A girder vehicle auxiliary driving recognition method and device, electronic equipment and storage medium

Abstract

The application discloses a kind of beam transport vehicle auxiliary driving identification methods, comprising the following steps: S1: constructing beam transport vehicle auxiliary driving identification system, the auxiliary driving identification system includes 360 degree video monitoring module and route rectification module;S2: obtain the real-time video image around the safety distance radius of beam transport vehicle, data preprocessing is carried out and is input to 360 degree video monitoring module, obtains detection identification result;S3: obtain the front road condition information, according to the lane line condition of current lane, corresponding distance calculation is carried out through route rectification network, and real-time calculation result is obtained;S4: according to the output result of 360 degree video monitoring module and route rectification module, corresponding auxiliary driving prompt result is generated.The application further provides a kind of device.The application further provides a kind of electronic equipment.The application further provides a kind of storage medium.The application solves the problems that existing auxiliary driving technology hardware cost is high, and there are limitations in dynamic obstacle detection and complex scene adaptability.

Description

Technical Field

[0001] This invention relates to the field of driver assistance recognition technology, and more specifically, to a driver assistance recognition method, device, electronic device, and storage medium for a beam transport vehicle. Background Technology

[0002] As a key transportation equipment in bridge construction, beam transport vehicles often face complex working conditions such as narrow construction access roads, slopes and turns, and nighttime construction. Due to the large load inertia and long braking distance, manual driving is easily affected by blind spots or judgment delays, which can lead to collisions and rollover accidents, seriously threatening construction safety and efficiency. The traditional driving mode that relies on manual observation and experience cannot meet the safety control requirements of high-precision real-time perception and rapid response.

[0003] Currently, the auxiliary driving technology for beam transport vehicles mainly uses lidar (LiDAR) and millimeter-wave radar. Although lidar has high accuracy, it has drawbacks such as high hardware cost, insufficient stability of rotating parts in vibration scenarios, and large computational load for point cloud data processing. Although millimeter-wave radar has strong anti-interference capabilities, it is not good at recognizing low obstacles and cannot meet the needs of complex scenarios. Both have limitations in dynamic obstacle detection and adaptability to complex scenarios. Summary of the Invention

[0004] To address the shortcomings or improvement needs of existing technologies, this invention provides a method, device, electronic device, and storage medium for assisted driving identification of beam transport vehicles. It solves the problems of high hardware costs and limitations in dynamic obstacle detection and adaptability to complex scenarios in existing assisted driving technologies.

[0005] According to an embodiment of the present invention, a method for assisted driving identification of a beam transport vehicle includes the following steps:

[0006] S1: Construct an auxiliary driving recognition system for the beam transport vehicle, which includes a 360-degree video monitoring module and a route correction module;

[0007] S2: Acquire real-time video images around the safe distance radius of the beam transport vehicle, perform data preprocessing, and input the data into the 360-degree video monitoring module to obtain the detection and recognition results;

[0008] S3: Obtain road condition information ahead, calculate the corresponding distance through the route correction network based on the lane markings of the current lane, and obtain the real-time calculation results;

[0009] S4: Based on the output results of the 360-degree video monitoring module and the route correction module, generate corresponding assisted driving prompts.

[0010] On the other hand, according to embodiments of the present invention, an auxiliary driving identification device for a beam transport vehicle is also provided, comprising:

[0011] Hierarchical backbone network: used to extract global feature information and generate feature maps;

[0012] Transformer interactive network: used for feature allocation and feature fusion;

[0013] Monitoring layer: Used to output monitoring results through the multi-scale decoupling head Detect;

[0014] Lane line information indicator submodule: used to determine the real-time status of the lane;

[0015] Virtual lane line generation submodule: used to virtually generate lane lines on the missing side;

[0016] Binocular distance calculation submodule: used to calculate the projected distance between the beam transport vehicle and the lane line.

[0017] On the other hand, according to an embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the above-described beam transporter assisted driving recognition method.

