Anti-collision early warning method for vehicle, computer program product and vehicle

Abstract

The invention relates to an anti-collision early warning method of a vehicle, a computer program product and the vehicle. The method comprises the following steps: acquiring a parameter value set corresponding to each operation parameter in a plurality of operation parameters of the vehicle running in a preset time period; the parameter value set comprises parameter values of the corresponding operation parameters at different time points; preprocessing the plurality of parameter value sets to obtain a time parameter matrix; inputting the time parameter matrix into a pre-trained anti-collision early warning model to obtain a plurality of early warning strategies output by the anti-collision early warning model and a probability value corresponding to each early warning strategy; and according to the probability value, determining a current target early warning strategy for the vehicle from a plurality of early warning strategies, and sending an early warning prompt to the vehicle according to the target early warning strategy. Therefore, the collision event of the vehicle in the driving process can be predicted more accurately and quickly, early warning is provided for a user in advance, the vehicle is prevented from colliding with other vehicles, the driving safety is improved, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle safety technology, and in particular to a vehicle anti-collision warning method, a computer program product and a vehicle. Background Art

[0002] With the rapid development of the automobile industry and people's increasing attention to traffic safety, automobile collision warning technology has become a research hotspot.

[0003] In the related art, for the anti-collision warning of automobiles, a single sensor and a simple algorithm are generally used to determine whether the vehicle has collided with other vehicles. For example, a radar sensor is used to detect the distance between the current vehicle and the vehicle in front, and when it is determined that the distance exceeds a preset distance threshold, it is determined that the vehicle has collided with other vehicles and an alarm is issued.

[0004] However, since sensors are easily affected by environmental factors (bad weather, electromagnetic interference, etc.), the above-mentioned detection of whether a vehicle has collided with other vehicles through a single sensor and a simple algorithm is prone to inaccurate detection, missed detection or false detection, resulting in an inability to accurately provide early warnings to users, affecting the user experience. Summary of the invention

[0005] The present application provides a vehicle anti-collision warning method, a computer program product and a vehicle to solve the technical problem in the prior art that when a single sensor and a simple algorithm are used to detect whether a vehicle has collided with other vehicles, inaccurate detection, missed detection or false detection may occur, thereby failing to accurately provide warnings to users and affecting user experience.

[0006] In a first aspect, the present application provides a vehicle anti-collision warning method, the method comprising:

[0007] Obtaining a parameter value set corresponding to each of the multiple operating parameters of the vehicle during a preset time period; the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points;

[0008] Preprocessing the plurality of parameter value sets to obtain a time parameter matrix;

[0009] Inputting the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each of the warning strategies;

[0010] According to the probability value, a current target warning strategy for the vehicle is determined from a plurality of warning strategies, and a warning prompt is issued to the vehicle according to the target warning strategy.

[0011] As an optional implementation, the operating parameters include operating parameters of the vehicle, operating parameters of a target vehicle, and operating parameters between the vehicle and the target vehicle, the target vehicle being the vehicle closest to the vehicle;

[0012] The step of obtaining a plurality of operating parameters of the vehicle during the preset time period, wherein each of the operating parameters corresponds to a parameter value set, includes:

[0013] Respectively measuring a speed set, an acceleration set, and a steering angle set of the vehicle traveling within the preset time period;

[0014] Determining a relative distance set and a relative speed set between the vehicle and the target vehicle, and a collision avoidance acceleration set of the target vehicle;

[0015] The speed set, the acceleration set, the steering angle set, the relative distance set, the relative speed set, and the anti-collision acceleration set are all determined as the parameter value set.

[0016] As an optional implementation, the anti-collision warning model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer; the anti-collision warning model outputs multiple warning strategies and a probability value corresponding to each warning strategy in the following manner:

[0017] Acquire the time parameter matrix through the input layer, and perform convolution calculation on the time parameter matrix through the convolution layer to obtain multiple convolution features;

[0018] Performing feature extraction on the multiple convolutional features through the pooling layer to obtain a pooling feature matrix;

[0019] Performing full connection calculation on the pooling feature matrix through the fully connected layer to obtain vector features;

[0020] The output layer performs probability calculation on the vector features to obtain a probability value corresponding to each of a plurality of pre-set early warning strategies, and outputs a plurality of the early warning strategies and the probability value corresponding to each of the early warning strategies.

[0021] As an optional implementation, the convolution layer includes N convolution kernels, where N is a positive integer, and the convolution layer performs convolution calculation on the time parameter matrix to obtain multiple convolution features, including:

[0022] For each of the convolution kernels, sliding the convolution kernel in the time parameter matrix according to a preset step size;

[0023] For the submatrix corresponding to each sliding of the convolution kernel, convolution calculation is performed on the submatrix to obtain the sub-features corresponding to the submatrix;

[0024] A matrix composed of multiple sub-features is determined as a convolution feature after the convolution kernel convolves the time parameter matrix.

[0025] As an optional implementation manner, performing convolution calculation on the submatrix to obtain sub-features corresponding to the submatrix includes:

[0026] Obtaining a weight matrix and a first bias item corresponding to the convolution kernel;

[0027] Performing weighted summation on the elements in the submatrix according to the weight matrix to obtain an initial convolution value;

[0028] The initial convolution value is added to the first bias term to obtain a sub-feature corresponding to the sub-matrix.

[0029] As an optional implementation, the extracting the multiple convolutional features through the pooling layer to obtain a pooling feature matrix includes:

[0030] Determining a pooling window of the pooling layer;

[0031] For each of the convolutional features, dividing the convolutional features into a plurality of sub-window matrices according to the pooling window;

[0032] Obtain the maximum value in each of the sub-window matrices, and use the matrix composed of the maximum values ​​as the sub-pooling feature matrix corresponding to the convolution feature;

[0033] Combine the multiple sub-pooling feature matrices to obtain a pooling feature matrix whose number of channels is the number of the convolution features.

[0034] As an optional implementation, the fully connected layer includes M neurons, where M is a positive integer, and the fully connected layer performs a fully connected calculation on the pooled feature matrix to obtain a vector feature, including:

[0035] Convert the pooled feature matrix into a one-dimensional vector;

[0036] For each of the neurons, obtaining a first weight vector and a second bias item corresponding to the neuron;

[0037] Performing weighted summation on the one-dimensional vector according to the first weight vector to obtain an initial sub-vector feature;

[0038] Adding the initial sub-vector feature to the second bias term to obtain a sub-vector feature;

[0039] The sub-vector features corresponding to each of the M neurons are combined to obtain the vector feature.

[0040] As an optional implementation, the output layer includes T neurons, each of which corresponds to an early warning strategy, and T is a positive integer. The probability calculation of the vector feature is performed through the output layer to obtain a probability value corresponding to each early warning strategy in a plurality of pre-set early warning strategies, including:

[0041] For each of the neurons, determining a second weight vector and a third bias item corresponding to the neuron;

[0042] performing weighted summation on the vector features according to the second weight vector to obtain a first value;

[0043] Adding the first value to the third bias term to obtain a second value;

[0044] The T second values ​​are calculated using a preset activation function to obtain a probability value corresponding to each second value.

[0045] As an optional implementation, the warning strategy includes normal driving, decelerated driving, and emergency braking;

[0046] The issuing of a warning prompt to the vehicle according to the target warning strategy includes:

[0047] When the target warning strategy is slow driving or sudden braking, a warning prompt sound corresponding to the target warning strategy is played through a preset sound playing module;

[0048] and / or,

[0049] When the target warning strategy is slow driving or sudden braking, the warning prompt symbol corresponding to the target warning strategy is displayed through a preset display interface.

[0050] In a second aspect, the present application provides a computer program product, which, when executed on a computer, enables the computer to execute any anti-collision warning method described in the first aspect.

[0051] In a third aspect, the present application provides a vehicle, comprising: a processor and a memory, wherein the processor is used to execute a vehicle anti-collision warning program stored in the memory to implement the vehicle anti-collision warning method described in any one of the first aspects.

