Visual blindness compensation algorithm model optimization method, system and vehicle based on DMS

Abstract

The present invention discloses a DMS-based visual blind spot algorithm model optimization method, system and vehicle, belonging to the field of vehicle blind spot monitoring technology. The present invention uses the DMS detection algorithm to judge the driver's driving steering intention, thereby judging the driver's blind spot field of view direction, and feeding back the driver's blind spot position to the whole vehicle, so that the first direction camera of the whole vehicle continues to call the first data format model weight of the surround perception algorithm for detection, and the second direction camera of the whole vehicle calls the second data format model weight of the surround perception algorithm for detection. The present invention can realize real-time and dynamic adjustment of the accuracy of the software detection algorithm, avoid wasting computing power on unnecessary tasks, and improve the flexibility of the whole vehicle algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle blind spot monitoring, and more specifically, to a method, system and vehicle for optimizing a visual blind spot compensation algorithm model based on a DMS (Driver Monitor System). Background Art

[0002] At present, the mainstream technology for autonomous driving and assisted driving is machine learning, especially deep learning. In terms of perception and detection tasks, deep learning models with neural networks as the core are mainly used to detect and perceive environmental factors such as pedestrians, vehicles, lane lines, etc. around the vehicle, laying a solid foundation for the subsequent control planning of the vehicle and providing solid protection.

[0003] However, there are currently two major challenges. On the one hand, as the number of on-board cameras continues to rise, the corresponding software algorithms that need to be embedded in the vehicle are becoming increasingly complex. This has caused the amount of data to be processed by existing domain control, on-board GPU and other processors and the complexity of software algorithms to rise sharply, and the requirements for the computing power of the vehicle have become increasingly stringent. Limited by the bottleneck that is difficult to break through in hardware computing power, coupled with the consideration of vehicle cost control, the computing power of the vehicle is difficult to support the high-performance operation of all detection algorithms. In desperation, the perception-type neural network model algorithm on the vehicle at this stage can only reduce the accuracy in order to ensure that all detection algorithms can operate normally under limited computing power conditions. However, this measure inevitably causes the detection accuracy to decline, resulting in the omission of some pedestrians and vehicles. On the other hand, during the driving of the vehicle, in certain specific scenes, such as turning, on-ramp and off-ramp, etc., the driver's attention is often focused on one side, which makes it easy to ignore the traffic conditions on the other side. Sometimes, due to interference from factors such as large traffic volume or complex scenes, even if obstacles around the vehicle have been detected, it is difficult for the driver to accurately determine their location.

[0004] In response to insufficient computing power, the current mainstream technical solution is to reduce the accuracy of the vehicle algorithm to ensure that all detection algorithms can run smoothly under limited computing power. Or based on past experience, the software algorithm accuracy of different cameras is manually adjusted. For example, the front camera uses float16 accuracy to ensure accurate capture of the road conditions ahead, while the rear camera uses int8 accuracy to meet basic monitoring needs. Through this differentiated configuration, the vehicle software algorithm can maintain normal operation under limited computing power conditions. However, there are obvious shortcomings in the current warning link of the surrounding environment of the whole vehicle. At this stage, the whole vehicle lacks precise part distinction for warnings of the surrounding environment. Whether the front, perimeter, or rear cameras capture anomalies or the radars detect risks, a unified alarm sound or voice broadcast is used to issue warnings. This model exposes many drawbacks, the most prominent of which are the lack of real-time performance, poor flexibility, and inaccurate information received by the driver.

[0005] Focusing on the vehicle computing power level, current technical solutions are difficult to achieve real-time dynamic adjustment of computing power accuracy. All accuracy settings are pre-planned and fixedly executed, and it is impossible to flexibly change the accuracy parameters according to the ever-changing driving environment and the driver's real-time driving behavior, which leads to rigid perception strategies and difficulty in adapting to complex and changing road conditions.

