CN110667593B

CN110667593B - Driving reminding method, device and equipment based on deep learning and storage medium

Info

Publication number: CN110667593B
Application number: CN201910846569.2A
Authority: CN
Inventors: 肖爽
Original assignee: Ping An Property and Casualty Insurance Company of China Ltd
Current assignee: Ping An Property and Casualty Insurance Company of China Ltd
Priority date: 2019-09-06
Filing date: 2019-09-06
Publication date: 2022-05-17
Anticipated expiration: 2039-09-06
Also published as: CN110667593A

Abstract

The invention discloses a driving reminding method, a driving reminding device, driving reminding equipment and a storage medium based on deep learning, wherein the driving reminding method based on the deep learning comprises the following steps: acquiring a monitoring video for a vehicle driver in real time to acquire image data of the monitoring video; inputting the image data serving as input data into a preset three-dimensional convolutional neural network model to judge whether hidden danger behaviors exist in the vehicle driver or not; and if the vehicle driver has hidden danger behaviors, outputting preset reminding information to remind the driver. The invention solves the technical problems that in the prior art, the dangerous driving actions of a vehicle driver are more, the probability of traffic accidents is easily increased, and the driving safety experience is influenced.

Description

Driving reminding method, device and equipment based on deep learning and storage medium

Technical Field

The invention relates to the technical field of neural networks, in particular to a driving reminding method, device, equipment and storage medium based on deep learning.

Background

In the vehicle driving process, a plurality of vehicle drivers have action behaviors influencing driving safety more or less, such as fatigue driving and the like, in the prior art, the action behaviors influencing the driving safety are usually identified by collecting the vehicle driving time and the like, and the action behaviors influencing the driving safety are identified by collecting the vehicle driving time and the like, so that the technical problem of low identification accuracy exists.

Disclosure of Invention

The invention mainly aims to provide a driving reminding method, a driving reminding device, driving reminding equipment and a storage medium based on deep learning, and aims to solve the technical problem that in the prior art, the recognition accuracy is low for dangerous driving action behaviors of a vehicle driver.

In order to achieve the above object, the present invention provides a driving reminding method based on deep learning, which is applied to a browser, and comprises:

acquiring a monitoring video for a vehicle driver in real time to acquire image data of the monitoring video;

inputting the image data serving as input data into a preset three-dimensional convolutional neural network model to judge whether hidden danger behaviors exist in the vehicle driver or not;

and if the vehicle driver has hidden danger behaviors, outputting preset reminding information to remind the driver.

Optionally, the step of inputting the image data into a preset three-dimensional convolutional neural network model as input data to determine whether the vehicle driver has hidden danger behaviors includes:

inputting the image data serving as input data into a preset three-dimensional convolution neural network model so as to perform convolution and pooling alternative processing on the input data for preset times to obtain an initial processing result;

and classifying the initial processing result to judge whether the vehicle driver has hidden danger behaviors.

Optionally, the step of inputting the image data as input data into a preset three-dimensional convolutional neural network model to perform convolution and pooling alternative processing on the input data for a preset number of times to obtain an initial processing result includes:

inputting the image data serving as input data into a preset three-dimensional convolution neural network model;

acquiring a plurality of preset action behavior characteristics identified aiming at the hidden danger behaviors of the driver in the preset three-dimensional convolutional neural network model and weight matrixes respectively corresponding to the statistical characteristics corresponding to the preset action behavior characteristics;

performing filtering convolution processing on the image data according to the preset action behavior characteristics and the weight matrix to obtain a convolution processing result;

performing pooling treatment on the convolution treatment result to obtain a pooling treatment result;

and performing convolution and pooling alternative processing on the pooling processing result again for corresponding times according to the preset times to obtain an initial processing result.

Optionally, the step of performing pooling on the convolution processing result to obtain a pooled processing result includes:

dividing the convolution processing result into a plurality of image matrixes with the same size and preset sizes;

acquiring a maximum pixel value in the image matrix with the preset size, and replacing the image matrix with the maximum pixel value to obtain a new image matrix;

and setting the new image matrix as the pooling processing result.