[0018] On the other hand, according to an embodiment of the present invention, a storage medium is also provided, which stores a computer program that, when executed, implements the above-described beam transporter assisted driving identification method.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] By constructing a dual-module system integrating 360-degree panoramic visual monitoring and intelligent route correction, a hierarchical backbone network is used to extract global features and combine it with Transformer to achieve multi-source information interaction. At the same time, virtual lane line generation technology is used to compensate for the lack of markings in complex road conditions. Combined with binocular distance calculation, high-precision pose perception is achieved. Through integrated vision and ranging perception, the shortcomings of traditional technologies in dynamic obstacle recognition and adaptability to complex scenarios are effectively overcome. The robustness of dynamic obstacle detection and lane keeping accuracy are significantly improved, and the risk of blind spots in manual driving is greatly reduced. Safety warning and real-time correction control in all construction scenarios are achieved with low hardware cost. Attached Figure Description

[0021] Figure 1 This is a flowchart of the beam transport vehicle assisted driving recognition method provided in the embodiments of the present invention;

[0022] Figure 2 This is a hierarchical backbone network structure diagram of the beam transport vehicle assisted driving recognition method provided in the embodiments of the present invention;

[0023] Figure 3 This is a structural diagram of the moving window module of the hierarchical backbone network of the beam transport vehicle assisted driving recognition method provided in this embodiment of the invention;

[0024] Figure 4 This is a Transformer interactive network structure diagram of the beam transport vehicle assisted driving recognition method provided in the embodiments of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] like Figure 1-4 As shown, Embodiment 1 of the present invention proposes an auxiliary driving identification method for a beam transport vehicle, comprising the following steps:

[0027] S1: Construct an auxiliary driving recognition system for the beam transport vehicle, which includes a 360-degree video monitoring module and a route correction module;

[0028] S2: Acquire real-time video images around the safe distance radius of the beam transport vehicle, perform data preprocessing, and input the data into the 360-degree video monitoring module to obtain the detection and recognition results;

[0029] S3: Obtain road condition information ahead, calculate the corresponding distance through the route correction network based on the lane markings of the current lane, and obtain the real-time calculation results;

[0030] S4: Based on the output results of the 360-degree video monitoring module and the route correction module, generate corresponding assisted driving prompts.

[0031] In one embodiment, a beam transport vehicle auxiliary driving recognition system is constructed, the auxiliary driving recognition system including a 360-degree video monitoring module and a route correction module, specifically including:

[0032] S11: Construct a 360-degree video monitoring module for the beam transport vehicle's auxiliary driving recognition system;

[0033] The 360-degree video monitoring module consists of three parts: a hierarchical backbone network, a Transformer interactive network, and a monitoring layer. The hierarchical backbone network is responsible for extracting global feature information and generating feature maps, the Transformer interactive network is responsible for feature allocation and feature fusion, and the monitoring layer outputs monitoring results through a multi-scale decoupled head Detect.

[0034] S12: Construct a route correction module for the auxiliary driving recognition system of the beam transport vehicle;

[0035] The route correction module comprises three parts: a lane line information indication submodule, a virtual lane line generation submodule, and a binocular distance calculation submodule. The lane line information indication submodule is responsible for judging the real-time lane status, the virtual lane line generation submodule is responsible for virtually generating lane lines on the missing side, and the binocular distance calculation submodule is responsible for calculating the projection distance between the beam transport vehicle and the lane line.

[0036] In one embodiment, real-time video images around the safety distance radius of the beam transport vehicle are acquired, preprocessed, and input into a 360-degree video monitoring module to obtain detection and recognition results, specifically including:

[0037] S21: Calculate the safe distance radius R of the beam transport vehicle. The expression for this process is:

[0038]

[0039] Among them, b x The maximum braking speed of the beam transport vehicle is indicated by h1, h2, and h3, which represent the driver's reaction time, the braking coordination time of the beam transport vehicle's brakes, and the deceleration increase time, respectively. x This indicates the current speed of the beam transport vehicle; r indicates the current length of the cargo being transported by the beam transport vehicle.

[0040] S22: Acquire 360-degree real-time video images within the safe distance radius R of the beam transport vehicle;

[0041] S23: Extract frames from video T to obtain the real-time frame image f of the beam transport vehicle;

[0042] S24: Perform preprocessing on image f, such as image scale normalization and image standardization, and then vectorize the preprocessed image information into feature vectors and input them into the 360-degree video monitoring module;

[0043] S25: The preprocessed and feature-vectorized image information is input into the hierarchical backbone network. The moving window module of the backbone network is responsible for extracting global image target feature information to form a feature map.