[0006] Looking at the driver warning, the current alarm mechanism is too general. As long as any sensor of the vehicle detects an obstacle from any direction, it will immediately trigger a unified alarm for the entire vehicle. This undoubtedly brings great trouble to the driver. Take the vehicle turning scene as an example. When the driver's sight is focused on the pedestrians on the turning side and has noticed them, the whole vehicle suddenly issues an alarm. The driver is very likely to misjudge that the warning target is only the pedestrian he has noticed, but he is completely unaware that there is an obstacle approaching the vehicle at high speed on the other side. In this way, the driver is very likely to miss the key danger points in the warning information, which will bury hidden dangers for driving safety.

[0007] To sum up, the existing autonomous driving and assisted driving technologies urgently need to be optimized and innovated in key dimensions such as computing power allocation, environmental warning, and driver information interaction in order to effectively improve driving safety and intelligent driving experience. Summary of the invention

[0008] An object of the present invention is to provide a DMS visual blind spot compensation algorithm model optimization method, system and vehicle to solve the aforementioned problem that when the driver performs driving behaviors such as turning, the driver's attention is focused on one side of the vehicle body and cannot take into account the blind spot on the other side, and the problem that the accuracy of the software detection algorithm cannot be adjusted in real time and dynamically.

[0009] According to a first aspect of the present invention, a method for optimizing a visual blindness compensation algorithm model based on DMS is provided, comprising:

[0010] Step S1: monitoring the driver's driving intention through the DMS;

[0011] Step S2: determining whether the driving intention is to turn in the first direction, if yes, proceeding to step S3, otherwise proceeding to step S4;

[0012] Step S3: Transmitting the driver's first direction steering information to the central domain control, and updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot;

[0013] Step S4: the vehicle cameras call the first data format model weight of the surround perception algorithm for detection;

[0014] Step S5: The state of the camera in the first direction of the vehicle remains unchanged, and the state of the camera in the second direction of the vehicle becomes a blind spot;

[0015] Step S6: The first direction camera of the whole vehicle continues to call the first data format model weight of the surround view perception algorithm for detection, and the second direction camera of the whole vehicle continues to call the second data format model weight of the surround view perception algorithm for detection.

[0016] Optionally, while the driver's driving intention is monitored by the DMS, the surround perception algorithm of the entire vehicle is initialized to the first data format model weight parameter type.

[0017] Optionally, before step S1, the method further includes:

[0018] Get the vehicle running status;

[0019] If the vehicle is in a cold start state or the vehicle is static, the weight parameters of the surround view perception algorithm are automatically increased from the first data format model weight parameters to the second data format model weight parameters.

[0020] Optionally, in step S6, after the first direction camera of the whole vehicle continues to call the first data format model weight of the surround perception algorithm for detection, it includes:

[0021] The DMS checks again whether the driver's steering intention has returned to normal. If not, the second direction camera continues to use the second data format model weight for detection, and the first direction camera continues to use the first data format model weight for detection;

[0022] If so, the entire vehicle camera resumes using the first data format model weights for detection.

[0023] Optionally, in step S6, the first direction camera of the whole vehicle includes a first direction front view camera, a first direction rear view camera, and a first direction surround view camera; the second direction camera of the whole vehicle includes a second direction front view camera, a second direction rear view camera, and a second direction surround view camera.

[0024] Optionally, in step S1, the driver's driving intention is monitored by the DMS, including monitoring the driver's head movements, monitoring the steering wheel holding posture, and performing eye tracking.

[0025] Optionally, the DMS-based visual blindness compensation algorithm model optimization method further includes:

[0026] Based on the obstacle information detected in the blind spot, DMS warns the driver by highlighting the ambient light or sounding an alarm through the speakers on one side.

[0027] According to a second aspect of the present invention, a DMS-based visual blindness compensation algorithm model optimization system is provided, including a DMS, wherein the DMS includes:

[0028] A driving intention monitoring module is used to monitor the driver's driving intention;

[0029] A state updating module, used for transmitting the driver's first direction steering information to the central domain control, and updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot;

[0030] The partition control module is used to enable the first direction camera of the whole vehicle to continue to call the first data format model weight of the surround perception algorithm for detection, and the second direction camera of the whole vehicle to call the second data format model weight of the surround perception algorithm for detection.