[0052] In a fourth aspect, the present application provides a vehicle anti-collision warning device, the device comprising:

[0053] An acquisition module, used to acquire a parameter value set corresponding to each of the multiple operating parameters of the vehicle during a preset time period; the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points;

[0054] A preprocessing module, used for preprocessing the plurality of parameter value sets to obtain a time parameter matrix;

[0055] An input module, used to input the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each of the warning strategies;

[0056] A determination module is used to determine a current target warning strategy for the vehicle from multiple warning strategies according to the probability value, and issue a warning prompt to the vehicle according to the target warning strategy.

[0057] In a fifth aspect, the present application provides a storage medium storing one or more programs, which can be executed by one or more processors to implement the vehicle anti-collision warning method described in any one of the first aspects.

[0058] The technical solution provided by the embodiment of the present application is to obtain a parameter value set corresponding to each operating parameter in a plurality of operating parameters of the vehicle during a preset time period, wherein the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points, pre-process the plurality of parameter value sets to obtain a time parameter matrix, input the time parameter matrix into a pre-trained anti-collision warning model, obtain a plurality of warning strategies output by the anti-collision warning model, and a probability value corresponding to each warning strategy, and determine the current target warning strategy for the vehicle from the plurality of warning strategies according to the probability value, and issue a warning prompt to the vehicle according to the target warning strategy. This technical solution integrates a plurality of parameter value sets involved in the vehicle operation process into a time parameter matrix, processes the time parameter matrix using a pre-trained anti-collision warning model, and determines the warning strategy that the vehicle currently needs to collect according to the output result of the anti-collision warning model. Through the training of the anti-collision warning model and the integration of a plurality of different parameter value sets, the collision event can be predicted quickly and accurately, thereby realizing a more accurate and rapid prediction of the collision event of the vehicle during driving, thereby providing a warning to the user in advance, preventing the vehicle from colliding with other vehicles, improving driving safety, and enhancing user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art texts are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0061] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0062] Figure 1 A schematic diagram of an application scenario corresponding to a vehicle anti-collision warning method provided in an embodiment of the present application;

[0063] Figure 2 A flowchart of an embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application;

[0064] Figure 3 A flowchart of another embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application;

[0065] Figure 4 A flowchart of another embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application;

[0066] Figure 5 A schematic diagram of the structure of an anti-collision warning model provided in an embodiment of the present application;

[0067] Figure 6 A flowchart of another embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application;

[0068] Figure 7 A flowchart of another embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application;

[0069] Figure 8 A block diagram of an embodiment of a vehicle anti-collision warning device provided in an embodiment of the present application;

[0070] Fig. 9 A schematic diagram of the structure of a vehicle provided in an embodiment of the present application. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the embodiments described in the text are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0072] The disclosure below provides many different embodiments or examples to realize the different structures of the present application. In order to simplify the disclosure of the present application, the parts and settings of specific examples are described in detail below. Of course, they are only examples, and the purpose is not to limit the present application. In addition, the present application can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0073] In order to solve the technical problem in the prior art that a single sensor and a simple algorithm are used to detect whether a vehicle collides with other vehicles, which easily leads to inaccurate detection, missed detection or false detection, thereby failing to accurately provide warnings to users and affecting user experience, the present application provides a vehicle anti-collision warning method, a computer program product and a vehicle, which can integrate multiple parameter value sets involved in the operation of the vehicle into a time parameter matrix, and use a pre-trained anti-collision warning model to process the time parameter matrix, and determine the warning strategy that the vehicle currently needs to collect according to the output result of the anti-collision warning model. It can quickly and accurately predict collision events through the training of the anti-collision warning model and the integration of multiple different parameter value sets, thereby achieving more accurate and rapid prediction of collision events of the vehicle during driving, thereby providing users with early warnings to prevent the vehicle from colliding with other vehicles, improving driving safety and enhancing user experience.

[0074] To facilitate understanding of the anti-collision warning method for a vehicle provided in an embodiment of the present application, the following first provides examples of application scenarios involving the anti-collision warning method for a vehicle provided in the present application.

[0075] join Figure 1 , is a schematic diagram of an application scenario corresponding to a vehicle anti-collision warning method provided in an embodiment of the present application. Figure 1 As shown, the application scenario 10 may include a vehicle 11 and a vehicle 12 .

[0076] Among them, the vehicle 11 can be located behind the vehicle 12 or in front of the vehicle 12 during driving, and this embodiment of the present application does not limit this.

[0077] The types, sizes, models, etc. of the vehicles 11 and 12 are not limited in this embodiment of the present application.

[0078] In actual applications, when a driver in vehicle 11 is driving vehicle 11 and there is at least one vehicle around, such as vehicle 12, if the distance between vehicle 11 and vehicle 12 is close, when vehicle 11 is traveling too fast or too slow, or turns at a large angle, it is easy for vehicle 11 to collide with vehicle 12.

[0079] In this regard, in the related art, the controller of vehicle 11 can detect the distance between itself and vehicle 12 through a pre-installed sensor, and when it is determined that the distance is greater than a preset distance threshold, it predicts that vehicle 11 may collide with vehicle 12 and promptly warns the driver of vehicle 11.

[0080] However, since sensors are easily affected by environmental factors (bad weather, electromagnetic interference, etc.), the above method is likely to lead to inaccurate detection, missed detection or false detection, resulting in the inability to accurately provide warnings to users, safety accidents, and affected user experience.

[0081] In response to the above technical problems, the present application provides a vehicle anti-collision warning method, which can accurately and quickly predict collision events of the vehicle during driving, thereby providing users with accurate warnings in advance and improving user experience.

[0082] The anti-collision warning method for a vehicle provided by the present application is further explained below with reference to specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation on the embodiments of the present application.

[0083] join Figure 2 , is a flow chart of an embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application. As an embodiment, the vehicle anti-collision warning method provided in an embodiment of the present application can be applied to vehicles, such as Figure 1 The vehicle 11 shown in the figure can also be applied to a controller of a vehicle. Figure 2 As shown, the process may include the following steps:

[0084] Step 201: Obtain a parameter value set corresponding to each operating parameter among a plurality of operating parameters of a vehicle traveling within a preset time period, wherein the parameter value set includes parameter values ​​of the corresponding operating parameter at different time points.

[0085] The above-mentioned preset time period refers to a preset time period from the current time point, for example, within 1 minute or 2 minutes from the current time point.

[0086] The above-mentioned operating parameters refer to the parameters involved in the vehicle's driving process, which may include but are not limited to: the vehicle's driving speed, acceleration, steering angle, relative distance to other vehicles, relative speed, and acceleration of other vehicles.

[0087] The above parameter value set refers to the parameter values ​​corresponding to the operating parameters of the vehicle at different time points. The parameter value set may include the same or different parameter values, that is, the operating parameters may correspond to the same parameter values ​​at different time points, or may correspond to different parameter values, and the embodiment of the present application does not limit this. For example, when the operating parameter is the vehicle speed, the parameter value set may include multiple vehicle speed values ​​of the vehicle within a preset time period, and the multiple vehicle speed values ​​may be the same or different, and the embodiment of the present application does not limit this.

[0088] In some embodiments of the present application, the executing entity of the embodiments of the present application can obtain parameter values ​​corresponding to each operating parameter at different time points among multiple operating parameters of the vehicle traveling within a preset time period, and obtain a parameter value set corresponding to each operating parameter.

[0089] As an optional implementation method, the execution entity of the embodiment of the present application can obtain in real time a plurality of parameter value sets of the vehicle traveling within a preset time period.

[0090] As another optional implementation method, the execution subject of the embodiment of the present application may periodically obtain multiple parameter value sets of the vehicle traveling within a preset time period.

[0091] As another optional implementation, the execution subject of the embodiment of the present application may obtain a plurality of parameter value sets of the vehicle traveling within a preset time period when the pre-installed sensor detects the presence of other vehicles within a preset range of the vehicle. The above-mentioned preset range may be within a preset distance from the vehicle in different directions of the vehicle.