Optionally, the step of classifying the initial processing result to determine whether the vehicle driver has a hidden danger behavior includes:

obtaining the type of the vehicle, and determining a classification threshold value of the vehicle according to the type of the vehicle;

comparing the processing value in the initial processing result with the classification threshold value;

and if the processing value is larger than the classification threshold value, determining that hidden danger behaviors exist in the vehicle driver.

Optionally, the obtaining the type of the vehicle, and the determining the classification threshold of the vehicle according to the type of the vehicle includes:

obtaining the type of the vehicle, and if the vehicle is a large passenger vehicle, determining the classification threshold value of the vehicle as a first classification threshold value;

if the vehicle is a bus, determining that the classification threshold value of the vehicle is a second classification threshold value;

and if the vehicle is a car, determining that the classification threshold of the vehicle is a third classification threshold, wherein the first classification threshold is smaller than the second classification threshold, and the second classification threshold is smaller than the third classification threshold.

Optionally, if the vehicle driver has hidden danger behaviors, outputting preset reminding information to remind the driver, wherein the step of reminding the driver comprises:

if the vehicle driver has hidden danger behaviors, determining the danger level of the hidden danger behaviors according to the difference value between the classification threshold value and the initial processing result;

and determining the reminding volume of the preset reminding information according to the danger level so as to remind the driver, wherein the higher the danger level is, the larger the reminding volume is.

The invention also provides a driving reminding device based on deep learning, which comprises the following components:

the system comprises an acquisition module, a display module and a control module, wherein the acquisition module is used for acquiring a monitoring video for a vehicle driver in real time so as to acquire image data of the monitoring video;

the judging module is used for inputting the image data serving as input data into a preset three-dimensional convolutional neural network model so as to judge whether hidden danger behaviors exist in the vehicle driver or not;

and the output module is used for outputting preset reminding information to remind the driver if the hidden danger behaviors exist in the vehicle driver.

Optionally, the determining module includes:

the input unit is used for inputting the image data serving as input data into a preset three-dimensional convolution neural network model so as to carry out convolution and pooling alternative processing on the input data for preset times to obtain an initial processing result;

and the judging unit is used for carrying out classification processing on the initial processing result so as to judge whether the vehicle driver has hidden danger behaviors.

Optionally, the input unit includes:

the input subunit is used for inputting the image data serving as input data into a preset three-dimensional convolutional neural network model;

the first obtaining subunit is configured to obtain a plurality of preset action behavior features identified for the hidden danger behavior of the driver in the preset three-dimensional convolutional neural network model, and weight matrices corresponding to the statistical characteristics corresponding to the plurality of preset action behavior features respectively;

the convolution subunit is used for performing filtering convolution processing on the image data according to the preset action behavior characteristics and the weight matrix to obtain a convolution processing result;

the pooling subunit is used for pooling the convolution processing result to obtain a pooling processing result;

and the initial processing subunit is used for performing convolution and pooling alternative processing on the pooling processing result again for corresponding times according to the preset times so as to obtain an initial processing result.

Optionally, the pooling subunit is configured to implement:

and setting the new image matrix as the pooling processing result.

Optionally, the determining unit includes:

the second acquiring subunit is used for acquiring the type of the vehicle and determining a classification threshold of the vehicle according to the type of the vehicle;

a comparison subunit, configured to compare the processing value in the initial processing result with the classification threshold;

and the determining subunit is used for determining that hidden danger behaviors exist in the vehicle driver if the processing value is greater than the classification threshold value.

Optionally, the second obtaining subunit is configured to implement:

Optionally, the output module includes:

the determining unit is used for determining the danger level of the hidden danger behaviors according to the difference value between the classification threshold value and the initial processing result if the hidden danger behaviors exist in the vehicle driver;

and the reminding unit is used for determining the reminding volume of the preset reminding information according to the danger level so as to remind the driver, wherein the higher the danger level is, the larger the reminding volume is.