[0044] S26: Input the feature maps formed by each moving window module of the backbone network into the Transformer interactive network, obtain new feature maps by weight allocation through the cross attention module and self attention module, propagate the feature maps along the upper and lower structure of the network, and perform feature fusion at each level to obtain meta-features;

[0045] S27: The monitoring results are output using a multi-scale decoupling head Detect of the monitoring layer.

[0046] In one embodiment, the hierarchical backbone network employs a moving window mechanism and includes four stages, such as... Figure 2As shown, the feature vectorized image information is first input into the feature merging module, transforming the image from H×W×3 to... The first stage includes a linear embedding module, which performs a linear transformation on the channel data of each pixel, changing the number of channels from 48 to 96. The subsequent three stages include a feature merging module, which performs downsampling operation by merging features, halving the width and height of the feature map and doubling its depth.

[0047] Each stage includes a moving window module, which comprises four sub-modules: LN (Layer Normalization), GSA (Global Spatial Attention), MLP (Multilayer Perceptron), and LSA (Local Spatial Attention). These sub-modules are used to model and integrate the target image features within the window. The moving window module structure is as follows: Figure 3 As shown, the channels of the input feature map Zi are divided into two groups Zi. 1 and Zi 2 The LN, GSA, and MLP modules and the LN, LSA, and MLP modules within the two channels are linearly connected, respectively. The feature maps are input into the two channels respectively, resulting in two sets of MLP outputs corresponding to GSA and LSA. This process is expressed as:

[0048]

[0049] Among them, Z i It is the feature mapping of the input. and It is the feature mapping after equal division. These are the outputs of GSA and LSA, respectively. For the output of the MLP,

[0050] Subsequently, the outputs of the two MLPs are combined through a 1×1 convolutional layer. The feature maps Z are fused to generate the enhanced feature map Z in the (i+1)th stage. i+1 The expression for this process is:

[0051]

[0052] In one embodiment, the Transformer interactive network structure is as follows: Figure 2 and 3 As shown, the feature map Z2 formed by the moving window module in stage 2 of the hierarchical backbone network serves as the key in the interaction framework, the feature map Z3 formed by the moving window module in stage 3 of the hierarchical backbone network serves as the query in the interaction framework, and the feature map Z4 formed by the moving window module in stage 4 of the hierarchical backbone network serves as the value in the interaction framework. The interaction process expression inside the first layer of the Transformer is as follows:

[0053]

[0054] Transformer initial output meta-feature F 1 for:

[0055]

[0056] The output meta-feature F of the (i+1)th layer of the Transformer i+1 for:

[0057]

[0058] Where Z2, Z3, and Z4 represent the feature maps of the second, third, and fourth stages of the hierarchical backbone network, respectively, and Q, K, and V represent the key, query, and value in the interaction framework, respectively. and This represents a linear projection matrix specifically tailored for cross-attention operations to facilitate alignment between different features, where the parameter d represents the common dimension of the feature vectors.

[0059] The output F of the first layer of the Transformer 1 When inputting into the second layer, the potential relationships between query tokens must first be identified and revealed through a self-attention mechanism. The query, key, and value for self-attention are then obtained from the input features. This process is expressed as follows:

[0060]

[0061] Where Q', K', and V' represent the query, key, and value of self-attention, respectively. Let represent the learnable projection heads used in the self-attention module. These projection heads are used to transform the input query, key, and value representations respectively, while preserving the original feature dimensions throughout the process. By maintaining the dimensionality of the features, these projection heads ensure that the self-attention mechanism can effectively manipulate the data without altering the underlying feature space, thereby facilitating interaction between components. The expression for multi-head self-attention is:

[0062]

[0063] Among them, Q i K i and V i These are the query, key, and value of the i-th attention head, respectively, where Attn(·) represents the feature result of the specific projection head, and the feature dimension d is preferably set to 512 / H, where H represents the number of attention heads.