[0092] As for how to set the parameter values ​​of each running parameter, you can refer to the following Figure 3 The process shown is explained and will not be described in detail here.

[0093] Step 202: Preprocess multiple parameter value sets to obtain a time parameter matrix.

[0094] The above-mentioned preprocessing refers to preprocessing the acquired parameter value set, and the processing method may include but is not limited to: outlier processing, missing value processing, and normalization processing.

[0095] The above-mentioned time parameter matrix refers to the parameter values ​​of multiple operating parameters corresponding to different time points during the driving process of the vehicle, that is, the time parameter matrix may include the correspondence between the time points and the parameter values ​​of multiple operating parameters, which can be a two-dimensional matrix of time points and multiple operating parameters.

[0096] In some embodiments of the present application, after obtaining the above-mentioned multiple parameter value sets during the vehicle driving process, the multiple parameter value sets can be preprocessed to obtain a more accurate correspondence between the time points and the multiple operating parameters, thereby obtaining a time parameter matrix for the vehicle driving within a preset time period.

[0097] As for how to preprocess multiple parameter value sets to obtain the above time parameter matrix, it can be described below. Figure 3 The process shown is explained and will not be described in detail here.

[0098] Step 203: input the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each warning strategy.

[0099] Step 204: According to the above probability value, determine the current target warning strategy for the vehicle from the multiple warning strategies mentioned above, and issue a warning prompt to the vehicle according to the above target warning strategy.

[0100] The following is a unified description of step 203 and step 204:

[0101] The above-mentioned anti-collision warning model refers to a pre-trained model used to predict whether there is a collision situation in the current operating state of the vehicle and to provide a response warning strategy.

[0102] The above-mentioned warning strategy refers to a warning strategy used to warn the driver of the vehicle based on the current operating status of the vehicle. For example, if the current operating status may cause a collision with other vehicles, then the corresponding warning strategy may be slow driving, so that the driver reduces the vehicle's driving speed according to the warning strategy to avoid collision with other vehicles.

[0103] The above-mentioned warning prompt refers to a warning prompt issued to the user, which is used to prompt the user of anti-collision strategies, such as deceleration, emergency braking, etc. The prompt method of the above-mentioned warning prompt can be a sound prompt, a symbol prompt of a preset color output through a visual interface, etc., and the embodiment of the present application does not limit this.

[0104] In some embodiments of the present application, after the executing entity of the embodiments of the present application determines the time parameter matrix of the vehicle within a preset time period, the above time parameter matrix can be input into a pre-trained anti-collision warning model. The anti-collision warning model can predict whether the vehicle collides with other vehicles under the current operating state based on the received time parameter matrix, and output multiple warning strategies and the probability value corresponding to each warning strategy.

[0105] Afterwards, the current target warning strategy for the vehicle can be determined from the above-mentioned multiple warning strategies according to the probability value, and a warning prompt can be issued to the vehicle according to the above-mentioned target warning strategy.

[0106] As an optional implementation method, the warning strategy with the largest probability value can be determined as the target warning strategy.

[0107] As another optional implementation, each of the above-mentioned multiple warning strategies may correspond to a priority. Based on this, when there are multiple warning strategies with the same probability value, the warning strategy with the highest priority may be determined as the target warning strategy.

[0108] Finally, the execution entity of the embodiment of the present application can issue a warning prompt to the vehicle according to the above-mentioned target warning strategy.

[0109] As an optional implementation, the warning strategy may include but is not limited to: normal driving, decelerated driving, and emergency braking. The normal driving is used to represent driving at the current driving speed, the decelerated driving is used to represent reducing the current driving speed according to a preset step speed, and the emergency braking is used to represent braking and stopping immediately.

[0110] Based on this, in an exemplary embodiment, when the target warning strategy is slow driving or sudden braking, a warning prompt sound corresponding to the target warning strategy can be played through a preset sound playing module (such as a speaker, an audio output device), such as "Please slow down!";

[0111] In another exemplary embodiment, when the current target warning strategy is slow driving or sudden braking, the warning prompt symbol corresponding to the target warning strategy can be displayed through a preset display interface. For example, the warning prompt symbol corresponding to slow driving is a yellow exclamation mark, and the warning prompt symbol corresponding to sudden braking is a red exclamation mark.

[0112] In another exemplary embodiment, when the target warning strategy is slow driving or sudden braking, the warning prompt sound corresponding to the target warning strategy can be played through the preset sound playback module, and the warning prompt symbol corresponding to the target warning strategy can be displayed through the preset display interface.

[0113] The technical solution provided by the embodiment of the present application is to obtain a parameter value set corresponding to each operating parameter in a plurality of operating parameters of the vehicle during a preset time period, wherein the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points, pre-process the plurality of parameter value sets to obtain a time parameter matrix, input the time parameter matrix into a pre-trained anti-collision warning model, obtain a plurality of warning strategies output by the anti-collision warning model, and a probability value corresponding to each warning strategy, and determine the current target warning strategy for the vehicle from the plurality of warning strategies according to the probability value, and issue a warning prompt to the vehicle according to the target warning strategy. This technical solution integrates a plurality of parameter value sets involved in the vehicle operation process into a time parameter matrix, processes the time parameter matrix using a pre-trained anti-collision warning model, and determines the warning strategy that the vehicle currently needs to collect according to the output result of the anti-collision warning model. Through the training of the anti-collision warning model and the integration of a plurality of different parameter value sets, the collision event can be predicted quickly and accurately, thereby realizing a more accurate and rapid prediction of the collision event of the vehicle during driving, thereby providing a warning to the user in advance, preventing the vehicle from colliding with other vehicles, improving driving safety, and enhancing user experience.

[0114] See also Figure 3 , which is a flow chart of an embodiment of another vehicle anti-collision warning method provided in an embodiment of the present application. Figure 3 The process shown in Figure 2 Based on the process shown, it is described how to obtain a parameter value set corresponding to each operating parameter in a plurality of operating parameters of the vehicle traveling within a preset time period, and how to pre-process the plurality of parameter value sets to obtain a time parameter matrix when the operating parameters include the operating parameters of the vehicle, the operating parameters of the target vehicle, and the operating parameters between the vehicle and the target vehicle. Figure 3 As shown, the process may include the following steps:

[0115] Step 301: respectively measure a set of speeds, a set of accelerations, and a set of steering angles of a vehicle traveling within a preset time period.

[0116] In some embodiments of the present application, the operating parameters of the vehicle may include the speed, acceleration, and steering angle of the vehicle. Based on this, the execution subject of the embodiments of the present application can measure the speed set, acceleration set, and steering angle set of the vehicle in a preset time period through a plurality of different parameter sensors pre-installed. The above-mentioned parameter sensors are sensors pre-installed on the vehicle for measuring the operating parameters of the vehicle, which may include but are not limited to: wheel speed sensors, acceleration sensors, and steering angle sensors.

[0117] As an optional implementation, the parameter sensors may include a wheel speed sensor, an acceleration sensor, and a steering angle sensor.

[0118] The wheel speed sensor may be installed at a wheel attachment of the vehicle, so that the wheel speed of the vehicle is collected through the wheel speed sensor, and the wheel speed is determined as the speed of the vehicle to obtain a corresponding speed set within a preset time period.

[0119] Among them, the above-mentioned acceleration sensor can be based on micro-electromechanical system technology, through the movable mass block set inside, when the vehicle accelerates or decelerates, the displacement generated by the inertia of the mass block is converted into an electrical signal to determine the magnitude and direction of the vehicle's acceleration, thereby obtaining the acceleration set of the vehicle within a preset time period.

[0120] The steering angle sensor may be installed on a steering column of a vehicle to measure the steering angle of the vehicle and obtain a set of steering angles of the vehicle within a preset time period.

[0121] Step 302: Determine a relative distance set and a relative speed set between the vehicle and the target vehicle, and an anti-collision acceleration set of the target vehicle.

[0122] Step 303: Determine the above speed set, acceleration set, steering angle set, relative distance set, relative speed set, and anti-collision acceleration set as a parameter value set.

