set, wherein the first detected object type set has the first detected object distribution position set in the window cut image; and calculate the first detected object size feature set of the first detected object type set based on the first detected object distribution position set.

[0060] Furthermore, the intelligent trajectory following device based on image recognition is also used for: when the target vehicle motion trajectory intersects with the prohibited driving area, rendering and marking the intersection trajectory to obtain an updated target vehicle motion trajectory.

[0061] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The method and specific examples of intelligent trajectory following based on image recognition in the aforementioned embodiment 1 are also applicable to an intelligent trajectory following device based on image recognition in this embodiment. Through the aforementioned detailed description of the method of intelligent trajectory following based on image recognition, those skilled in the art can clearly understand the intelligent trajectory following device based on image recognition in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here.

[0062] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for intelligent trajectory following based on image recognition, characterized in that: include: receiving positioning timing information from a target vehicle; When the positioning timing information has a null time zone, obtaining a start time adjacent position and an end time adjacent position of the null time zone; Downloading road image information from a road network database according to the adjacent position at the starting time and the adjacent position at the end time; receiving null-time zone road image information from the target vehicle; Performing similarity analysis on the road image information and the road image information of the null-value time zone to obtain the null-value time zone embedding position time series information; After compensating the positioning timing information according to the embedded position timing information of the null time zone, the movement trajectory of the target vehicle is obtained.

2. The method for intelligent trajectory following based on image recognition as claimed in claim 1, characterized in that: Downloading road image information from a road network database according to the adjacent position at the start time and the adjacent position at the end time includes: Uploading the adjacent positions at the starting time and the adjacent positions at the end time to a road network database, and downloading the first connected road image to the Nth connected road image; The first connected road image includes a first traveling direction left image, a first traveling direction right image and a first traveling direction front image; Until the Nth connected road image includes the Nth traveling direction left image, the Nth traveling direction right image and the Nth traveling direction front image.

3. The method for intelligent trajectory following based on image recognition as claimed in claim 1, characterized in that: Receiving null-time zone road image information from the target vehicle, including: The null-value time zone road image information includes a road image in front of the vehicle, a road image on the left side of the vehicle, and a road image on the right side of the vehicle.

4. The method for intelligent trajectory following based on image recognition as claimed in claim 1, characterized in that: Performing similarity analysis on the road image information and the road image information of the null value time zone to obtain the null value time zone embedding position time series information, including: Obtaining a first null-value moment image of null-value time zone road image information; Obtaining window size information of the image at the first null value moment; Randomly shearing the road image information according to the window size information to obtain a window shearing image; The similarity between the window cut image and the first null value moment image is compared. When the similarity is greater than or equal to a similarity threshold, the acquisition position of the window cut image is set to the embedding position of the first null value moment image, and the null value time zone embedding position timing information is added.

5. The method for intelligent trajectory following based on image recognition as claimed in claim 4, characterized in that: Comparing the similarity between the window clipping image and the image at the first null value moment includes: Extracting a first detected object type set, a first detected object size feature set, and a first detected object distribution position set from the window clipping image; Extracting a second detected object type set, a second detected object size feature set, and a second detected object distribution position set of the first null value moment image; Based on the first detected object distribution position set and the second detected object distribution position set, the proportion of detected objects having the same type as the first detected object type set and the second detected object type set at the same position and the same size as the first detected object size feature set and the second detected object size feature set is analyzed to obtain the similarity.

6. The method for intelligent trajectory following based on image recognition as claimed in claim 5, characterized in that: Extracting a first detected object type set, a first detected object size feature set, and a first detected object distribution position set from the window clipping image, comprising: Get item classification channel; Inputting the window cut image into the object classification channel to obtain the first detected object type set, wherein the first detected object type set has the first detected object distribution position set in the window cut image; The first detected object size feature set of the first detected object type set is calculated according to the first detected object distribution position set.

7. The method for intelligent trajectory following based on image recognition as claimed in claim 1, characterized in that: After compensating the positioning timing information according to the null time zone embedded position timing information, a target vehicle motion trajectory is obtained, including: When the target vehicle motion trajectory intersects with the prohibited driving area, the intersection trajectory is rendered and marked to obtain an updated target vehicle motion trajectory.

8. An intelligent trajectory following device based on image recognition, characterized in that: The steps for implementing the method of intelligent trajectory following based on image recognition as described in any one of claims 1 to 7 include: A positioning timing information receiving module, wherein the positioning timing information receiving module is used to receive positioning timing information from a target vehicle; A module for obtaining adjacent positions of a null-value time zone, wherein the module is used for obtaining adjacent positions of a start time and an end time of a null-value time zone when the positioning timing information has a null-value time zone; A road image information downloading module, the road image information downloading module is used to download the road image information from the road network database according to the adjacent position at the starting time and the adjacent position at the end time; A null-value time zone road image information receiving module, the null-value time zone road image information receiving module is used to receive null-value time zone road image information from the target vehicle; A similarity analysis module, the similarity analysis module is used to perform similarity analysis on the road image information and the road image information of the null value time zone to obtain the null value time zone embedding position time sequence information; A positioning timing information compensation module is used to compensate the positioning timing information according to the null value time zone embedded position timing information to obtain the target vehicle movement trajectory.

Images