[0064] In one embodiment, road condition information ahead is obtained, and based on the lane markings of the current lane, the corresponding distance is calculated using a route correction network to obtain real-time calculation results, specifically including:

[0065] S31: Obtain road condition information ahead of the beam transport vehicle, and determine the existence of the actual lane line in the lane where the vehicle is currently located through the lane line information indication submodule in the route correction module;

[0066] The confidence scores C for the left and right sides of the lane will be obtained in real time. L , Input to the lane line information indicator submodule, if C L A value ≥0.8 and a continuous tracking frame count greater than 5 indicate the existence of a real lane line on the left; similarly, if C R A value ≥0.8 and a continuous tracking frame count greater than 5 indicate that the real lane line on the right exists; otherwise, it does not exist and virtual lane line generation is required.

[0067] S32: If there are no real lane lines on either side of the vehicle, lane lines are simulated based on map guide lines and preset road width, and the simulated virtual lane lines are used as the lane lines of the lane where the vehicle is currently located.

[0068] Map guide line parameterization uses a cubic polynomial to fit the map path:

[0069] f map (x)=a0+a1x+a2x 2 +a3x 3 ,

[0070] The coefficients a0, a1, a2, and a3 are provided by the high-precision map;

[0071] Symmetrical offset along the normal direction of the guide line The expression for generating the left and right virtual lane lines is:

[0072]

[0073] Among them, y L (x), y R (x) represent the left and right virtual lane lines respectively. To indicate the direction of the guide line curvature, W is a configurable preset standard road width parameter. This refers to the lateral distance offset to both sides;

[0074] S33: If there is a real lane line on only one side of the vehicle, then the lane line is simulated based on the real lane line and the preset road width to obtain the virtual lane line on the other side, and the real lane line and the virtual lane line are used as the lane line of the lane where the vehicle is currently located.

[0075] For generating virtual lane lines on one side, first set the detected real lane line as y. real (x) = kx + b, offset by W along the direction of the real lane line normal, to generate a virtual lane line on the other side, the expression is:

[0076]

[0077] When k > 0, the lane is shifted to the right to generate a virtual right lane line; when k < 0, the lane is shifted to the left to generate a virtual left lane line.

[0078] S34: If there are real lane lines on both sides of the vehicle, then the real lane lines are used as the lane lines of the lane the vehicle is currently in.

[0079] The detected dual-lane lines are smoothed using a filtering process, expressed as follows:

[0080] y filtered (t)=α·y real (t)+(1-α)·y filtered (t-1),

[0081] Where α is the filtering coefficient used to suppress detection jitter; y(t) represents the lane line in the current frame;

[0082] S35: Calculate the distance between the left and right sides of the beam transport vehicle and the lane lines through the binocular distance calculation submodule in the route correction module, and obtain the real-time calculation results;

[0083] First, using stereo matching, the disparity map D(u,v) is calculated using the SGBM algorithm, with the following expression:

[0084] D(u,v)=arg min d (∑ p∈N(u,v) |I L (p)-I R (pd)|+λ·smoothness(d)),

[0085] Where N(u,v) is the matching window centered on the pixel, and I L (p) is the brightness value of pixel p in the left camera image, I R (pd) is the brightness value of the corresponding point in the right camera image after horizontal displacement d, d is the candidate disparity value, λ is the weight coefficient of the smoothing term, and smoothness(d) is the disparity smoothing constraint term.

[0086] Subsequently, the lateral distance is calculated based on the binocular geometric model using three-dimensional coordinate solving, and the expression is:

[0087]

[0088] Among them, B is the binocular baseline, f is the camera focal length, and c x is the abscissa of the image center;

[0089] Finally, by projecting the left and right nearest lane line points, the lane line distance is calculated, and the expression is:

[0090]

[0091] Among them, k and b are the parameters of the corresponding side lane line equation, and (X vehicle , Y vehicle ) is the projection coordinate of the vehicle center;

[0092] In one embodiment, according to the output results of the 360-degree video monitoring module and the route deviation correction module, the corresponding assisted driving prompt results are generated, specifically including:

[0093] S41: Set the 360-degree warning distance range as D1, and the route deviation correction warning distance as D2;

[0094] S42: When the beam transport vehicle is driving normally along the planned lane line, if an abnormal target is detected within the D1 range, a voice prompt is given to assist the driver; if the distance d between the beam transport vehicle and the lane line during driving side < D2, it is prompted that the corresponding side is about to press the line, and the driver is reminded to pay attention to whether the route is deviated;

[0095] S43: When the beam transport vehicle needs to change lanes due to special circumstances, that is, when d side tends to 0, if a vehicle or pedestrian is detected within the D1 distance range on the corresponding lane-changing side, it is prompted that there is a vehicle or pedestrian on the corresponding side, and the driver is reminded to change lanes carefully.

