CN-115935294-B - Method, device, equipment and medium for establishing driver intention recognition model

CN115935294BCN 115935294 BCN115935294 BCN 115935294BCN-115935294-B

Abstract

The invention provides a method, a device, equipment and a medium for establishing a driver intention recognition model, which relate to the technical field of intelligent interaction of automobiles and comprise the following steps of collecting camera image data, radar data and vehicle sensor data; and processing according to the acquired information to obtain sample data, putting the sample data into a training model to train to obtain a driver intention recognition model, and inputting new sample data to recognize to obtain a driver intention result after the driver intention recognition model is obtained. According to the method and the device for identifying the intention of the driver, the intention of the driver is accurately identified according to the intention identification model of the driver, and the practicability is very high.

Inventors

GU XIUYING
HE JING
Yuan Zhangkai
KUANG FEI

Assignees

重庆长安汽车股份有限公司

Dates

Publication Date: 20260505
Application Date: 20230103

Claims (8)

1. A method for creating a model for identifying the intention of a driver, comprising the steps of: acquiring vehicle interior and exterior data in different scenes, wherein the vehicle interior and exterior data comprise image data, radar data and vehicle sensor data, and the vehicle sensor data comprise a vehicle ignition state, a vehicle speed, a vehicle window state, an interior and exterior temperature, an interior and exterior humidity and a vehicle surrounding light brightness; Processing the image data to generate different visual perception events, and processing the radar data to generate perception targets and state data, wherein the visual perception events comprise whether a driver is in place, whether the driver is tired, whether a co-driver is someone, whether the driver smokes, and whether the driver drinks water; encoding the visual perception event, the perception target and the state data, and carrying out fusion processing on the visual perception event, the perception target and the state data and the vehicle sensor data to obtain sample data; Training the neural network fused with the attention mechanism by utilizing the sample data to acquire a driver intention recognition model; the step of encoding the visual perception event, the perception target and the state data and fusing the visual perception event, the perception target and the state data with the vehicle sensor data to obtain sample data comprises the following steps: Carrying out segmentation processing on all data according to the width of a preset time window; Processing each piece of data according to the data type, wherein the processing comprises taking an average value in a time window for continuous data; For the event data, taking the latest state in the time window; performing constraint processing on the processed data; And carrying out fusion processing on the data subjected to constraint processing to obtain the sample data.
2. The driver intent recognition model building method according to claim 1, characterized in that processing the image data to generate different visual perception events, in particular comprises: Processing the image data into pixels and frame rates supported by an image recognition algorithm; and inputting the processed image data into an image recognition algorithm to acquire the visual perception event.
3. The method for constructing a model for identifying driver's intention according to claim 1, wherein the radar data is processed to generate a perception target and status data, specifically comprising: Radar data is input to a radar recognition algorithm to acquire the perceived target and state data; the perceived target includes other vehicles in the vicinity, and the state includes the distance of the vehicle from the host vehicle.
4. The driver intention recognition model building method according to claim 1, wherein the constraint process includes a bucket process, a normalization process, or a normalization process.
5. The method for creating a model for identifying driver's intention as claimed in claim 1, wherein the step of encoding the visual perception event, the perception object and the status data comprises: Inquiring a pre-established corresponding table of the visual perception event to encode the visual perception event acquired by the image recognition algorithm; and inquiring a pre-established corresponding table of the perceived targets and states for encoding a plurality of perceived targets and states which are acquired by a radar identification algorithm and are closest to the vehicle.
6. A driver intention recognition model creation apparatus, characterized by comprising: The system comprises a data acquisition module, a data acquisition module and a data processing module, wherein the data acquisition module is used for acquiring vehicle interior and exterior data in different scenes, the vehicle interior and exterior data comprise image data, radar data and vehicle sensor data, and the vehicle sensor data comprise a vehicle ignition state, a vehicle speed, a vehicle window state and vehicle interior and exterior temperature or vehicle interior and exterior humidity; The processing data module is used for processing the image data to generate different visual perception events and processing the radar data to generate perception targets and state data, wherein the visual perception events comprise whether a driver is in place, whether the driver is tired, whether a person is driven by a assistant, whether the driver smokes and whether the driver drinks water; The fusion data module is used for encoding the visual perception event, the perception target and the state data and carrying out fusion processing on the visual perception event, the perception target and the state data and the vehicle sensor data so as to obtain sample data; The training data module is used for training the neural network fused with the attention mechanism by utilizing the sample data so as to acquire a driver intention recognition model; the step of encoding the visual perception event, the perception target and the state data and fusing the visual perception event, the perception target and the state data with the vehicle sensor data to obtain sample data comprises the following steps: Carrying out segmentation processing on all data according to the width of a preset time window; Processing each piece of data according to the data type, wherein the processing comprises taking an average value in a time window for continuous data; For the event data, taking the latest state in the time window; performing constraint processing on the processed data; And carrying out fusion processing on the data subjected to constraint processing to obtain the sample data.
7. An electronic device, comprising: One or more processors; Storage means for storing one or more programs that, when executed by the one or more processors, cause the electronic device to implement the driver intention recognition model building method of any one of claims 1 to 5.
8. A computer-readable storage medium having stored thereon computer-readable instructions which, when executed by a processor of a computer, cause the computer to perform the driver intention recognition model building method according to any one of claims 1 to 5.

