An environmental perception-based electronic outside rearview mirror image optimization display method and electronic rearview mirror

Abstract

The application discloses an environment perception-based electronic outside rearview mirror image optimization display method and an electronic rearview mirror, and belongs to the technical field of electronic rearview mirrors. The method comprises the following steps: collecting the environment information of a vehicle and the state information of the vehicle in real time, marking the image frame of the electronic rearview mirror and the collected data with the same time reference, and constructing a frame-synchronized environment vector; performing multi-modal evidence fusion based on the frame-synchronized environment vector, forming a probabilistic judgment on the environment state, and obtaining a final state decision; calculating a final optimization parameter set according to the result of the state decision by using a parameter smooth transition model; and delivering the final optimization parameter set to a DSP chip, so that the DSP chip generates an optimized electronic rearview mirror image for final presentation. The application effectively improves the accuracy and robustness of the electronic outside rearview mirror image optimization.

Description

Technical Field

[0001] This invention relates to the field of electronic rearview mirror technology, and in particular to an image optimization display method for electronic exterior rearview mirrors based on environmental perception, and an electronic rearview mirror. Background Technology

[0002] The rapid development of electronic rearview mirror technology has greatly contributed to vehicle intelligence. Generally, traditional vehicle perception systems rely on single sensors or simple data fusion techniques. However, these traditional technologies often face numerous challenges when dealing with complex and ever-changing environments. For example, traditional vehicle environmental perception systems often lack environmental adaptability, making it difficult to quickly switch and accurately adapt to different scenarios, especially in environments such as highways, urban roads, and tree-lined roads where road conditions change rapidly. Sensors need to dynamically adjust according to the current road conditions to maintain efficient operation. Secondly, existing multi-sensor fusion systems generally have shortcomings, failing to fully consider the combined impact of complex and changing environmental factors and the vehicle's real-time driving status on sensor performance, making it difficult to achieve optimal sensor weight allocation in various scenarios. This, in turn, leads to distortion and deformation of the displayed image in the electronic rearview mirror. Summary of the Invention

[0003] One of the objectives of this invention is to provide an image optimization display method for electronic rearview mirrors based on environmental perception, so as to solve the problem that electronic rearview mirrors in the prior art lack environmental adaptive capabilities and are difficult to achieve rapid switching and accurate adaptation of displayed images in different scenarios.

[0004] This invention is achieved through the following technical solution: a method for optimizing the display of electronic rearview mirror images based on environmental perception, comprising the following steps: S100, real-time acquisition of vehicle environmental information and vehicle state information, and assigning the same time reference to the image frames of the electronic rearview mirror and the acquired data, thereby constructing a frame-synchronized environmental vector; S200, performing multimodal evidence fusion based on the frame-synchronized environmental vector to form a probabilistic judgment of the environmental state, obtaining the final state decision, and evaluating the confidence level of the probabilistic judgment to address the challenges of sensor noise or malfunction; S300, making... The final optimized parameter set is calculated based on the state decision results using a parameter smoothing transition model. The parameter smoothing transition model includes: finding and obtaining the basic optimized parameter set from the environment-parameter mapping table based on the state decision results; then fine-tuning the basic optimized parameter set in combination with the illumination value; updating the basic optimized parameter set using a first-order inertial filter; and dynamically adjusting the filter coefficients of the filter in combination with the confidence level; S400: sending the final optimized parameter set to the DSP chip; the DSP chip performs image processing operations based on the final optimized parameter set to generate the final optimized electronic rearview mirror image.

[0005] Furthermore, the frame synchronization environment vector is constructed as follows: for continuously changing physical quantities, linear interpolation is used for alignment; for discrete state quantities, a forward padding strategy is adopted to encapsulate all data that is time-aligned with the completion of a single frame image into a standardized frame synchronization environment vector data packet.

[0006] Furthermore, when the continuously changing physical quantity is the illumination value, linear interpolation alignment is achieved using the following formula:

[0007] ,in, Let be the illumination values ​​that are perfectly aligned temporally for the nth frame image. The most recent illumination sensor reading prior to the image frame timestamp. This represents the first illumination sensor reading after the image frame timestamp; For timestamps, For the nth frame image, Represents the precise timestamp of the nth frame of the image; For reading timestamp, Indicates reading Timestamp.