[0031] Optionally, the DMS-based visual blindness compensation algorithm model optimization system further includes:

[0032] The warning module is connected to the DMS and is used to warn the driver based on the obstacle information detected by the DMS by highlighting the ambient light or emitting an alarm tone through a single-sided speaker.

[0033] According to a third aspect of the present invention, a vehicle is provided, comprising a DMS-based visual blindness compensation algorithm model optimization system.

[0034] The DMS-based visual blindness compensation algorithm model optimization method, system, and vehicle disclosed in the present invention have the following technical effects:

[0035] Firstly, the present invention uses the detection algorithm of DMS to judge the driver's driving steering intention, thereby judging the direction of the driver's blind spot field of view, and feeding back the driver's blind spot position to the whole vehicle, so that the first direction camera of the whole vehicle continues to call the first data format model weight of the surround perception algorithm for detection, and the second direction camera of the whole vehicle calls the second data format model weight of the surround perception algorithm for detection. The whole vehicle improves the parameter accuracy of the software algorithm of the sensor in the blind spot part to ensure that in the driver's blind spot, the sensor can better help the driver complete the environmental perception work and can see farther and more accurately;

[0036] Secondly, the present invention ensures that under limited computing power conditions, the accuracy of each sensor can be dynamically adjusted, and the accuracy of the software algorithms of sensors in different parts can be improved by about 10%-20% according to the real-time driving status. The computing power is saved at 10TOPS, avoiding the waste of computing power on unnecessary tasks and improving the flexibility of the entire vehicle algorithm.

[0037] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0039] Figure 1 A flowchart of a DMS-based visual blindness compensation algorithm model optimization method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0041] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0042] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0043] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0044] The present invention proposes an embodiment of a DMS-based visual blindness compensation algorithm model optimization method. It should be noted that in the early stage of technology research and development, a refined training process is implemented for all detection models equipped on the vehicle. A double-precision parameter training strategy is adopted, that is, for the same detection model, special training is carried out using the int8 low-precision parameter type and the float32 high-precision parameter type. After the training task is successfully completed, the generated weight file adapted to different precision parameters is properly stored in a specific storage area of ​​the vehicle configuration system for subsequent operation calls. It should be emphasized that the two parameter types, int8 and float32, show significant differences in the storage performance requirements of the vehicle domain control. Compared with int8, the float32 data type occupies a considerable amount of storage space due to its inherent characteristics. In the limited storage resource system of the vehicle domain control, it is often necessary to open up a larger storage space to accommodate related data.

[0045] Moreover, when entering the actual calculation phase, the float32 parameter type has a higher level of precision and retains more significant digits after the decimal point, which inevitably requires more computing resources in the calculation process, covering the central processing unit (CPU) calculation time, graphics processing unit (GPU) computing power allocation and other aspects, and the calculation time is greatly increased. However, it cannot be ignored that although the float32 parameter type has disadvantages such as high storage and computing costs, it can bring significant gains to the vehicle's environmental perception performance. The detection model trained using the float32 parameter type can achieve a significant improvement of 10% - 20% in detection result accuracy compared to the int8 type. In complex traffic scenarios, it is more sensitive and accurate in identifying various potential obstacles, and can accurately capture the contour details of small non-motor vehicles and special-shaped obstacles, greatly enhancing the reliability and accuracy of the vehicle's perception of the surrounding environment.

[0046] Specifically, if Figure 1 As shown, the DMS-based visual blindness compensation algorithm model optimization method includes:

[0047] Step S1: monitoring the driver's driving intention through the DMS;

[0048] Specifically, DMS monitors the driver's driving intention, including monitoring the driver's head movements, monitoring the steering wheel holding posture, and tracking the eye gaze.

[0049] Step S2: Determine whether the driving intention is to turn in the first direction. If so, proceed to step S3, otherwise proceed to step S4; specifically, the first direction turning can be set to turn left, and the second direction turning can be set to turn right.

[0050] Step S3: Transmitting the driver's first direction steering information to the central domain control, updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot;

[0051] Specifically, if it is detected that the driver is turning his head to the left, it means that the driver is observing the road condition on the left side, and the right side of the vehicle is the driver's blind spot.