[0096] The present invention constructs a dual-module system that integrates 360-degree panoramic vision monitoring and intelligent route deviation correction, uses a hierarchical backbone network to extract global features and combines Transformer to achieve multi-source information interaction. At the same time, a virtual lane line generation technology is adopted to compensate for the missing signs under complex road conditions, and binocular distance calculation is combined to achieve high-precision pose perception. Through visual and ranging integrated perception, it effectively overcomes the defects of traditional technologies in dynamic obstacle recognition and complex scene adaptability, significantly improves the robustness of dynamic obstacle detection and the lane keeping accuracy, greatly reduces the risk of manual driving blind spots, and realizes safety warning and real-time deviation correction control for the entire construction scene at a low hardware cost.

[0097] According to an embodiment of the present invention, there is also provided a beam transport vehicle assisted driving recognition device, which includes:

[0098] Hierarchical backbone network: used to extract global feature information to generate feature maps;

[0099] Transformer interactive network: used for feature allocation and feature fusion;

[0100] Monitoring layer: Used to output monitoring results through the multi-scale decoupling head Detect;

[0101] Lane line information indicator submodule: used to determine the real-time status of the lane;

[0102] Virtual lane line generation submodule: used to virtually generate lane lines on the missing side;

[0103] Binocular distance calculation submodule: used to calculate the projected distance between the beam transport vehicle and the lane line.

[0104] According to an embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the above-described beam transport vehicle assisted driving recognition method.

[0105] According to an embodiment of the present invention, a storage medium is also provided, which stores a computer program that, when executed, implements the above-described assisted driving identification method for beam transport vehicles.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for assisting driving and identifying beam transport vehicles, characterized in that, Includes the following steps: S1: Construct an auxiliary driving recognition system for the beam transport vehicle, which includes a 360-degree video monitoring module and a route correction module; Step S1 includes the following sub-steps: S11: Construct a 360-degree video monitoring module for the beam transporter's assisted driving recognition system. The 360-degree video monitoring module includes a hierarchical backbone network, a Transformer interactive network, and a monitoring layer. S12: Construct a route correction module for the auxiliary driving recognition system of the beam transport vehicle. The route correction module includes a lane line information indication submodule, a virtual lane line generation submodule, and a binocular distance calculation submodule. S2: Acquire real-time video images around the safe distance radius of the beam transport vehicle, perform data preprocessing, and input the data into the 360-degree video monitoring module to obtain the detection and recognition results; Step S2 includes the following sub-steps: S21: Calculate the safe distance radius of the beam transport vehicle; S22: Acquire 360-degree real-time video images within the safe distance radius of the beam transport vehicle; S23: Extract frames from the video to obtain real-time frame images of the beam transport vehicle; S24: Perform preprocessing on the image, such as image scale normalization and image standardization, and then convert the preprocessed image information into feature vectors and input them into the 360-degree video monitoring module; S25: The preprocessed and feature-vectorized image information is input into the hierarchical backbone network. The moving window module of the backbone network is responsible for extracting global image target feature information to form a feature map. S26: Input the feature maps formed by each moving window module of the backbone network into the Transformer interactive network, obtain new feature maps by weight allocation through the cross attention module and self attention module, propagate the feature maps along the upper and lower structure of the network, and perform feature fusion at each level to obtain meta-features; S27: The monitoring results are output using the multi-scale decoupling head Detect of the monitoring layer; S3: Obtain road condition information ahead, calculate the corresponding distance through the route correction network based on the lane markings of the current lane, and obtain the real-time calculation results; Step S3 includes the following sub-steps: S31: Obtain road condition information ahead of the beam transport vehicle, and determine the existence of the actual lane line in the lane where the vehicle is currently located through the lane line information indication submodule in the route correction module; S32: If there are no real lane lines on either side of the vehicle, lane lines are simulated based on map guide lines and preset road width, and the simulated virtual lane lines are used as the lane lines of the lane where the vehicle is currently located. S33: If there is a real lane line on only one side of the vehicle, then the lane line is simulated based on the real lane line and the preset road width to obtain the virtual lane line on the other side, and the real lane line and the virtual lane line are used as the lane line of the lane where the vehicle is currently located. S34: If there are real lane lines on both sides of the vehicle, then the real lane lines are used as the lane lines of the lane the vehicle is currently in. S35: Calculate the distance between the left and right sides of the beam transport vehicle and the lane lines through the binocular distance calculation submodule in the route correction module, and obtain the real-time calculation results; S4: Based on the output results of the 360-degree video monitoring module and the route correction module, generate corresponding assisted driving prompts.