[0123] The following is a unified description of step 302 and step 303:

[0124] The above-mentioned target vehicle refers to the vehicle that is currently closest to the vehicle (the vehicle where the executing entity of the embodiment of the present application is located), that is, the vehicle that may collide with the vehicle. It may be a vehicle located in front of, behind, or on both sides of the vehicle. The embodiment of the present application does not impose any restrictions on this.

[0125] In some embodiments of the present application, the operating parameters may include, in addition to the operating parameters of the vehicle, the operating parameters between the vehicle and the target vehicle (such as the relative distance and relative speed between the vehicle and the target vehicle, etc.), and the operating parameters of the target vehicle (such as the acceleration of the target vehicle (hereinafter referred to as "anti-collision acceleration")).

[0126] Based on this, the execution subject of the embodiment of the present application can determine the relative distance set, relative speed set, and anti-collision acceleration set of the target vehicle between the vehicle and the target vehicle within a preset time period.

[0127] As an implementation method, a relative distance set and a relative speed set between the vehicle and the target vehicle within a preset time period, as well as a collision avoidance acceleration set of the target vehicle can be determined based on a preset radar sensor. The above-mentioned radar sensor refers to a radar sensor pre-installed on the vehicle, which can be a laser radar sensor or an infrared radar sensor, and the present application embodiment does not limit this.

[0128] As an optional implementation method, the executing entity of the embodiment of the present application can send a laser beam to the target vehicle through a lidar sensor, receive the reflected light of the target vehicle, and calculate the relative distance and relative speed between the vehicle and the target vehicle based on the time difference between the emitted light and the reflected light, thereby obtaining a relative distance set and a relative speed set.

[0129] Among them, for the measurement of the acceleration of the target vehicle, the execution entity of the embodiment of the present application can perform continuous relative distance and speed measurements on the target vehicle through a laser radar, and determine the acceleration of the target vehicle based on the speed change and time interval before and after, thereby obtaining the acceleration set of the vehicle within a preset time period.

[0130] Based on step 301 and step 302, the vehicle's speed set, acceleration set, steering angle set, relative distance set between the vehicle and the target vehicle, relative speed set, and anti-collision acceleration set of the target vehicle can be obtained. The execution entity of the embodiment of the present application can determine the above-mentioned speed set, acceleration set, steering angle set, relative distance set between the vehicle and the target vehicle, relative speed set, and anti-collision acceleration set of the target vehicle as a parameter value set of the vehicle within a preset time period.

[0131] Step 304: for each parameter value set, perform outlier processing, missing value processing, and normalization processing on the parameter values ​​in the parameter value set to obtain a new parameter value set including multiple new parameter values.

[0132] Step 305: merge multiple new parameter value sets to obtain a time parameter matrix, where the time parameter matrix includes new parameter values ​​of different operating parameters corresponding to each time point.

[0133] The following is a unified description of step 304 and step 305:

[0134] The above-mentioned outlier processing can be a Z-score analysis method, which identifies outliers by calculating the standard deviation distance (Z-score) of each data point from the mean. For example, data points with a Z-score value greater than 3 or less than -3 are regarded as outliers and deleted.

[0135] The missing value processing mentioned above means that when data at a certain moment is missing, linear interpolation can be used to fill it. Among them, linear interpolation is a data interpolation method, which can be used to estimate the unknown data values ​​between these points when discrete data points are known.

[0136] The above normalization process can be to use the minimum to maximum normalization method to scale the data to the range of [0, 1]. This method can first determine the minimum value (Xmin) and the maximum value (Xmax) of a set of data. For each value X in the data set, the normalized new value X' can be calculated by the following formula (1), so that each value in the data set is scaled to the range of [0, 1]:

[0137] X'=(X-Xmin) / (Xmax-Xmin) Formula (1)

[0138] Wherein, the X' is the new value after normalization, the X is the value to be normalized, the Xmin is the minimum value in the data set, and the Xmax is the maximum value in the data set.

[0139] In some embodiments of the present application, after obtaining the above-mentioned parameter value sets, the parameter values ​​in each of the above-mentioned parameter value sets can be processed for outliers, missing values, and normalization to obtain a new parameter value set containing multiple new parameter values.

[0140] Afterwards, a plurality of the above-mentioned new parameter value sets may be combined to obtain a time parameter matrix, and the above-mentioned time parameter matrix may include new parameter values ​​of different operating parameters corresponding to each time point within a preset time period.

[0141] For example, assuming that the above new parameter value sets may include the following three: {a1, a2, a3, a4}, {b1, b2, b3, b4}, {c1, c2, c3, c4}, and continuing to assume that the time points corresponding to the above parameter value sets are t1, t2, t3, and t4, respectively, after merging the above new parameter value sets, the following time parameter matrix can be obtained:

[0142]

[0143] The technical solution provided by the embodiment of the present application measures the speed set, acceleration set, and steering angle set of the vehicle traveling in a preset time period through a plurality of pre-installed different parameter sensors, determines the relative distance set, relative speed set, and anti-collision acceleration set between the vehicle and the target vehicle through a preset radar sensor, determines the above speed set, acceleration set, steering angle set, relative distance set, relative speed set, and anti-collision acceleration set as parameter value sets, performs outlier processing, missing value processing, and normalization processing on the parameter values ​​in the above parameter values ​​for each parameter value set, obtains a new parameter value set containing a plurality of new parameter values, merges the plurality of new parameter value sets, obtains a time parameter matrix, and the above time parameter matrix includes new parameter values ​​of different operating parameters corresponding to each time point. This technical solution can avoid the situation where one of the sensors is damaged and the anti-collision event cannot be predicted by measuring with different sensors. It can reduce the error of the measured parameter values ​​by performing outlier processing, missing value processing, and normalization processing on multiple parameter value sets, and achieve more accurate determination of the time parameter matrix of the vehicle traveling within a preset time period, thereby improving the accuracy of anti-collision event prediction.

[0144] See also Figure 4 , which is a flow chart of an embodiment of another vehicle anti-collision warning method provided in an embodiment of the present application. Figure 4 The process shown in Figure 1 Based on the process shown in the figure, it describes how to determine multiple warning strategies and the probability value corresponding to each warning strategy through the anti-collision warning model when the anti-collision warning model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. Figure 4 As shown, the process may include the following steps:

[0145] Step 401: Obtain a time parameter matrix through an input layer, and perform convolution calculation on the time parameter matrix through a convolution layer to obtain multiple convolution features.

[0146] In some embodiments of the present application, the above-mentioned anti-collision warning model can be as follows: Figure 5 As shown, see Figure 5 , is a schematic diagram of the structure of an anti-collision warning model provided in an embodiment of the present application. Figure 5 As shown, the anti-collision warning model may include an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0147] Based on this, after obtaining the above-mentioned time parameter matrix, the executing entity of the embodiment of the present application can input the above-mentioned time parameter matrix into the above-mentioned anti-collision warning model. The anti-collision warning model can obtain the above-mentioned time parameter matrix through the input layer, and perform convolution calculation on the above-mentioned time parameter matrix through the convolution layer to obtain multiple convolution features.

[0148] As an optional implementation, the convolution layer may include N convolution kernels, where N is a positive integer. For example, the convolution layer may include 32 convolution kernels. Based on this, when performing convolution calculation on the time parameter matrix through the convolution layer, the convolution kernel may be slid in the time parameter matrix according to a preset step size for each convolution kernel. The preset step size refers to the distance that the preset convolution kernel slides each time in the time parameter matrix.

[0149] After that, for the submatrix corresponding to each sliding of the convolution kernel, the submatrix can be convolved to obtain the sub-features corresponding to the submatrix. The above submatrix is ​​the matrix mapped by the convolution kernel in the time parameter matrix.