[0008] Furthermore, when the discrete state variable is the weather state, the forward filling strategy is implemented using the following formula:

[0009] ,in, The time-aligned weather conditions determined for the nth frame image; Weather status value; Indicates constraints. Weather status value timestamp, This represents the maximum value.

[0010] Furthermore, the frame synchronization environment vector data packet can be represented by the following equation:

[0011] ,

[0012] in, The unified data input for the nth frame is the frame synchronization environment vector packet, which integrates the image itself and information such as the illumination value, weather conditions, and vehicle state vectors that are precisely synchronized with it. This refers to the vehicle's status.

[0013] Furthermore, multimodal evidence fusion is calculated through an environmental state probability model. The evidence score for each potential environmental state is obtained by weighted summation of the membership degrees from different sensors after transformation. The scenario with the highest score is the current state, and the difference between the highest score and the second highest score is used to quantify the confidence level of this judgment. The membership degree is obtained by smoothly mapping the environmental data from the frame synchronization environment vector to the interval [0, 1] using the Sigmoid function.

[0014] Furthermore, for a scene with basic lighting, the membership degree is calculated using the following formula:

[0015] ,

[0016] ,in, This represents the degree of membership of the environment to a daytime scene under given illumination values. This represents the membership degree of the environment to a nighttime scene under given illumination values; The symbol for the natural exponential function is . These are the curve control parameters for daytime. For curve control parameters at night, The preset daytime light threshold, The preset light threshold for nighttime; The calculated illumination values ​​are those that are perfectly aligned in time for the nth frame image.

[0017] Furthermore, the formula for calculating the evidence score when the potential environmental state is a tunnel entrance scenario is as follows:

[0018] ,

[0019] ,

[0020] in, Indicates in a given data packet Under the given conditions, the overall score for nighttime scenes; Indicates in a given data packet Under the given conditions, the overall score for entering the tunnel scenario; For the weighting coefficients of the light sensor, Nighttime membership is based on the illumination sensor; The weighting coefficient for image brightness. Nighttime membership is based on image brightness; To calculate the illumination sensor weighting coefficients for the overall score of the tunnel scene, To calculate the image brightness weighting coefficients for the overall score of the tunnel scene, The weighting coefficient for the rate of change of illumination. The membership degree of the tunnel scene is based on the rate of change of illumination;

[0021] Furthermore, the final state decision is calculated using the following formula: The confidence level is calculated using the following formula: ,in, The most likely environmental state for the final decision in frame n. It is an element in the set of all environment states. For current decision Confidence level, To adjust the coefficients of the confidence curve, To achieve the highest score, This is the second highest score.

[0022] Furthermore, the basic optimization parameter set and brightness fine-tuning parameters are calculated using the following formula: , ,in, For the basic target parameter set of the nth frame, For the preset environment-parameter mapping table, This is the adjustment amount for the brightness parameter calculated based on the current illuminance value. This is the adjustment coefficient for the brightness parameter adjustment amount; It is the natural logarithm function; For the current scenario The next reference illuminance; To ensure that the logarithmic function is always a meaningful positive number.

[0023] Furthermore, the first-order inertial filter is shown in the following equation:

[0024] ,

[0025] ,

[0026] ,in, This is the set of optimization parameters to be applied in the final frame n. The parameter set that was finally applied in the previous frame. These are the smoothing filter coefficients dynamically calculated for the nth frame; Let n be the final target parameter set for the nth frame. For an adjustment vector function, This represents the lower bound of the filter coefficients. This is the upper limit of the filter coefficients. This is a power-law exponent of the confidence level, used to enhance sensitivity to high confidence intervals.

[0027] Another aspect of the present invention provides an electronic rearview mirror, an image sensor, a processor, a display device, and a memory. The processor, memory, image sensor, and display device are all electrically connected. A computer program is stored in the memory. When the computer program is executed by the processor, it implements any of the above-mentioned methods for optimizing the display of electronic rearview mirror images based on environmental perception.

[0028] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0029] 1. This invention effectively solves the problem of sensor data time misalignment by unifying image frames with sensor data such as illumination, weather, and vehicle status to the same time base, thereby improving the reliability of environment recognition and parameter optimization. This enables the model to accurately determine the environment based on the real state at the moment of exposure when processing any image frame. At the same time, it overcomes the fragility of traditional IF-THEN hard threshold rule-based judgment methods through multimodal evidence fusion, effectively suppressing the impact of noise or failure of a single sensor on the judgment result, and improving the accuracy and robustness of environment judgment.