In addition, to achieve the above object, the present invention also provides a driving reminding apparatus based on deep learning, including: a memory, a processor, a communication bus, and a deep learning based driving reminder stored on the memory,

the communication bus is used for realizing communication connection between the processor and the memory;

the processor is used for executing the deep learning-based driving reminding program to realize the following steps:

Further, to achieve the above object, the present invention also provides a storage medium storing one or more programs, the one or more programs being executable by one or more processors for:

The method comprises the steps of acquiring a monitoring video for a vehicle driver in real time to acquire image data of the monitoring video; inputting the image data serving as input data into a preset three-dimensional convolutional neural network model to judge whether hidden danger behaviors exist in the vehicle driver or not; and if the vehicle driver has hidden danger behaviors, outputting preset reminding information to remind the driver. In this embodiment, the identification of the hidden danger behaviors of the vehicle driver is performed by presetting the three-dimensional convolutional neural network model, and the monitoring video for the vehicle driver is acquired in real time in the application, so that the identification of the hidden danger behaviors of the vehicle driver is performed by presetting the three-dimensional convolutional neural network model, wherein the preset three-dimensional convolutional neural network model is a model which can accurately identify the hidden danger behaviors after training, therefore, in the application, the hidden danger behaviors of the vehicle driver can be accurately identified, such as playing a mobile phone, fighting or fatigue driving, and the like, and the technical problem of low identification accuracy of the hidden danger behaviors in the prior art is solved.

Drawings

Fig. 1 is a schematic flow chart of a driving reminding method based on deep learning according to a first embodiment of the present invention;

fig. 2 is a detailed flowchart of a step of inputting the image data as input data into a preset three-dimensional convolutional neural network model to determine whether a hidden danger behavior exists in the vehicle driver in a second embodiment of the deep learning-based driving reminding method of the present invention;

fig. 3 is a schematic device structure diagram of a hardware operating environment related to the method according to the embodiment of the present invention.

The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.

Detailed Description

It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.

The invention provides a driving reminding method based on deep learning, and in a first embodiment of the driving reminding method based on deep learning, referring to fig. 1, the driving reminding method based on deep learning comprises the following steps:

step S10, acquiring a monitoring video for a vehicle driver in real time to acquire image data of the monitoring video;

step S20, inputting the image data as input data into a preset three-dimensional convolutional neural network model to judge whether hidden danger behaviors exist in the vehicle driver;

and step S30, if the vehicle driver has hidden danger behaviors, outputting preset reminding information to remind the driver.

The method comprises the following specific steps:

in this embodiment, the vehicle is provided with a monitor for recording, the monitor starts the video recording function only after the vehicle is started, and the video recording function is for the vehicle driver, that is, the monitor records the video of the vehicle driver.

The monitoring in the vehicle sends the monitoring video of the vehicle driver acquired in real time to the driving reminding device based on the deep learning in the embodiment in real time, namely the driving reminding device based on the deep learning acquires the monitoring video for the vehicle driver in real time, the monitoring video is composed of a series of image data, and after the monitoring video is acquired, the image data of the monitoring video is acquired.

after the image data is acquired, if the data volume of the image data is larger than the preset data volume, firstly, the image data is segmented to obtain each segmented image block data, it should be noted that the preset data volume is changeable, a preset mapping relationship exists between the preset data volume and the cpu performance of the driving reminding device based on the deep learning, if the cpu performance of the driving reminding device based on the deep learning is improved, the preset data volume is increased, otherwise, the preset data volume is decreased.

The process of segmenting the image data can be carried out in the preset three-dimensional convolution neural network model or not.

In this embodiment, it should be noted that the convolutional neural network model is a trained model capable of performing hidden danger behavior determination, and the convolutional neural network model is a three-dimensional convolutional network model, that is, in this embodiment, the input parameters are related to the time parameter characteristic Z in addition to the spatial characteristics (X, Y) of the motion behavior.