[0150] For example, assuming that the above convolution kernel can be a 3*3 structure, and continuing to assume that the time parameter matrix is ​​a 5*5 matrix, and continuing to assume that the above preset step size is 1, then the submatrix obtained by the first mapping of the convolution kernel in the time parameter matrix is ​​the 3*3 submatrix starting from the upper left corner in the time parameter matrix. After moving according to the above preset step size, the submatrix corresponding to the first three rows and the 2nd to the 4th columns in the time parameter matrix can be obtained. After the convolution kernel is moved horizontally, it can move downward until it moves to the lower right corner of the time parameter matrix, thereby dividing the time parameter matrix into multiple submatrices.

[0151] As an exemplary implementation, when performing convolution calculation on a submatrix to obtain a sub-feature corresponding to the submatrix, a weight matrix and a first bias term corresponding to the convolution kernel may be obtained. Afterwards, the elements in the submatrix may be weighted and summed according to the weight matrix to obtain an initial convolution value. Finally, the initial convolution value may be added to the first bias term to obtain the sub-feature corresponding to the submatrix. Among them, the weight matrix and the first bias term are values ​​obtained by pre-training the convolution layer, and the weights in the weight matrix correspond one-to-one to the elements in the submatrix.

[0152] For example, assuming that the convolution kernel is a 3*3 matrix, then the submatrix obtained by the above convolution kernel is also a 3*3 matrix. Correspondingly, the above weight matrix is ​​also a 3*3 matrix. Based on this, the corresponding sub-features can be obtained for the submatrix by the following formula (II):

[0153]

[0154] Among them, the above y is the above sub-feature, and the above wij is the weight of the i-th row and j-th column in the weight matrix. ij is the element in the i-th row and j-th column in the submatrix, and the above b is the above-mentioned first bias term.

[0155] Finally, the matrix composed of multiple sub-features can be determined as the convolution feature after the convolution kernel convolves the time parameter matrix.

[0156] According to the above steps, each convolution kernel can obtain a convolution feature corresponding to a time parameter matrix, so N convolution features can be obtained.

[0157] Step 402: extract multiple convolutional features through a pooling layer to obtain a pooling feature matrix.

[0158] The above pooling layer is used to reduce the data dimension and the amount of calculation while retaining the main features. Based on this, the pooling layer can be used to extract the multiple convolutional features obtained above to obtain a pooling feature matrix after reducing the dimension.

[0159] The above pooling feature matrix is ​​a feature matrix after main feature extraction of multiple convolutional features.

[0160] In an embodiment of the present application, in order to extract the main features in the convolution features, the anti-collision warning model in the embodiment of the present application can extract the convolution features through a pooling layer after convolving the time parameter matrix to obtain a corresponding pooling feature matrix.

[0161] As an optional implementation, a pooling window of the pooling layer may be determined, and for each convolution feature, the convolution feature may be divided into a plurality of sub-window matrices according to the pooling window. The pooling window is a pre-set unit matrix for extracting features, for example, the pooling window may be 2*2, and on this basis, the time parameter matrix may be divided according to the pooling window to obtain a plurality of 2*2 sub-window matrices.

[0162] Afterwards, the maximum value of each sub-window matrix can be obtained, and the matrix composed of the maximum values ​​can be used as the sub-pooling feature matrix corresponding to the convolution feature. Finally, multiple sub-pooling feature matrices can be combined to obtain a pooling feature matrix with the number of channels equal to the number of convolution features (such as N above).

[0163] For example, assuming that the pooling window is 2*2 and the convolution feature is 4*4, the pooling window can divide the above time parameter matrix into 4 sub-window matrices, and for each window matrix, the maximum value in the window matrix can be extracted, and finally a 2*2 sub-pooling feature matrix is ​​obtained. Continuing to assume that the number of the above convolution features is N, after combining the N sub-pooling feature matrices, a pooling feature matrix with a shape of 2*2*N can be obtained.

[0164] Step 403: Perform full connection calculation on the above pooling feature matrix through a fully connected layer to obtain vector features.

[0165] The above fully connected layer is used to further integrate the pooled feature matrix output by the pooling layer.

[0166] In some embodiments of the present application, the pooled feature matrix output by the pooling layer can be further integrated through a fully connected layer and mapped to the output layer.

[0167] As an optional implementation, the fully connected layer may include M neurons, where M is a positive integer, such as 64. Based on this, the pooled feature matrix may be fully connected by M neurons to obtain corresponding vector features.

[0168] As an exemplary implementation method, the above-mentioned pooling feature matrix can be converted into a one-dimensional vector. For example, the above-mentioned pooling feature matrix is ​​(H, W, C), where H is the number of rows, W is the number of columns, and C is the number of channels. Then, converting it into a one-dimensional vector can obtain a vector of length H*W*C.

[0169] Based on this, the above one-dimensional vector can be input into the fully connected layer, and for each neuron in the fully connected layer, the first weight vector and the second bias item corresponding to the neuron can be obtained. After that, the one-dimensional vector can be weighted summed according to the first weight vector to obtain the initial sub-vector feature, and the above sub-vector feature can be added to the second bias item to obtain the sub-vector feature.

[0170] For example, suppose the one-dimensional vector above is (z1, z2, ..., z n ), the first weight vector is (w 11 , w 12 ,…,w 1n ), the second bias term is b1, then the initial sub-vector feature can be calculated by the following formula (III):

[0171]

[0172] Among them, the above d1 is the initial sub-vector feature, the above w 1i is the i-th weight vector, the above z 1i is the i-th vector element in the one-dimensional vector, and the above b1 is the above second bias term.

[0173] Finally, the sub-vector features corresponding to each neuron in the above M neurons can be combined to obtain the above vector features, and the vector features are fully connected to the neurons in the output layer. For example, the initial sub-vector feature obtained by the first neuron is d1, and the initial sub-vector feature obtained by the Mth neuron is d M, then the vector feature can be (d1, d2, ... d M ).

[0174] Step 404: Probability calculation is performed on the vector features through the output layer to obtain the probability value corresponding to each of the pre-set multiple warning strategies, and multiple warning strategies and the probability value corresponding to each warning strategy are output.

[0175] The above-mentioned early warning strategies are possible early warning strategies that are pre-set in the output layer.

[0176] In some embodiments of the present application, multiple warning strategies can be set in advance in the output layer, and the probability value corresponding to each warning strategy under the current operating condition of the vehicle can be determined through the above-mentioned output layer, and multiple warning strategies corresponding to the current operating condition of the vehicle and the probability value corresponding to each warning strategy can be output.

[0177] As an optional implementation, the output layer may include T neurons, each neuron may correspond to an early warning strategy, and T is a positive integer.

[0178] Based on this, when performing probability calculation on the vector features output by the fully connected layer, the second weight vector and the third bias item corresponding to each neuron can be determined. The above weight vector and the third bias item are obtained by pre-training the output layer.

[0179] Afterwards, the vector features may be weighted and summed according to the second weight vector to obtain a first value, and the first value may be added to the third bias term to obtain a second value.

[0180] For example, suppose the output vector feature of the fully connected layer is D = (d1, d2, ..., d 64 ), the second weight vector k1 of the first neuron in the output layer = (k 11 , k 12 , …, k 1，64 ), the second value m1 of the first neuron in the output layer can be obtained by the following formula (IV):

[0181]

[0182] Among them, the above m1 is the second value, and the above k 1i is the i-th weight vector in the weight vector, the above d i is the i-th vector feature in the vector feature, and b2 is the third bias term.

[0183] Finally, the preset activation function can be used to calculate the T second values ​​to obtain the probability value corresponding to each second value, and the probability value is the probability value of the warning strategy corresponding to the second value.

[0184] As an exemplary implementation, the probability value of each second value corresponding to the early warning strategy can be determined by the following formula (V):

[0185]

[0186] The above j is the probability value corresponding to the j-th neuron, the above mj refers to the second value corresponding to the j-th neuron, and the above mn refers to the probability value corresponding to the n-th neuron, wherein in formula (V), the total number of neurons is 3 for example.