[0052] Step S4: The whole vehicle camera calls the first data format model weight of the surround perception algorithm for detection; specifically, the whole vehicle camera includes a whole vehicle first direction camera and a whole vehicle second direction camera.

[0053] Step S5: the state of the camera in the first direction of the vehicle remains unchanged, and the state of the camera in the second direction of the vehicle becomes a blind spot; that is, if it is detected that the driver turns his head to the left, the state of the camera on the left side remains unchanged, and the right side becomes a blind spot.

[0054] Step S6: The first direction camera of the whole vehicle continues to call the first data format model weight of the surround view perception algorithm for detection, and the second direction camera of the whole vehicle continues to call the second data format model weight of the surround view perception algorithm for detection.

[0055] Specifically, the first data format model weight is the int8 data format weight, and the second data format model weight is the Float32 high-precision data format weight; when the driver turns his head to the left, the sensors in the right blind spot (right front camera, right rear camera, right surround camera, right laser blind spot radar) will use the Float32 high-precision data format weight to perform algorithm reasoning to assist the driver in completing the environmental perception work in the blind spot. The sensors on the left side continue to use the int8 data format weight to perform algorithm reasoning to avoid wasting computing power on unnecessary tasks.

[0056] In an embodiment of the present invention, while the driver's driving intention is monitored by the DMS, the surround perception algorithm of the whole vehicle is initialized to the first data format model weight parameter type. Specifically, the perception algorithm of the whole vehicle is initialized to the int8 parameter type, reserving the computing power and space required for subsequent high precision.

[0057] In the embodiment of the present invention, before step S1, the method further includes:

[0058] Get the vehicle running status;

[0059] If the vehicle is in a cold start state or is static, the weight parameters of the surround view perception algorithm are automatically upgraded from the first data format model weight parameters to the second data format model weight parameters. For example, the weight parameters of the surround view perception algorithm can be upgraded from int8 to float32 to enhance the perception of the driver's blind spots under and around the vehicle body in a static state.

[0060] It should be noted that before the static start of the vehicle, the driver needs to walk around the vehicle to observe whether there are obstacles, personnel, animals, etc. around the vehicle body, but in most cases, the driver will omit this step. Relying on the vehicle's own surround perception to perform static perception around the vehicle, but because the surround camera generally has low pixels, the detection effect is not very good, thus ignoring some small obstacles such as cats and dogs. At this time, the present invention can allocate higher parameter accuracy to the surround detection algorithm in a static situation, improve the detection accuracy of the algorithm, and help the driver to better perform static blind spot detection. On the other hand, in the static situation where the car is stopped, due to obstacles such as vehicles and pillars around the vehicle body, the alarm system will always sound for early warning. At this time, if there are other obstacles around the vehicle body, such as a cat, the driver will find it difficult to distinguish whether the vehicle is alerting the pillars around the vehicle body or the cat. The driver mistakenly thinks that it is an early warning of the pillars around the vehicle body, resulting in some collision risks with the cat.

[0061] The present invention assists the driver in determining the specific direction from which the alarm is issued through an enhanced warning system in the vehicle. Taking the atmosphere light as an example, when the atmosphere light signal is connected, it can be divided into left front, left rear, right front, and right rear according to the atmosphere lights on the four doors. When the driver turns his head to the right, he finds a pillar on the right. At this time, the DMS detects that the driver's attention is on the right side of the vehicle, so only a voice alarm is needed. However, a cat may appear in the blind spot on the left side of the driver. At this time, the driver cannot distinguish which obstacle is being warned by the alarm sound. At this time, the atmosphere light on the left side can flash red to remind the driver that there is also an obstacle in the blind spot on the left. Since there are many reflective planes in the car, such as window glass and rearview mirrors inside the car, the driver can use peripheral vision to observe the warning. At this time, the driver will know that there are also obstacles in the blind spot on the other side, thereby avoiding danger.