[0030] 2. This invention achieves a damping effect through confidence level, making parameter adjustments smoother and enabling smooth transitions of parameters when the environment changes. It can also calculate the basic target parameter set and brightness fine-tuning parameters based on the environmental state and continuous environmental data, and update the current parameters through a first-order inertial filter, which ensures fast response and avoids unstable jumps, effectively improving the stability and reliability of the system.

[0031] 3. This invention can flexibly switch and accurately adapt to different environments. Through the effective fusion and time alignment of multi-source asynchronous sensor data, it realizes image adjustment based on comprehensive and efficient perception of the vehicle's surrounding environment, providing users with more accurate image display in complex environments. Attached Figure Description

[0032] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0033] Figure 1 This is a flowchart of the method provided in Embodiment 1 of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0035] Example 1

[0036] Current electronic rearview mirrors typically use multiple single-sensor information based on IF-THEN hard threshold rules to determine the environment and adjust the image. In the context of multi-source information fusion perception in vehicles, differences in data processing and transmission times can lead to situations where the timestamp of a later received message precedes that of an earlier message. Especially during vehicle movement, the temporal relationships of sensor data become even more complex, making the effective fusion and processing of this heterogeneous data a pressing issue.

[0037] To address these issues, this embodiment discloses an image optimization display method for electronic exterior rearview mirrors based on environmental perception. The content disclosed in this embodiment can not only improve the accuracy of environmental perception, but also achieve image optimization display based on environmental perception. Figure 1 A flowchart of the method in this embodiment is shown, from Figure 1 As can be seen from this embodiment, it includes the following steps:

[0038] Step 1: Based on the sensors deployed on the vehicle body, real-time environmental information such as light and weather around the vehicle, as well as the vehicle's own status information, these multi-source asynchronous sensor data are unified with the image frames collected by the electronic rearview mirror to the same time base, thereby constructing a standardized frame synchronization environment vector.

[0039] Generally, in actual data acquisition, image frames and sensors for illumination, weather, and vehicle status often sample at different frequencies, resulting in clock drift or communication delays. Directly using raw sensor readings can easily lead to time misalignment, weakening the reliability of subsequent environmental recognition and parameter optimization. Therefore, in this embodiment, any given image frame is first aligned temporally to the environmental data at the corresponding moment, ensuring that the environmental data processed by the model strictly corresponds to the current image. This ensures that when the subsequent algorithm processes any given image frame, the environmental information it relies on is the true state at the moment of image exposure, rather than outdated data from before or after.

[0040] Specifically, in this embodiment, different alignment strategies can be adopted based on the physical characteristics of the data type. For physical quantities like illumination, which are inherently continuously changing, their state between two sampling points can be reasonably estimated through interpolation. However, for discrete state quantities like weather and vehicle status, whose changes are step-like, interpolation has no physical meaning, and their state should remain unchanged between two updates. In this case, the nearest past value principle should be used for forward padding. Finally, all data that is time-aligned with a single frame image is encapsulated into a standardized data package, providing a unified and clean input for subsequent models.

[0041] For example, in this embodiment, linear interpolation of continuous data can be expressed by the following formula:

[0042] ,

[0043] in, To calculate the illumination values ​​that are perfectly aligned in time for the nth frame image, The most recent illumination sensor reading prior to the image frame timestamp. This represents the first illumination sensor reading after the image frame timestamp; For timestamps, For the nth frame image, Represents the precise timestamp of the nth frame of the image; For reading timestamp, Indicates reading Timestamp.

[0044] For example, in this embodiment, the alignment of discrete data adopts a forward padding strategy, as shown in the following formula:

[0045] ,

[0046] in, The time-aligned weather conditions determined for the nth frame image; Weather status value; This indicates a constraint condition. Weather status value timestamp, This represents the maximum value.

[0047] For example, in this embodiment, the standardized data packet can be represented by the following formula:

[0048] ,

[0049] in, The unified data input for the nth frame is the frame synchronization environment vector packet, which integrates the image itself and information such as the illumination value, weather conditions, and vehicle state vectors that are precisely synchronized with it. This represents the vehicle's status. By encapsulating the data in this way, multi-source heterogeneous information can be managed within a unified framework, ensuring its semantics are consistent with the time reference, which facilitates subsequent multimodal fusion.