[0187] The technical solution provided by the embodiment of the present application is that, in the case where the anti-collision warning model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer, the time parameter matrix is ​​obtained through the input layer, and the time parameter matrix is ​​convoluted through the convolution layer to obtain multiple convolution features. The multiple convolution features are extracted through the pooling layer to obtain a pooling feature matrix, and the above pooling feature matrix is ​​fully connected through the fully connected layer to obtain vector features. The vector features are calculated by probability through the output layer to obtain the probability value corresponding to each of the multiple pre-set warning strategies, and multiple warning strategies and the probability value corresponding to each warning strategy are output. This technical solution can more quickly and accurately output multiple warning strategies of the vehicle under the current working conditions and the probability value corresponding to each warning strategy by performing convolution, feature extraction, feature integration, feature full connection, and calculation of the probability value of each warning strategy on the time parameter matrix through the anti-collision warning model.

[0188] See also Figure 6 , which is a flow chart of an embodiment of another vehicle anti-collision warning method provided in an embodiment of the present application. Figure 6 The process shown in Figure 4 Based on the process shown in the figure, this paper describes how the anti-collision warning model is trained. Figure 6 As shown, the process may include the following steps:

[0189] Step 601: Obtain a predetermined training sample set, the training sample set including multiple time parameter matrix samples and a probability value sample set corresponding to each time parameter matrix sample, the probability value sample set including the probability value corresponding to each warning strategy in multiple preset warning strategies.

[0190] Step 602: Select a target sample set with a preset data volume from the above training sample set.

[0191] Step 603: Use the target sample set to train the initial model to obtain an initial probability value set corresponding to each time parameter matrix sample output by the initial model, wherein the initial probability value set includes the initial probability value corresponding to each warning strategy in the preset multiple warning strategies.

[0192] The following is a unified description of steps 601 to 603:

[0193] The above-mentioned time parameter matrix sample refers to the correspondence between the parameter values ​​of a plurality of predetermined operating parameters of the vehicle traveling within a preset time period and time.

[0194] The above-mentioned probability value sample set is a plurality of pre-determined warning strategies that may be implemented by the vehicle corresponding to the above-mentioned time parameter matrix samples, and the probability value corresponding to each warning strategy.

[0195] In some embodiments of the present application, a training sample set for training an anti-collision warning model can be obtained in advance. Thereafter, in order to improve the training efficiency of the model, a target sample set with a preset data volume (for example, 64) can be selected from the above training sample set, and the initial model can be trained using the target sample set to obtain an initial probability value set corresponding to each time parameter matrix sample output by the initial model. The above initial probability value set includes the initial probability value corresponding to each warning strategy among the preset multiple warning strategies.

[0196] As an optional implementation method, the execution subject of the embodiment of the present application can obtain the training sample set input by the user through a preset visual interface.

[0197] As another optional implementation, the execution subject of the embodiment of the present application may obtain a pre-stored training sample set from a preset storage medium.

[0198] In addition, the initial model may be directly trained using the acquired training sample set, and the embodiments of the present application do not impose any restrictions on this.

[0199] As for how to use the target sample set to train the initial model, we will explain it in the following. Figure 7 The process shown is explained and will not be described in detail here.

[0200] Step 604: Determine whether the initial model meets the preset stop training condition based on the probability value sample set and the corresponding initial probability value set. If so, execute step 605; if not, execute step 606.

[0201] Step 605: determine the initial model as the anti-collision warning model.

[0202] Step 606: Optimize the parameters in the initial model using a preset optimization algorithm according to a preset step size to obtain a new initial model.

[0203] Step 607: determine the new initial model as the initial model and return to step 603.

[0204] The following is a unified description of steps 604 to 607:

[0205] The above-mentioned stopping training conditions refer to the pre-set conditions for stopping training, which may include but are not limited to: the number of training times reaches a preset value (for example, 200), the accuracy, recall rate, F1 value (the average of accuracy and recall rate) and other evaluation indicators of the initial probability value set output by the initial model are greater than the preset indicator threshold, etc.

[0206] In some embodiments of the present application, after the initial model is trained using the target sample set to obtain the initial probability value set corresponding to each time parameter matrix sample output by the initial model, it is possible to determine whether the initial model meets the preset stop training condition based on the probability value sample set corresponding to the target sample set and the corresponding initial probability value set, that is, to determine whether the initial model has met the preset stop training condition based on the probability value sample set and the initial probability value set corresponding to the same time parameter matrix sample.

[0207] As an optional implementation, it may be determined whether the current number of training times is greater than or equal to a preset number threshold. Optionally, when it is determined that the number of training times is greater than the number threshold, it is determined that the initial model meets a preset stop training condition.

[0208] As another optional implementation, the current accuracy, recall, and F1 value of the initial model may be determined based on the above probability value sample set and the initial probability value set. Optionally, when it is determined that the above accuracy, recall, and F1 value are all greater than their respective corresponding preset thresholds, it is determined that the initial model currently meets the preset stop training condition.

[0209] Optionally, when it is determined that a preset stop training condition is currently met, it can be determined that the initial model has completed training, and the initial model can be determined as the anti-collision warning model obtained through training.

[0210] Optionally, when it is determined that the preset stop training condition is not met at present, the parameters in the initial model (such as the weight parameters of each layer) can be optimized according to the preset step size using a preset optimization algorithm to obtain an optimized new initial model. Afterwards, the new initial model can be returned as the initial model to execute the above step 603, that is, the target sample set is reused to train the new initial model. Among them, the above optimization algorithm can be the Adam optimization algorithm, or other algorithms, which is not limited in the embodiment of the present application, and the above preset step size can be a pre-set parameter adjustment step size, such as 0.001, or other values, which is not limited in the embodiment of the present application.

[0211] The technical solution provided by the embodiment of the present application is to obtain a predetermined training sample set, select a target sample set of a preset data volume from the training sample set, and use the target sample set to train the initial model to obtain an initial probability value set corresponding to each time parameter matrix sample output by the initial model, wherein the initial probability value set includes the initial probability value corresponding to each warning strategy in the preset multiple warning strategies, and determine whether the initial model meets the preset stop training condition according to the probability value sample set and the corresponding initial probability value set. If so, the initial model is determined as an anti-collision warning model; if not, the parameters in the initial model are optimized according to the preset step length using the preset optimization algorithm to obtain a new initial model, and the new initial model is determined as the initial model to re-use the target sample set to train the new initial model. This technical solution realizes more efficient and accurate training of the anti-collision warning model by presetting the sample data volume of the training model, and in the process of training the initial model according to the target sample set, the initial model is continuously optimized using the preset optimization algorithm, and the parameters of the initial model are adjusted according to the preset step length.

[0212] See also Figure 7 , is a flow chart of an embodiment of a vehicle anti-collision warning method provided in an embodiment of the present application. Figure 7 The process shown in Figure 6 Based on the process shown in FIG. 1 , it describes how to train the initial model using the target sample set when the initial model includes an initial input layer, an initial convolutional layer, an initial pooling layer, an initial fully connected layer, and an initial output layer. Figure 7 As shown, the process may include the following steps:

[0213] Step 701: For each time parameter matrix sample in the target sample set, the time parameter matrix sample is obtained through the initial input layer, and the time parameter matrix sample is convolved through the initial convolution layer to obtain multiple initial convolution features.

[0214] In some embodiments of the present application, the initial model may include an initial input layer, an initial convolutional layer, an initial pooling layer, an initial fully connected layer, and an initial output layer.

[0215] Based on this, when the executing entity of the embodiment of the present application uses the above-mentioned time parameter matrix samples in the target sample set to train the initial model, the above-mentioned time parameter matrix samples can be input into the above-mentioned initial model. The initial model can obtain the above-mentioned time parameter matrix samples through the initial input layer, and perform convolution calculation on the above-mentioned time parameter matrix samples through the initial convolution layer to obtain multiple initial convolution features.

[0216] As an optional implementation, the initial convolution layer may include N initial convolution kernels, where N is a positive integer. For example, the initial convolution layer may include 32 initial convolution kernels. Based on this, when performing convolution calculation on the time parameter matrix samples through the initial convolution layer, the initial convolution kernel may be slid in the time parameter matrix samples according to a preset step length for each initial convolution kernel. The preset step length refers to the distance that the preset initial convolution kernel slides in the time parameter matrix samples each time.