Description

Method, device, equipment and medium for establishing driver intention recognition model Technical Field The invention relates to the technical field of intelligent interaction of automobiles, in particular to a method, a device, equipment and a medium for establishing an intention recognition model of a driver. Background In recent years, with the continuous development of deep learning technology, how to recognize semantics from texts or images has been developed to other fields, such as automobile driving, drivers and passengers can generate many scenes with clear intentions in each driving process from unlocking vehicles, such as unlocking vehicles by drivers and starting engines, the intention of the drivers is very likely to be driving vehicles out, the characteristics of the scenes can be recognized by changing all information of the vehicles in a period of time, the prior art generally defines limited scenes by a product manager, and the prediction of the user intention is realized by monitoring limited conditions. Disclosure of Invention In view of the above-mentioned drawbacks of the prior art, the present invention provides a method, apparatus, device and medium for creating a driver intention recognition model, so as to solve the above-mentioned technical problems. The invention provides a method for establishing a driver intention recognition model, which comprises the following steps: Acquiring vehicle interior and exterior data in different scenes, wherein the vehicle interior and exterior data comprises image data, radar data and vehicle sensor data; Processing the image data to generate different visual perception events, processing the radar data to generate perception targets and state data; encoding the visual perception event, the perception target and the state data, and carrying out fusion processing on the visual perception event, the perception target and the state data and the vehicle sensor data to obtain sample data; and training the neural network fused with the attention mechanism by using the sample data to acquire a driver intention recognition model. In one embodiment of the present invention, the processing the image data to generate different visual perception events specifically includes: Processing the image data into pixels and frame rates supported by an image recognition algorithm; inputting the processed image data into the image recognition algorithm to acquire the visual perception event; the visual perception event includes whether the driver is in place, whether the driver is tired, and whether the secondary drive is someone. In one embodiment of the present invention, the processing the radar data to generate the perception target and the state data specifically includes: radar data is input to the radar recognition algorithm to acquire the perceived target and state data; the perceived target includes other vehicles in the vicinity, and the state includes the distance of the vehicle from the host vehicle. In one embodiment of the present invention, the vehicle sensor data specifically includes: the vehicle sensor data comprises the ignition state of the vehicle, the vehicle speed, the state of the vehicle window and the temperature or humidity inside and outside the vehicle. In an embodiment of the present invention, the step of encoding the visual perception event, the perception target and the status data and performing fusion processing with the vehicle sensor data to obtain sample data includes: Carrying out segmentation processing on all data according to the width of a preset time window; Processing each piece of data according to the data type, wherein the processing comprises taking an average value in a time window for continuous data; For the event data, taking the latest state in the time window; performing constraint processing on the processed data; And carrying out fusion processing on the data subjected to constraint processing to obtain the sample data. In an embodiment of the present invention, the constraint processing includes a bucket processing, a normalization processing, or a normalization processing. In an embodiment of the present invention, the step of encoding the visual perception event, the perception object and the status data specifically includes: inquiring a pre-established corresponding table of the visual perception event to encode the visual perception event acquired by the image recognition algorithm; Inquiring a pre-established corresponding table of the perceived targets and states to encode a plurality of perceived targets and states which are acquired by the radar identification algorithm and are closest to the vehicle; in one embodiment of the present invention, the driver intention recognition model automatically learns and calculates the contribution of the input data to the output data by introducing an attention mechanism, and automatically recognizes the input data having a larger effect on each scene. The inventio