[0050] Step 2: Based on the constructed data packet, a probabilistic judgment on the environmental state is formed through multimodal evidence fusion, and the credibility of the judgment is evaluated to address the challenges of sensor noise or failure.

[0051] In this embodiment, after obtaining data packets that are precisely synchronized with each frame of image, a multimodal evidence fusion environment state probability model is constructed to achieve a comprehensive judgment of the environment state and quantify its confidence level.

[0052] Generally, traditional methods rely on a rule-based approach using an IF-THEN hard threshold. This method is inherently fragile and prone to misjudgments due to noise or malfunction of a single sensor, making it particularly susceptible to errors in automotive environments. For instance, when a car enters a shaded area, the light sensor may be briefly obscured by trees, potentially causing the IF-THEN hard threshold to misjudge the vehicle as entering a tunnel. Therefore, a more robust judgment criterion should comprehensively utilize all available information sources (such as light sensor readings, image content, and weather data) as the basis for judgment, employing cross-validation to jointly determine the most probable environmental state.

[0053] Specifically, the environmental state probability model in this embodiment calculates an evidence score for each potential environmental state (e.g., daytime, nighttime, tunnel entrance, rainy day, etc.). This evidence score is obtained by weighted summation of transformed membership degrees from different sensors. The membership degree is obtained by smoothly mapping transformed environmental data from different sensors to the interval [0, 1] using a nonlinear function (e.g., the Sigmoid function), representing the degree of confidence that the current reading supports a judgment of a certain scenario.

[0054] Ultimately, the scenario with the highest score is selected as the current state, and the difference between the highest and second-highest scores can be used to quantify the level of confidence in this judgment (i.e., confidence level).

[0055] For example, in this embodiment, the membership degree of the basic lighting scene (day / night) can be calculated by the following formula:

[0056] ,

[0057] ,

[0058] in, This represents the degree of membership of the environment to a daytime scene under given illumination values. This represents the membership degree of the environment to a nighttime scene under given illumination values; The symbol for the natural exponential function is . These are the curve control parameters for daytime. These are the curve control parameters for nighttime. These two parameters are used to control the steepness of the function curve and are positive constants. The preset daytime light threshold, The preset light threshold for nighttime.

[0059] For example, in this embodiment, to describe the rapidly darkening environment at the tunnel entrance, the rate of change of illumination can be obtained using the following formula, and its membership degree can be calculated accordingly:

[0060] ,

[0061] ,

[0062] in, This indicates the membership degree of entering the tunnel based on the rate of change of illumination. These are the curve control parameters for the membership degree of the tunnel. The rate of change of illumination, A positive threshold represents the magnitude of the rate of change that constitutes a dramatic change; For the m-th frame, The illumination values ​​are for the previous m frames. This refers to the frame images of the first m frames.

[0063] For example, taking nighttime and tunnel entrance as examples, the calculation process of the environmental state score in this embodiment can be shown in the following formula:

[0064] ,

[0065] ,

[0066] in, Indicates in a given data packet Under the given conditions, the overall score for nighttime scenes; Indicates in a given data packet Under the given conditions, the overall score for entering the tunnel scenario; For the weighting coefficients of the light sensor, Nighttime membership is based on the illumination sensor; The weighting coefficient for image brightness. Nighttime membership is based on image brightness; To calculate the illumination sensor weighting coefficients for the overall score of the tunnel scene, To calculate the image brightness weighting coefficients for the overall score of the tunnel scene, The weighting coefficient for the rate of change of illumination. The membership degree of the tunnel scene is based on the rate of change of illumination.

[0067] For example, in this embodiment, the environmental state score for rainy days can be expressed as follows:

[0068] ,

[0069] in, The overall score for rainy day scenarios. Weighting of weather data, Membership degree for rainy days calculated based on weather sensor readings; For image analysis weights, This represents the intensity of rain streak features based on image analysis.

[0070] For example, in this embodiment, the environmental state score of the glare scene can be expressed as follows:

[0071] ,

[0072] in, Scoring is given for the environmental condition of the scene where the high beams of the following vehicle illuminate the vehicle (glare).