[0217] Afterwards, for each sub-initial matrix corresponding to the sliding of the initial convolution kernel, the sub-initial matrix can be convolved to obtain the sub-initial features corresponding to the sub-initial matrix. The above sub-initial matrix is ​​the matrix mapped by the initial convolution kernel in the time parameter matrix sample.

[0218] For example, assuming that the above initial convolution kernel can be a 3*3 structure, and continuing to assume that the time parameter matrix sample is a 5*5 matrix, and continuing to assume that the above preset step size is 1, then the sub-initial matrix obtained by the first mapping of the initial convolution kernel in the time parameter matrix sample is the 3*3 sub-initial matrix starting from the upper left corner in the time parameter matrix sample. After moving according to the above preset step size, the sub-initial matrix corresponding to the first three rows and the 2nd to the 4th columns in the time parameter matrix sample can be obtained. And so on. After the initial convolution kernel is moved horizontally, it can be moved downward until it moves to the lower right corner of the time parameter matrix sample, thereby dividing the time parameter matrix sample into multiple sub-initial matrices.

[0219] As an exemplary implementation, when performing convolution calculation on the sub-initial matrix to obtain the sub-initial features corresponding to the sub-initial matrix, the initial weight matrix and the first initial bias term corresponding to the initial convolution kernel can be obtained. Afterwards, the elements in the sub-initial matrix can be weighted and summed according to the initial weight matrix to obtain the initial convolution value. Finally, the initial convolution value can be added to the first initial bias term to obtain the sub-initial features corresponding to the sub-initial matrix. Among them, the initial weight matrix and the first initial bias term are values ​​obtained by pre-training the initial convolution layer, and the weights in the initial weight matrix correspond one-to-one to the elements in the sub-initial matrix.

[0220] Finally, the matrix composed of multiple sub-initial features can be determined as the initial convolution feature after the initial convolution kernel convolves the time parameter matrix sample.

[0221] According to the above steps, each initial convolution kernel can obtain an initial convolution feature corresponding to a time parameter matrix sample, so N initial convolution features can be obtained.

[0222] Step 702: extract features from multiple initial convolutional features through an initial pooling layer to obtain an initial pooling feature matrix.

[0223] The above initial pooling layer is used to reduce the data dimension and the amount of calculation while retaining the main features. Based on this, the initial pooling layer can be used to extract features from the multiple initial convolutional features obtained above to obtain an initial pooling feature matrix after reducing the dimension.

[0224] The above initial pooling feature matrix is ​​a feature matrix after main features are extracted from multiple initial convolution features.

[0225] In an embodiment of the present application, in order to extract the main features in the initial convolution features, the initial model in the embodiment of the present application can extract the initial convolution features through the initial pooling layer after convolving the time parameter matrix samples to obtain the corresponding initial pooling feature matrix.

[0226] As an optional implementation, an initial pooling window of the initial pooling layer may be determined, and for each initial convolution feature, the initial convolution feature may be divided into a plurality of sub-initial window matrices according to the initial pooling window. The initial pooling window is a pre-set unit matrix for extracting features, for example, the initial pooling window may be 2*2, on this basis, the time parameter matrix samples may be divided according to the initial pooling window, and a plurality of 2*2 sub-initial window matrices may be obtained.

[0227] Afterwards, the maximum value of each sub-initial window matrix can be obtained, and the matrix composed of the maximum values ​​can be used as the sub-initial pooling feature matrix corresponding to the initial convolution feature. Finally, multiple sub-initial pooling feature matrices can be combined to obtain an initial pooling feature matrix with the number of channels equal to the number of initial convolution features (such as N above).

[0228] For example, assuming that the initial pooling window is 2*2 and the initial convolution feature is 4*4, the initial pooling window can divide the above time parameter matrix samples into 4 sub-initial window matrices, and for each initial window matrix, the maximum value in the initial window matrix can be extracted, and finally a 2*2 sub-initial pooling feature matrix is ​​obtained. Continuing to assume that the number of the above initial convolution features is N, after combining the N sub-initial pooling feature matrices, an initial pooling feature matrix with a shape of 2*2*N can be obtained.

[0229] Step 703: Perform full connection calculation on the initial pooling feature matrix through the initial fully connected layer to obtain initial vector features.

[0230] The above initial fully connected layer is used to further integrate the initial pooling feature matrix output by the initial pooling layer.

[0231] In some embodiments of the present application, the initial pooling feature matrix output by the initial pooling layer can be further integrated through the initial fully connected layer and mapped to the initial output layer.

[0232] As an optional implementation, the initial fully connected layer may include M initial neurons, where M is a positive integer, such as 64. Based on this, the initial pooling feature matrix may be fully connected by the M initial neurons to obtain the corresponding initial vector features.

[0233] As an exemplary implementation method, the above-mentioned initial pooling feature matrix can be converted into an initial one-dimensional vector. For example, the above-mentioned initial pooling feature matrix is ​​(H, W, C), where H is the number of rows, W is the number of columns, and C is the number of channels. Then, converting it into an initial one-dimensional vector can obtain an initial vector of length H*W*C.

[0234] Based on this, the initial one-dimensional vector can be input into the initial fully connected layer, and for each initial neuron in the initial fully connected layer, the first initial weight vector and the second initial bias item corresponding to the initial neuron can be obtained. Afterwards, the initial one-dimensional vector can be weighted summed according to the first initial weight vector to obtain the first initial sub-vector feature, and the first sub-vector feature can be added to the second initial bias item to obtain the initial sub-vector feature.

[0235] Finally, the initial sub-vector features corresponding to each of the M initial neurons can be combined to obtain the initial vector features, and the initial vector features are fully connected to the initial neurons in the initial output layer. For example, the initial sub-vector feature obtained by the first initial neuron is d1, and the initial sub-vector feature obtained by the Mth initial neuron is d M , then the initial vector feature can be (d1, d2, ... d M ).

[0236] Step 704: Probability calculation is performed on the initial vector features through the initial output layer to obtain the initial probability value corresponding to each of the pre-set multiple early warning strategies, and multiple early warning strategies and the initial probability value corresponding to each early warning strategy are output.

[0237] The above-mentioned early warning strategies are possible early warning strategies that are pre-set in the initial output layer.

[0238] In some embodiments of the present application, multiple warning strategies can be set in advance in the initial output layer, and the initial probability value corresponding to each warning strategy under the current operating condition of the vehicle can be determined through the above-mentioned initial output layer, and multiple warning strategies corresponding to the current operating condition of the vehicle and the initial probability value corresponding to each warning strategy can be output.

[0239] As an optional implementation, the initial output layer may include T initial neurons, each initial neuron may correspond to an early warning strategy, and T is a positive integer.

[0240] Based on this, when the probability calculation is performed on the initial vector features output by the initial fully connected layer, the second initial weight vector and the third initial bias item corresponding to each initial neuron can be determined. The second initial weight vector and the third initial bias item are both obtained by pre-training the initial output layer.

[0241] Afterwards, the initial vector features may be weighted and summed according to the second initial weight vector to obtain a first initial value, and the first initial value may be added to the third initial bias term to obtain a second initial value.

[0242] Finally, the preset activation function can be used to calculate the T second initial values ​​to obtain the initial probability value corresponding to each second initial value, and the initial probability value is the initial probability value of the early warning strategy corresponding to the second initial value.

[0243] As an exemplary implementation, the total value of all current second initial values ​​may be determined, and then each second initial value may be divided by the total value to obtain an initial probability value corresponding to each second initial value.