2. The method for assisting driving and identifying a beam transport vehicle as described in claim 1, characterized in that: The hierarchical backbone network in step S25 adopts a moving window mechanism and includes four stages. The first stage includes a linear embedding module, which performs a linear transformation on the channel data of each pixel. The subsequent three stages include a feature merging module, which performs downsampling operation by merging features to halve the width and height of the feature map and double its depth. Each stage includes a moving window module, which comprises four sub-modules: LN, GSA, MLP, and LSA. These modules are used to model and integrate the target image features within the window. Specifically, the modules include: dividing the input feature map channels into two groups; linearly connecting the LN, GSA, and MLP modules and the LN, LSA, and MLP modules within each group of channels; inputting the feature map into each of the two channels to obtain two sets of MLP outputs corresponding to GSA and LSA, respectively; and then fusing the outputs of the two sets of MLPs to obtain the enhanced feature map for the corresponding stage.

3. The method for assisting driving and identifying beam transport vehicles as described in claim 2, characterized in that: The feature maps formed by the second, third, and fourth stage moving window modules of the hierarchical backbone network serve as the key, query, and value inputs to the Transformer interaction network in the interaction framework. The linear projection matrix of the cross-attention module promotes the alignment of different features, obtaining initial output meta-features. These meta-features are then propagated along the upper and lower layers of the network. The meta-features received from the upper layer are used by the learnable projection head of the self-attention module to obtain new keys, queries, and values ​​for the Transformer interaction network, which are then used as inputs to the lower layer structure to obtain the feature results of the corresponding projection head. The feature results and the received meta-features are then fused to obtain the output meta-features of the corresponding layer.

4. The method for assisting driving and identifying a beam transport vehicle as described in claim 1, characterized in that, Step S4 includes the following sub-steps: S41: Set the 360-degree warning distance range and the route correction warning distance; S42: When the beam transport vehicle is driving normally along the planned lane line, if an abnormal target is detected within the 360-degree warning distance range, a voice prompt will be given to assist the driver. If the distance between the beam transport vehicle and the lane line is less than the route correction warning distance during the journey, the driver will be prompted to pay attention to whether the vehicle is about to cross the line on the corresponding side and whether it is deviating from the route. S43: When a beam transport vehicle encounters a special situation and needs to change lanes, that is, when the distance from the lane line approaches 0, if a vehicle or pedestrian is detected within a 360-degree warning distance range on the corresponding lane side, the driver will be prompted to be cautious when changing lanes as there is a vehicle or pedestrian on the corresponding side.

5. A beam transport vehicle auxiliary driving identification device, using the beam transport vehicle auxiliary driving identification method as described in claim 1, characterized in that, include: Hierarchical backbone network: used to extract global feature information and generate feature maps; Transformer interactive network: used for feature allocation and feature fusion; Monitoring layer: Used to output monitoring results through the multi-scale decoupling head Detect; Lane line information indicator submodule: used to determine the real-time status of the lane; Virtual lane line generation submodule: used to virtually generate lane lines on the missing side; Binocular distance calculation submodule: used to calculate the projected distance between the beam transport vehicle and the lane line.

6. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the beam transport vehicle assisted driving recognition method according to any one of claims 1-4.

7. A storage medium, characterized in that, The device contains a computer program that, when executed, implements the beam transport vehicle assisted driving identification method according to any one of claims 1-4.

Images