[0073] This refers to the proportion of pixels whose brightness values ​​exceed the saturation threshold within a specific region of interest (ROI) of the image. The maximum saturation value, It is an exponential adjustment factor.

[0074] For example, in this embodiment, the formula for calculating the final state decision and confidence level can be shown as follows:

[0075] ,

[0076] ,

[0077] in, The most likely environmental state for the final decision in frame n. It is an element in the set of all possible environmental states, such as night, tunnel entrance, rain, etc. Current environmental state Confidence level, To adjust the coefficients of the confidence curve, To achieve the highest score, This is the second highest score.

[0078] Step 3: After obtaining the identified current environmental state and specific confidence level, the basic optimization parameter set is retrieved from the preset environment-parameter mapping table based on the environmental state. Then, the basic optimization parameter set is fine-tuned based on the illumination value.

[0079] Generally, directly converting high-level cognitive signals (e.g., entering a tunnel) into low-level image processing parameters (e.g., brightness, contrast) for switching would cause abrupt changes in the electronic rearview mirror screen, resulting in a poor user experience. Therefore, a smooth parameter transition model should ensure a smooth and gradual parameter adjustment process, allowing for natural adaptation by the human eye. Simultaneously, the speed of parameter adjustment should be intelligent; it should respond quickly when the system's judgment of environmental changes is highly certain, and remain stable when the judgment is less clear, avoiding frequent oscillations between the two modes.

[0080] Specifically, in this embodiment, a parameter smooth transition model based on confidence damping can be constructed to ensure the response speed and visual comfort of the subsequent image parameter adjustment process.

[0081] The parameter smoothing transition model introduces a smoothing filtering mechanism with confidence damping. First, a basic set of optimized parameters is obtained by looking up a mapping table based on the identified environmental state and continuous environmental data. Then, brightness fine-tuning parameters are obtained based on illumination values. Next, instead of directly applying this target value, the current parameters are updated using a first-order inertial filter (i.e., exponential moving average). The filter coefficients are not fixed but dynamically adjusted based on the confidence level output from the previous step. High confidence corresponds to large coefficients, allowing the parameters to quickly approach the target; low confidence corresponds to small coefficients, causing the parameters to change slowly, acting as a damper to suppress unstable jumps.

[0082] For example, in this embodiment, the basic optimization parameter set and the brightness fine-tuning parameters can be calculated using the following formula:

[0083] ,

[0084] ,

[0085] in, For the basic target parameter set of the nth frame, A preset environment-parameter mapping table is used to map scene labels to a set of basic parameters (an example lookup table is shown in Table 1). This refers to the adjustment amount of the brightness parameter calculated based on the current illumination value (i.e., the brightness fine-tuning parameter). This is an adjustment coefficient for the brightness parameter adjustment amount, used to control the intensity of the adjustment; It is the natural logarithm function; For the current scenario The next reference illuminance; It is a positive number, used to prevent when A value of 0 results in the logarithmic function being undefined.

[0086] For example, in this embodiment, the first-order inertial filter that updates the current parameters can be represented by the following equation:

[0087] ,

[0088] ,

[0089] ,

[0090] in, This is the set of optimization parameters to be applied in the final frame n. The parameter set that was finally applied in the previous frame. These are the smoothing filter coefficients dynamically calculated for the nth frame; Let n be the final target parameter set for the nth frame. For an adjustment vector function, based on the included Environment description with continuous values Calculate the fine-tuning amount for each component in the basic parameter set; This represents the lower bound of the filter coefficients. This is the upper limit of the filter coefficients. This is a power-law exponent of the confidence level, used to enhance sensitivity to high confidence intervals.