[0244] The technical solution provided by the embodiment of the present application is that, when the anti-collision warning model includes an initial input layer, an initial convolution layer, an initial pooling layer, an initial fully connected layer, and an initial output layer, the time parameter matrix samples are obtained through the initial input layer, and the time parameter matrix samples are convoluted by the initial convolution layer to obtain multiple initial convolution features, the initial pooling layer performs feature extraction on the multiple initial convolution features to obtain an initial pooling feature matrix, the initial fully connected layer performs full connection calculation on the above initial pooling feature matrix to obtain initial vector features, and the initial output layer performs probability calculation on the initial vector features to obtain the initial probability value corresponding to each of the multiple pre-set warning strategies, and outputs multiple warning strategies and the initial probability value corresponding to each warning strategy. This technical solution can train the initial model more quickly and accurately by performing operations such as convolution, feature extraction, feature integration, feature full connection, and calculation of the probability value of each warning strategy on the time parameter matrix samples through the initial model, thereby obtaining a more accurate anti-collision warning model.

[0245] See also Figure 8 , is a block diagram of an embodiment of a vehicle anti-collision warning device provided in an embodiment of the present application. Figure 8 As shown, the device may include:

[0246] An acquisition module 81 is used to acquire a parameter value set corresponding to each of the multiple operating parameters of the vehicle during a preset time period; the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points;

[0247] A preprocessing module 82, used for preprocessing the plurality of parameter value sets to obtain a time parameter matrix;

[0248] An input module 83 is used to input the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each of the warning strategies;

[0249] The determination module 84 is used to determine the current target warning strategy for the vehicle from multiple warning strategies according to the probability value, and issue a warning prompt to the vehicle according to the target warning strategy.

[0250] like Fig. 9 As shown, it is a schematic diagram of the structure of a vehicle provided in an embodiment of the present application, including a processor 91, a communication interface 92, a memory 93 and a communication bus 94, wherein the processor 91, the communication interface 92, and the memory 93 communicate with each other through the communication bus 94.

[0251] Memory 93, used for storing computer programs;

[0252] In one embodiment of the present application, the processor 91 is used to execute the program stored in the memory 93 to implement the anti-collision warning method for a vehicle provided by any one of the aforementioned method embodiments, including:

[0253] Obtaining a parameter value set corresponding to each of the multiple operating parameters of the vehicle during a preset time period; the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points;

[0254] Preprocessing the plurality of parameter value sets to obtain a time parameter matrix;

[0255] Inputting the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each of the warning strategies;

[0256] According to the probability value, a current target warning strategy for the vehicle is determined from a plurality of warning strategies, and a warning prompt is issued to the vehicle according to the target warning strategy.

[0257] The embodiment of the present application also provides a computer program product, which, when executed on a computer, enables the computer to execute any of the above-mentioned anti-collision warning methods.

[0258] An embodiment of the present application further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the vehicle anti-collision warning method provided in any of the aforementioned method embodiments are implemented.

[0259] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0260] Through the text of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0261] It should be understood that the terms used in the text are only for the purpose of the specific example implementation of the text, and are not intended to be limited. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used in the text can also be represented to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of the stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or their combinations. The method steps, processes, and operations of the text in the text are not interpreted as necessarily requiring them to be performed in the specific order of the text or description, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps can be used.

[0262] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can 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 range consistent with the principles and novel features applied for herein.

Claims

1. A vehicle anti-collision warning method, characterized in that: The method comprises: Obtaining a parameter value set corresponding to each of the multiple operating parameters of the vehicle during a preset time period; the parameter value set includes parameter values ​​of the corresponding operating parameters at different time points; Preprocessing the plurality of parameter value sets to obtain a time parameter matrix; Inputting the time parameter matrix into a pre-trained anti-collision warning model to obtain a plurality of warning strategies output by the anti-collision warning model and a probability value corresponding to each of the warning strategies; According to the probability value, a current target warning strategy for the vehicle is determined from a plurality of warning strategies, and a warning prompt is issued to the vehicle according to the target warning strategy.

2. The method according to claim 1, characterized in that: The operating parameters include operating parameters of the vehicle, operating parameters of a target vehicle, and operating parameters between the vehicle and the target vehicle, the target vehicle being the vehicle closest to the vehicle; The step of obtaining a plurality of operating parameters of the vehicle during the preset time period, wherein each of the operating parameters corresponds to a parameter value set, includes: Respectively measuring a speed set, an acceleration set, and a steering angle set of the vehicle traveling within the preset time period; Determining a relative distance set and a relative speed set between the vehicle and the target vehicle, and a collision avoidance acceleration set of the target vehicle; The speed set, the acceleration set, the steering angle set, the relative distance set, the relative speed set, and the anti-collision acceleration set are all determined as the parameter value set.

3. The method according to claim 1, characterized in that The anti-collision warning model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer; the anti-collision warning model outputs multiple warning strategies and the probability value corresponding to each warning strategy in the following manner: Acquire the time parameter matrix through the input layer, and perform convolution calculation on the time parameter matrix through the convolution layer to obtain multiple convolution features; Performing feature extraction on the multiple convolutional features through the pooling layer to obtain a pooling feature matrix; Performing full connection calculation on the pooling feature matrix through the fully connected layer to obtain vector features; The output layer performs probability calculation on the vector features to obtain a probability value corresponding to each of a plurality of pre-set early warning strategies, and outputs a plurality of the early warning strategies and the probability value corresponding to each of the early warning strategies.

4. The method according to claim 3, characterized in that: The convolution layer includes N convolution kernels, where N is a positive integer. The convolution layer performs convolution calculation on the time parameter matrix to obtain multiple convolution features, including: For each of the convolution kernels, sliding the convolution kernel in the time parameter matrix according to a preset step size; For the submatrix corresponding to each sliding of the convolution kernel, convolution calculation is performed on the submatrix to obtain the sub-features corresponding to the submatrix; A matrix composed of multiple sub-features is determined as a convolution feature after the convolution kernel convolves the time parameter matrix.

5. The method according to claim 4, characterized in that The performing convolution calculation on the sub-matrix to obtain sub-features corresponding to the sub-matrix includes: Obtaining a weight matrix and a first bias item corresponding to the convolution kernel; Performing weighted summation on the elements in the submatrix according to the weight matrix to obtain an initial convolution value; The initial convolution value is added to the first bias term to obtain a sub-feature corresponding to the sub-matrix.

6. The method according to claim 3, characterized in that: The step of extracting features from the plurality of convolutional features through the pooling layer to obtain a pooling feature matrix includes: Determining a pooling window of the pooling layer; For each of the convolutional features, dividing the convolutional features into a plurality of sub-window matrices according to the pooling window; Obtain the maximum value in each of the sub-window matrices, and use the matrix composed of the maximum values ​​as the sub-pooling feature matrix corresponding to the convolution feature; Combine the multiple sub-pooling feature matrices to obtain a pooling feature matrix whose number of channels is the number of the convolution features.

7. The method according to claim 3, characterized in that The fully connected layer includes M neurons, where M is a positive integer. The fully connected layer performs a fully connected calculation on the pooled feature matrix to obtain a vector feature, including: Convert the pooled feature matrix into a one-dimensional vector; For each of the neurons, obtaining a first weight vector and a second bias item corresponding to the neuron; According to the first weight vector, performing weighted summation on the one-dimensional vector to obtain an initial sub-vector feature; Adding the initial sub-vector feature to the second bias term to obtain a sub-vector feature; The sub-vector features corresponding to each of the M neurons are combined to obtain the vector feature.

8. The method according to claim 3, characterized in that The output layer includes T neurons, each of which corresponds to an early warning strategy, and T is a positive integer. The probability calculation of the vector feature is performed through the output layer to obtain a probability value corresponding to each early warning strategy in a plurality of pre-set early warning strategies, including: For each of the neurons, determining a second weight vector and a third bias item corresponding to the neuron; performing weighted summation on the vector features according to the second weight vector to obtain a first value; Adding the first value to the third bias term to obtain a second value; The T second values ​​are calculated using a preset activation function to obtain a probability value corresponding to each second value.

9. A computer program product, characterized in that When the method is run on a computer, the computer is enabled to execute the anti-collision warning method described in any one of claims 1 to 8.

10. A vehicle, characterized in that: include: A processor and a memory, wherein the processor is used to execute a vehicle anti-collision warning program stored in the memory to implement the vehicle anti-collision warning method according to any one of claims 1 to 8.

Images