The preset three-dimensional convolution neural network model training process comprises the following steps: acquiring a basic mathematical model, acquiring known action behavior data of a vehicle driver, wherein the action behavior data comprises hidden danger behaviors and non-hidden danger behaviors, extracting action behavior data of a preset proportion, such as 70% of action behavior data, from the action behavior data to serve as training data to train the basic mathematical model, taking the other 30% of action behavior data as test data to test the trained model to finally obtain a preset three-dimensional convolutional neural network model, extracting image characteristics of the hidden danger behaviors in advance in the training process, such as image characteristics of mobile phone playing behaviors in the driving process, image characteristics of beating behaviors and the like, acquiring various statistical characteristics in the image characteristics after obtaining the image characteristics of the mobile phone playing behaviors in the driving process, the image characteristics of the beating behaviors and the like, if the statistical characteristics of the image characteristics of the mobile phone playing behaviors in the driving process are obtained, the statistical characteristics specifically comprise image profiles of the mobile phone playing behaviors, image change trends and the like, after the statistical characteristics of the image characteristics are obtained, the weights of the statistical characteristics of the image characteristics are adjusted, wherein the adjustment of the weights of the statistical characteristics is sequentially adjusted according to expected difference values of prediction results output by corresponding models in the training process and actual results, so that a sub-preset three-dimensional convolutional neural network model corresponding to the mobile phone playing behaviors can be accurately predicted, similarly, training of sub-preset three-dimensional convolutional neural network models corresponding to other image characteristics is carried out, and the preset three-dimensional convolutional neural network model is finally obtained through combination.

Specifically, referring to fig. 2, the step of inputting the image data into a preset three-dimensional convolutional neural network model as input data to determine whether the vehicle driver has hidden danger behaviors includes:

step S21, inputting the image data as input data into a preset three-dimensional convolution neural network model so as to perform convolution and pooling alternative processing on the input data for preset times to obtain an initial processing result;

the biggest advantage of presetting the three-dimensional convolutional neural network is that weights are shared in convolutional layers, namely, the same weight library is used for all image block data.

And after the image data is input into a preset three-dimensional convolution neural network model as input data, performing convolution and pooling alternative processing on image block data in the image data by the preset three-dimensional convolution neural network model for preset times to obtain an initial processing result.

In this embodiment, after the image data is input into a preset three-dimensional convolutional neural network model as input data, the image data is alternately and circularly processed for preset times, such as convolution and pooling, to obtain an initial processing result.

Specifically, the step of inputting the image data into a preset three-dimensional convolutional neural network model as input data to perform convolution and pooling alternative processing on the input data for a preset number of times to obtain an initial processing result includes:

step A1, inputting the image data as input data into a preset three-dimensional convolution neural network model;

step A2, obtaining a plurality of preset action behavior characteristics identified aiming at the hidden danger behaviors of the driver in the preset three-dimensional convolutional neural network model and weight matrixes respectively corresponding to the statistical characteristics corresponding to the preset action behavior characteristics;

the convolution process can be understood as: the statistical characteristics of one part of the image features are the same as those of other parts, namely, the statistical characteristics learned in the part can also appear in the other part, so that the learned statistical characteristics are used as a detector and applied to any part of the image features, namely, the statistical characteristics learned by the small-range image are convoluted with the image features of the original large-size image, and mathematically, the convolution can be that a characteristic matrix of the corresponding image is multiplied by a plurality of detection matrixes in advance to obtain a convolution processing result.

In the embodiment, a plurality of preset action behavior characteristics, which are identified aiming at the hidden danger behaviors of the driver, in the preset three-dimensional convolutional neural network model are obtained, wherein the preset action behavior characteristics comprise a play mobile phone behavior characteristic, a beating behavior characteristic and a fatigue driving behavior characteristic, after the play mobile phone behavior characteristic, the beating behavior characteristic and the fatigue driving behavior characteristic are obtained, the pre-stored statistical characteristics of the behavior characteristics are obtained, and the pre-stored weight matrix of the statistical characteristics is obtained;

step A3, performing filtering convolution processing on the image data according to the preset action behavior characteristics and the weight matrix to obtain a convolution processing result;

and performing filtering convolution processing on the image data according to the preset action behavior characteristics and the weight matrix, namely multiplying a pixel matrix corresponding to the image data by a pixel matrix with statistical characteristics corresponding to the preset action behavior characteristics according to the weight matrix, and finally summing weights to obtain a convolution processing result.