[0091] Table 1. Preset Environment-Parameter Mapping Table

[0092]

[0093] It should be noted that the environment-parameter mapping table shown is only an example to explain the mapping rules in this embodiment. Those skilled in the art can add / delete content from the table according to actual circumstances. Specifically, when the output... =Nighttime rainy day, detected light intensity When it is 0.5 Lux, according to By looking up the table and loading the basic parameter set from the fourth row, we obtain... Subsequent fine-tuning steps are all based on these initial values. Table 1 shows the baseline state during sunny daytime: ample light and high signal-to-noise ratio mean no need to adjust brightness or color; only low-level noise reduction is needed to maintain purity, and moderate sharpening is required to see distant vehicle details. At night, brightening and noise reduction are key, requiring a significant increase in brightness; moderately increasing contrast to separate objects from the dark background; slightly reducing saturation to suppress color noise; enabling high-intensity noise reduction; and very low sharpening, otherwise noise will be amplified, resulting in pixelation. At tunnel entrances, wide dynamic range (WDR) should be prioritized. This parameter setting aims to assist the ISP's WDR algorithm, increasing brightness to enhance dark details; high contrast to enhance stereoscopic effect; and taking a median value for noise reduction and sharpening to accommodate different conditions inside and outside the tunnel. Nighttime rainy days present the most complex scenarios. Combating blur is crucial in these situations. Brightness should be slightly increased below that of a clear night to avoid overexposure of ground reflections; contrast should be significantly increased to penetrate rain and fog; saturation should be increased to make warning colors like lights stand out more; noise reduction should be set to the highest level; and sharpening should be kept to the lowest level to avoid sharpening rain streaks and splashes, which would create a cluttered image. In backlight / glare situations, highlights need to be reduced, shadows brightened, and overall brightness increased to clearly see backlit objects; contrast is used to enhance shadow details; saturation should be slightly reduced to avoid color overflow in bright areas; in these situations, the light is usually sufficient, so noise reduction is lower; higher sharpening is needed to define the outlines of objects blurred by strong light. In overcast conditions, de-dusting and contrast enhancement are needed, so brightness should be slightly increased; the core is to increase contrast to prevent the image from appearing flat; saturation should be moderately increased to make colors more vibrant; the light on overcast days is generally acceptable, so only low noise reduction and medium sharpening are required.

[0094] Step 4: Obtain This final optimized parameter set is sent to the DSP chip, which performs image processing operations based on the final optimized parameter set to generate the final optimized electronic rearview mirror image.

[0095] Specifically, firstly, the parameter set in vector form The data is unpacked to obtain specific parameter values. Each specific parameter value is then sent to the register or memory address of the corresponding processing unit in the DSP chip pipeline to prepare for subsequent processing.

[0096] For example, the final set of optimized parameters obtained after unpacking can be as follows:

[0097] ,

[0098] To enhance the integrity, we added a sharpening parameter to the original settings. Each parameter represents: Final brightness, Final contrast, Final color saturation, Final noise reduction intensity, Final sharpening intensity.

[0099] DSP chip will convert numerical values Write the value to the intensity control register of the noise reduction module in the image signal processor (ISP); and Send the values ​​to the tone mapping module to build the mapping curve that the module will use; Write the value to the saturation gain register of the color space adjustment module; Write to the intensity control register of the "Detail Enhancement / Sharpening Module".

[0100] The image data is processed according to the data in each register to obtain the final processed image data, which is then sent to the screen driver IC for display.

[0101] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing the display of electronic exterior rearview mirror images based on environmental perception, characterized in that, The method for optimizing the display of the rearview mirror image includes: S100: Real-time acquisition of vehicle environmental information and vehicle status information, and marking the image frames of the electronic rearview mirror with the acquired data with the same time reference, thereby constructing a frame synchronization environment vector. S200. Based on the frame synchronization environment vector, perform multimodal evidence fusion to form a probabilistic judgment of the environment state and obtain the final state decision. And assess the confidence level of the probabilistic judgment to address the challenges of sensor noise or failure; S300. The final set of optimized parameters is calculated based on the results of the state decision using the parametric smooth transition model. The parameter smooth transition model includes: finding and obtaining the basic optimization parameter set from the environment-parameter mapping table based on the results of state decisions. Then, the brightness fine-tuning parameters calculated from the illumination value are combined to fine-tune the basic optimization parameter set, and a first-order inertial filter is used to update the basic optimization parameter set. The filter coefficients of the filter are dynamically adjusted in combination with the confidence level. S400 sends the final optimized parameter set to the DSP chip. The DSP chip performs image processing operations based on the final optimized parameter set to generate the final optimized electronic rearview mirror image. The basic optimization parameter set and brightness fine-tuning parameters are calculated using the following formula: ， ， in, For the basic target parameter set of the nth frame, For the preset environment-parameter mapping table, This is the adjustment amount for the brightness parameter calculated based on the current illuminance value. This is the adjustment coefficient for the brightness parameter adjustment amount; It is the natural logarithm function; For the current scenario The next reference illuminance; To ensure that the logarithmic function always contains positive numbers; The illumination value is calculated to be perfectly aligned with the nth frame image in time.

2. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 1, characterized in that, The frame synchronization environment vector is constructed according to the following method: For continuously changing physical quantities, alignment is achieved through linear interpolation. For discrete state variables, a forward-filling strategy is adopted. All data aligned with the completion time of a single frame image is encapsulated into a standardized frame synchronization environment vector data packet.

3. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 2, characterized in that, When the continuously changing physical quantity is the illumination value, linear interpolation alignment is achieved using the following formula: ， in, Let be the illumination values ​​that are perfectly aligned temporally for the nth frame image. The most recent illumination sensor reading prior to the image frame timestamp. This represents the first illumination sensor reading after the image frame timestamp; For timestamps, For the nth frame image, Represents the precise timestamp of the nth frame of the image; For reading timestamp, Indicates reading timestamp; When the discrete state variable is a weather state, the forward filling strategy is implemented using the following formula: ， in, The time-aligned weather conditions determined for the nth frame image; Weather status value; Indicates constraints. Weather status value timestamp, This represents the maximum value.

4. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 1, characterized in that, The multimodal evidence fusion is calculated using an environmental state probability model. The evidence score for each potential environmental state is obtained by weighted summation of membership degrees from different sensors after transformation. The scenario with the highest score is the current state, and the difference between the highest and second-highest scores is used to quantify the confidence level of this judgment. The membership degree is obtained by smoothly mapping the environmental data from the frame synchronization environment vector to the interval [0,1] using the Sigmoid function.

5. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 4, characterized in that, The membership degree is calculated using the following formula for a basic lighting scene: ， ， in, This represents the degree of membership of the environment to a daytime scene under given illumination values. This represents the membership degree of the environment to a nighttime scene under given illumination values; The symbol for the natural exponential function is . These are the curve control parameters for daytime. For curve control parameters at night, The preset daytime light threshold, The preset light threshold for nighttime; The calculated illumination values ​​are those that are perfectly aligned in time for the nth frame image.

6. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 4, characterized in that, When the potential environmental state is a tunnel entrance scenario, the formula for calculating the evidence score is as follows: ， ， in, Indicates in a given data packet Under the given conditions, the overall score for nighttime scenes; Indicates in a given data packet Under the given conditions, the overall score for entering the tunnel scenario; For the weighting coefficients of the light sensor, Nighttime membership is based on the illumination sensor. The weighting coefficient for image brightness. Nighttime membership is based on image brightness; To calculate the illumination sensor weighting coefficients for the overall score of the tunnel scene, To calculate the image brightness weighting coefficients for the overall score of the tunnel scene, The weighting coefficient for the rate of change of illumination. The membership degree of the tunnel scene is based on the rate of change of illumination.

7. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 6, characterized in that, The final state decision is calculated using the following formula: ， The confidence level is calculated using the following formula: ， in, The most likely environmental state for the final decision in frame n. It is an element in the set of all environment states. For current decision confidence level To adjust the coefficients of the confidence curve, To achieve the highest score, This is the second highest score.

8. The method for optimizing the display of electronic exterior rearview mirror images based on environmental perception according to claim 1, characterized in that, The first-order inertial filter is shown in the following equation: ， ， ， in, This is the set of optimization parameters to be applied in the final frame n. The parameter set that was finally applied in the previous frame. These are the smoothing filter coefficients dynamically calculated for the nth frame; Let n be the final target parameter set for the nth frame. For an adjustment vector function, This represents the lower bound of the filter coefficients. This is the upper limit of the filter coefficients. This is a power-law exponent of the confidence level, used to enhance sensitivity to high confidence intervals; This is the set of basic target parameters for the nth frame; For the current scenario The confidence level.

9. An electronic rearview mirror, comprising an image sensor, a processor, a display device, and a memory, wherein the processor and the memory are electrically connected to the image sensor and the display device, characterized in that, The memory stores computer programs. When the computer program is executed by the processor, it implements the method for optimizing the display of electronic exterior rearview mirror images based on environmental perception as described in any one of claims 1 to 8.

Images