[0062] In the embodiment of the present invention, in step S6, after the first direction camera of the whole vehicle continues to call the first data format model weight of the surround perception algorithm for detection, it includes:

[0063] The DMS checks again whether the driver's steering intention has returned to normal. If not, the second direction camera continues to use the second data format model weight for detection, and the first direction camera continues to use the first data format model weight for detection;

[0064] If so, the entire vehicle camera resumes using the first data format model weights for detection.

[0065] In an embodiment of the present invention, in step S6, the first direction camera of the whole vehicle includes a first direction front view camera, a first direction rear view camera, and a first direction surround view camera; the second direction camera of the whole vehicle includes a second direction front view camera, a second direction rear view camera, and a second direction surround view camera.

[0066] In an embodiment of the present invention, the DMS-based visual blindness compensation algorithm model optimization method further includes:

[0067] Based on the obstacle information detected in the blind spot, DMS warns the driver by highlighting the ambient light or sounding an alarm on one side of the speaker. When the blind spot is critical and visual and auditory warnings alone are not enough to avoid the risk of collision, the system decisively connects to the vehicle control (chassis control, power control and other multi-system integrated control) signal link to directly intervene in the vehicle's driving status.

[0068] In another embodiment of the present invention, taking the scenario where the driver makes a right turn as an example, the driver often cares about whether there are motor vehicles, non-motor vehicles and pedestrians on the right side, and sometimes continues to pay attention to whether there are non-motor vehicles approaching and interweaving quickly on the right rear side, thus completely ignoring the situation on the left side. This situation is particularly obvious on large vehicles such as commercial vehicles. At this time, there is a high probability that the problem of whether there are other vehicles on the left side of the vehicle will be ignored, resulting in dangers such as collisions. When there is a normal vehicle on the left side, the driver ignores the danger of the vehicle coming from the left side because his attention is on the right side, and may invade the lane of the vehicle on the left side after turning right, causing dangers such as collisions. The embodiment of the present invention can help the driver pay attention to the environmental conditions on the right side while improving the parameter accuracy of the left sensor to improve the perception performance within the left blind spot, which can help the driver to perceive the blind spot farther and more accurately, and help the driver to decelerate, brake and other operations in an emergency.

[0069] The present invention also provides an embodiment of a visual blindness compensation algorithm model optimization system based on DMS, including DMS, wherein the DMS includes:

[0070] A driving intention monitoring module is used to monitor the driver's driving intention;

[0071] A state update module, used for transmitting the driver's first direction steering information to the central domain control, updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot;

[0072] The partition control module is used to enable the first direction camera of the whole vehicle to continue to call the first data format model weight of the surround perception algorithm for detection, and the second direction camera of the whole vehicle to call the second data format model weight of the surround perception algorithm for detection.

[0073] In an embodiment of the present invention, the DMS-based visual blindness compensation algorithm model optimization system also includes a warning module, which is connected to the DMS and is used to warn the driver based on the obstacle information detected by the DMS by highlighting the ambient light or issuing an alarm tone through a single-sided speaker.

[0074] It should be noted that the information interaction, execution process and other contents between the above-mentioned modules are based on the same concept as the method embodiment of the present application, and are a system corresponding to the above-mentioned DMS-based visual blindness compensation algorithm model optimization method. All implementation methods in the above-mentioned method embodiment are applicable to the embodiment of the system. Its specific functions and technical effects brought about by it can be found in the method embodiment part, which will not be repeated here.

[0075] In the DMS-based visual blindness compensation algorithm model optimization solution of the embodiment of the present invention, the perception effect and perception accuracy on one side of the driver's blind spot are provided, which can enable the driver to perform perception tasks around the vehicle body better and faster when the driver has a blind spot.

[0076] It provides a flexible algorithm accuracy improvement system, which can switch the perception algorithm model accuracy of different sensors according to the real-time vehicle driving status, so as to bring out the sensor performance required by users on limited computing power. It improves the early warning system, and by connecting to the interior ambient light and the interior display screen, when different sensors detect obstacles at the same time, obstacles in the blind spot will not be missed due to the same early warning sound, which can help the driver to judge the environment around the vehicle more accurately.