Step A4, performing pooling treatment on the convolution treatment result to obtain a pooling treatment result;

and after the convolution processing result is obtained, performing pooling treatment on the convolution processing result, wherein the pooling treatment comprises maximum pooling and mean pooling, and after the convolution processing result is subjected to pooling, obtaining a pooling treatment result.

Specifically, the step of performing pooling processing on the convolution processing result to obtain a pooling processing result includes:

a step B1 of dividing the convolution processing result into a plurality of image matrices of a preset size that are uniform in size;

the specific pooling process is as follows: and dividing the convolution processing result into a plurality of image matrixes with the same size and preset size, such as a plurality of 3-by-3-dimensional image matrixes.

Step B2, obtaining the maximum pixel value in the image matrix with the preset size, and replacing the image matrix with the maximum pixel value to obtain a new image matrix;

obtaining a maximum pixel value in the image matrix with the preset size, and replacing the maximum pixel value with the image matrix with the preset size to obtain a new image matrix, wherein if the maximum pixel value in the image matrix with 3 x 3 dimensions is 1, the image matrix with 3 x 3 dimensions is replaced with 1, and a convolution processing result includes a plurality of image matrices with 3 x 3 dimensions, so that a new image matrix can be obtained finally.

Step B3, setting the new image matrix as the pooling result.

And setting the new image matrix as the pooling processing result.

And step A5, performing convolution and pooling alternative processing on the pooling processing result again for corresponding times according to the preset times to obtain an initial processing result.

The above-mentioned a1-a4 is an alternative process of convolution and pooling, and in this embodiment, a preset number of times of convolution and pooling are required to obtain the initial processing result.

And step S22, classifying the initial processing result to judge whether the vehicle driver has hidden danger behaviors.

After the initial processing result is obtained, the initial processing result is classified to judge whether hidden danger behaviors exist in the vehicle driver, if the initial processing result is larger than a certain threshold value, the hidden danger behaviors exist in the vehicle driver, and if the initial processing result is smaller than the certain threshold value, the hidden danger behaviors do not exist in the vehicle driver.

In this embodiment, if the vehicle driver has a hidden danger behavior, a preset reminding message is output to remind the driver, where the preset reminding message may be content of "please notice that there is a hidden danger behavior now, which affects driving safety" and the like.

Further, the present invention provides another embodiment of the driving reminding method based on deep learning, in this embodiment, the step of classifying the initial processing result to determine whether the hidden danger behavior exists for the vehicle driver includes:

step C1, obtaining the type of the vehicle, and determining the classification threshold value of the vehicle according to the type of the vehicle;

in this embodiment, after the initial processing result is obtained, the initial processing result is input into a preset classifier of a three-dimensional convolutional neural network, and before the classifier performs classification processing, a classification threshold of a vehicle needs to be determined according to the type of the vehicle, where the types of the vehicles are different and the classification threshold is different.

The step of obtaining the type of the vehicle and determining the classification threshold value of the vehicle according to the type of the vehicle comprises the following steps:

step D1, obtaining the type of the vehicle, and if the vehicle is a large passenger vehicle, determining the classification threshold of the vehicle as a first classification threshold;

if the current vehicle is detected to be a bus or a large passenger vehicle, determining that the classification threshold of the vehicle is a first classification threshold, wherein the first classification threshold may be 0.90.

Step D2, if the vehicle is a bus, determining that the classification threshold of the vehicle is a second classification threshold;

and if the current vehicle is detected to be the bus, determining that the classification threshold of the vehicle is a second classification threshold, wherein the second classification threshold can be 0.95.