[0077] The present invention also provides an embodiment of a vehicle, including the above-mentioned DMS-based visual blindness compensation algorithm model optimization system. Through the detection of the driver's state in the vehicle by DMS, the driver's visual blind spot at this time is determined, and the algorithm accuracy is improved and converted through the driver's blind spot. In addition, the traditional single sound warning is improved to a multi-sensory warning, and obstacles around the vehicle body are distinguished and warned through methods such as ambient lights and screen displays.

[0078] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A DMS-based visual blindness compensation algorithm model optimization method, characterized in that: include: Step S1: monitoring the driver's driving intention through the DMS; Step S2: determining whether the driving intention is to turn in the first direction, if yes, proceeding to step S3, otherwise proceeding to step S4; Step S3: Transmitting the driver's first direction steering information to the central domain control, and updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot; Step S4: the vehicle cameras call the first data format model weight of the surround perception algorithm for detection; Step S5: The state of the camera in the first direction of the vehicle remains unchanged, and the state of the camera in the second direction of the vehicle becomes a blind spot; Step S6: The first direction camera of the whole vehicle continues to call the first data format model weight of the surround view perception algorithm for detection, and the second direction camera of the whole vehicle continues to call the second data format model weight of the surround view perception algorithm for detection.

2. The DMS-based visual blindness compensation algorithm model optimization method according to claim 1, characterized in that: While the driver's driving intention is monitored through the DMS, the surround perception algorithm of the entire vehicle is initialized to the first data format model weight parameter type.

3. The DMS-based visual blindness compensation algorithm model optimization method according to claim 1, characterized in that: Before step S1, the method further includes: Get the vehicle running status; If the vehicle is in a cold start state or the vehicle is static, the weight parameters of the surround view perception algorithm are automatically increased from the first data format model weight parameters to the second data format model weight parameters.

4. The DMS-based visual blindness compensation algorithm model optimization method according to claim 1, characterized in that: In step S6, after the first direction camera of the vehicle continues to call the first data format model weight of the surround perception algorithm for detection, it includes: The DMS checks again whether the driver's steering intention has returned to normal. If not, the second direction camera continues to use the second data format model weight for detection, and the first direction camera continues to use the first data format model weight for detection; If so, the entire vehicle camera resumes using the first data format model weights for detection.

5. The DMS-based visual blindness compensation algorithm model optimization method according to claim 1, characterized in that: In step S6, the first direction cameras of the whole vehicle include a first direction front view camera, a first direction rear view camera, and a first direction surround view camera; the second direction cameras of the whole vehicle include a second direction front view camera, a second direction rear view camera, and a second direction surround view camera.

6. The DMS-based visual blindness compensation algorithm model optimization method according to claim 1, characterized in that: In step S1, the driver's driving intention is monitored by the DMS, including monitoring the driver's head movements, monitoring the steering wheel holding posture, and tracking the eye gaze.

7. The DMS-based visual blindness compensation algorithm model optimization method according to any one of claims 1 to 6, characterized in that: Also includes: Based on the obstacle information detected in the blind spot, DMS warns the driver by highlighting the ambient light or sounding an alarm through the speakers on one side.

8. A DMS-based visual blindness compensation algorithm model optimization system, characterized in that: comprising a DMS, the DMS comprising: A driving intention monitoring module is used to monitor the driver's driving intention; A state updating module, used for transmitting the driver's first direction steering information to the central domain control, and updating the driver's state to the driver's first direction steering through the central domain control, and updating the driver's second direction to the blind spot; The partition control module is used to enable the first direction camera of the whole vehicle to continue to call the first data format model weight of the surround perception algorithm for detection, and the second direction camera of the whole vehicle to call the second data format model weight of the surround perception algorithm for detection.

9. The DMS-based visual blindness compensation algorithm model optimization system according to claim 8, characterized in that: Also includes: The warning module is connected to the DMS and is used to warn the driver based on the obstacle information detected by the DMS by highlighting the ambient light or emitting an alarm tone through a single-sided speaker.

10. A vehicle, characterized in that: Including the DMS-based visual blindness compensation algorithm model optimization system described in claim 9.

Images