Step D3, if the vehicle is a car, determining that the classification threshold of the vehicle is a third classification threshold, wherein the first classification threshold is smaller than the second classification threshold, and the second classification threshold is smaller than the third classification threshold.

If the vehicle is a car, determining that the classification threshold of the vehicle is a third classification threshold, where the second classification threshold may be 0.98, where the first classification threshold is smaller than the second classification threshold, and the second classification threshold is smaller than the third classification threshold, because detection of hidden danger behaviors of a vehicle with a large number of passengers is stricter, so as to ensure driving safety of the vehicle.

Step C2, comparing the processing value in the initial processing result with the classification threshold value;

after the initial processing result is obtained, the initial processing result is compared with the corresponding classification threshold value respectively, if the vehicle is a car, the initial processing result is compared with a third classification threshold value, if the vehicle is a bus, the initial processing result is compared with the second classification threshold value, if the vehicle is a bus or a large passenger carrying vehicle, the initial processing result is compared with the third classification threshold value.

And step C3, if the processing value is larger than the classification threshold value, determining that hidden danger behaviors exist in the vehicle driver.

And if the processing value is greater than the classification threshold value, determining that the hidden danger behaviors exist in the vehicle driver, and if the vehicle is a car, determining that the hidden danger behaviors exist in the vehicle driver if the initial processing result is greater than the third classification threshold value.

In the embodiment, by acquiring the type of the vehicle, the classification threshold of the vehicle is determined according to the type of the vehicle; comparing the processing value in the initial processing result with the classification threshold value; if the processing value is greater than the classification threshold value, it is determined that hidden danger behaviors exist in the vehicle driver, and in the embodiment, different classification threshold values are distinguished for different vehicles, so that the potential safety hazard reminding requirements of different vehicles are met better.

Further, the present invention provides another embodiment of the driving reminding method based on deep learning, in which if the vehicle driver has hidden danger behaviors, the step of outputting preset reminding information to remind the driver includes:

step S31, if the vehicle driver has hidden danger behaviors, determining the danger level of the hidden danger behaviors according to the difference value between the classification threshold value and the initial processing result;

and if the vehicle driver has hidden danger behaviors, determining the danger level of the hidden danger behaviors according to the difference value between the classification threshold value and the initial processing result, wherein the larger the difference value is, the higher the danger level of the hidden danger behaviors is.

Step S32, determining the reminding volume of the preset reminding information according to the danger level so as to remind the driver, wherein the higher the danger level is, the larger the reminding volume is.

And determining the reminding volume of the preset reminding information according to the danger level so as to remind the driver, wherein the higher the danger level is, the larger the reminding volume is, so that the reminding effect is improved.

In this embodiment, if the vehicle driver has a hidden danger behavior, determining a danger level of the hidden danger behavior according to a difference value between the classification threshold and the initial processing result; and determining the reminding volume of the preset reminding information according to the danger level so as to remind the driver, wherein the higher the danger level is, the larger the reminding volume is, so that the reminding effect is improved.

Referring to fig. 3, fig. 3 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present invention.

The driving reminding device based on deep learning in the embodiment of the invention can be a PC, and can also be terminal devices such as a smart phone, a tablet computer, an electronic book reader, an MP3(Moving Picture Experts Group Audio Layer III, dynamic video Experts compression standard Audio Layer 3) player, an MP4(Moving Picture Experts Group Audio Layer IV, dynamic video Experts compression standard Audio Layer 4) player, a portable computer and the like.

As shown in fig. 3, the driving reminding device based on deep learning may include: a processor 1001, such as a CPU, a memory 1005, and a communication bus 1002. The communication bus 1002 is used for realizing connection communication between the processor 1001 and the memory 1005. The memory 1005 may be a high-speed RAM memory or a non-volatile memory (e.g., a magnetic disk memory). The memory 1005 may alternatively be a memory device separate from the processor 1001 described above.

Optionally, the deep learning based driving reminding device may further include a target user interface, a network interface, a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. The target user interface may comprise a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional target user interface may also comprise a standard wired interface, a wireless interface. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface).

Those skilled in the art will appreciate that the deep learning based driving alert device configuration shown in fig. 3 does not constitute a limitation of the deep learning based driving alert device and may include more or fewer components than shown, or some components in combination, or a different arrangement of components.

As shown in fig. 3, a memory 1005, which is a kind of computer storage medium, may include therein an operating system, a network communication module, and a deep learning based driving reminder. The operating system is a program that manages and controls the hardware and software resources of the deep learning based driving alert device, supporting the operation of the deep learning based driving alert program as well as other software and/or programs. The network communication module is used for realizing communication among components in the memory 1005 and communication with other hardware and software in the deep learning based driving reminding device.

In the deep learning based driving alert device shown in fig. 3, the processor 1001 is configured to execute a deep learning based driving alert program stored in the memory 1005, and implement any one of the steps of the deep learning based driving alert method described above.

The specific implementation of the driving reminding device based on deep learning of the invention is basically the same as that of the driving reminding method based on deep learning, and is not described herein again.

Optionally, the determining module includes:

Optionally, the input unit includes:

Optionally, the pooling subunit is configured to implement:

and setting the new image matrix as the pooling processing result.

Optionally, the determining unit includes:

the second obtaining subunit is used for obtaining the type of the vehicle and determining a classification threshold value of the vehicle according to the type of the vehicle;

Optionally, the second obtaining subunit is configured to implement:

Optionally, the output module includes:

The specific implementation of the driving reminding device based on deep learning of the invention is basically the same as that of each embodiment of the driving reminding method based on deep learning, and is not described herein again.

The invention provides a storage medium, which stores one or more programs, which can be executed by one or more processors to implement the steps of any one of the deep learning-based driving reminding methods.

The specific implementation of the storage medium of the present invention is substantially the same as that of each embodiment of the driving reminding method based on deep learning, and is not described herein again.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the present specification and drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims

1. A driving reminding method based on deep learning is characterized by comprising the following steps:

obtaining the type of the vehicle, and determining a classification threshold of the vehicle according to the type of the vehicle;

if the processing value is larger than the classification threshold value, determining that hidden danger behaviors exist in the vehicle driver;

2. The deep learning-based driving reminding method according to claim 1, wherein the step of inputting the image data as input data into a preset three-dimensional convolution neural network model to perform convolution and pooling alternating processing for a preset number of times on the input data to obtain an initial processing result comprises:

acquiring a plurality of preset action behavior characteristics identified aiming at the hidden danger behaviors of the driver in the preset three-dimensional convolutional neural network model and weight matrixes respectively corresponding to the preset action behavior characteristics;

pooling the convolution processing result to obtain a pooled processing result;

3. The driving reminding method based on deep learning as claimed in claim 2, wherein the step of pooling the convolution processing result to obtain a pooled processing result comprises:

and setting the new image matrix as the pooling processing result.

4. The deep learning-based driving reminding method according to claim 1, wherein the step of obtaining the type of the vehicle and determining the classification threshold of the vehicle according to the type of the vehicle comprises:

5. The driving reminding method based on deep learning of claim 1, wherein the step of outputting preset reminding information to remind the driver if hidden danger behaviors exist in the vehicle driver comprises the following steps:

6. A driving reminding device based on deep learning is characterized in that the driving reminding device based on deep learning comprises:

the judging module is used for inputting the image data serving as input data into a preset three-dimensional convolution neural network model so as to carry out convolution and pooling alternative processing on the input data for preset times to obtain an initial processing result;

7. A deep learning based driving alert device, characterized in that the deep learning based driving alert device comprises: a memory, a processor, a communication bus, and a deep learning based driving reminder stored on the memory,

the processor is configured to execute the deep learning based driving alert procedure to implement the steps of the deep learning based driving alert method according to any one of claims 1 to 5.

8. A storage medium having stored thereon a deep learning based driving reminder, which when executed by a processor implements the steps of the deep learning based driving reminder method of any of claims 1-5.