Automatic emergency braking pre-judgment method based on vehicle type recognition and brake lamp response characteristics

Abstract

The invention discloses an automatic emergency braking pre-judgment method based on vehicle type recognition and brake lamp response characteristics, and the method achieves the high-confidence judgment of the type of a front vehicle through the construction of a double-branch vehicle recognition network fusing brand semantics and model fine-grained characteristics. A vehicle type brake lamp response characteristic query table is further introduced, the brake signal response difference of the vehicle design level serves as prior information to be fused into the risk assessment process, and when the deceleration behavior without the brake lamp explicit prompt is detected, the collision time threshold value of the automatic emergency braking system is dynamically corrected. Therefore, the method achieves the conversion from the passive response depending on the explicit braking signal to the active pre-judgment control fusing the vehicle type prior and dynamic characteristics, and can effectively improve the longitudinal driving safety of the new energy vehicles in the mixed driving scene.

Description

Technical Field

[0001] This invention belongs to the field of intelligent vehicle safety control technology, and in particular relates to an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics. Background Technology

[0002] With the increasing number of new energy vehicles on the road, urban roads and highways are gradually entering a traffic pattern where new energy vehicles and traditional fuel vehicles share the road. New energy vehicles generally employ energy recovery braking strategies, which generate longitudinal deceleration when the accelerator pedal is released. However, some models do not illuminate their brake lights during this process. Since drivers of following vehicles and existing driver assistance systems typically rely on brake lights as the primary visible signal of the vehicle ahead's deceleration intention, when deceleration occurs without brake light indication, the perception of the deceleration status of the vehicle ahead is easily delayed. In heavy traffic or when following distances are close, this significantly increases the risk of rear-end collisions, becoming a typical safety hazard in environments where new energy vehicles share the road.

[0003] To address the risk of deceleration from the vehicle ahead, existing automatic emergency braking systems primarily rely on relative distance, relative speed, and longitudinal acceleration information obtained from millimeter-wave radar or visual sensors, triggering braking intervention based on collision time or distance thresholds. Some methods aim for earlier intervention by increasing acceleration detection sensitivity or adjusting trigger thresholds earlier; however, these methods are largely based on instantaneous dynamic characteristics and are susceptible to interference from factors such as road gradient and natural coasting. They struggle to accurately distinguish between active deceleration and non-braking deceleration states, posing a risk of false triggering and impacting system stability and driving comfort.

[0004] Furthermore, existing methods generally treat the vehicle in front as a target with consistent behavior, lacking differentiation on the design differences in brake light response strategies among different vehicle models. This makes it difficult to address the structural perception deficiencies caused by new energy vehicles not illuminating their brake lights during the energy recovery deceleration phase. Therefore, there is an urgent need for an active prediction method that can combine prior information about the vehicle model in front with deceleration behavior characteristics. This method should be able to identify potential deceleration risks in advance without relying on explicit brake light warnings, and reasonably adjust the triggering timing of the automatic emergency braking system to improve longitudinal driving safety in mixed-traffic scenarios involving new energy vehicles. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics. This method solves the problem that the lack of explicit brake light prompts during the energy recovery deceleration process of new energy vehicles can easily lead to delayed perception of the deceleration behavior of the vehicle in front by the driver of the following vehicle, resulting in the risk of rear-end collisions.

[0006] The technical problem to be solved by this invention is achieved through the following technical solution: This invention provides an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics, comprising: Acquire the target image of the target vehicle ahead, and preprocess the target image to obtain the target image tensor; The target image tensor is input into the pre-constructed brand recognition branch network and model recognition branch network respectively to obtain the brand prediction probability distribution and model prediction probability distribution; and random forward propagation is performed based on the Monte Carlo random deactivation mechanism to obtain random brand prediction probability distribution and random model prediction probability distribution. Based on the brand prediction probability distribution, model prediction probability distribution, random brand prediction probability distribution and random model prediction probability distribution, the fusion model prediction probability distribution is determined. The vehicle model identification result is determined based on the maximum a posteriori probability principle and the fused model prediction probability distribution. Based on the preset vehicle model brake light response characteristic lookup table and vehicle model identification results, determine the brake light response characteristics of the target vehicle ahead; If the brake light response characteristics indicate that the target vehicle ahead is decelerating without a visible brake light warning, the collision time threshold of the automatic emergency braking system will be adjusted.

[0007] This invention provides an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics. It constructs a dual-branch vehicle recognition architecture that integrates brand semantics and fine-grained model features, and introduces a prediction stability-driven gating fusion mechanism to achieve high-confidence identification of the preceding vehicle's model. Furthermore, it introduces a vehicle model brake light response characteristic lookup table, incorporating differences in braking signal response at the vehicle design level as prior information into the risk assessment process. When deceleration behavior without explicit brake light warning is detected, the collision time threshold of the automatic emergency braking system is dynamically corrected. Thus, this invention achieves a shift from a passive response relying on explicit braking signals to an active predictive control integrating vehicle model priors and dynamic characteristics, effectively improving longitudinal driving safety in mixed-traffic scenarios involving new energy vehicles.

[0008] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the vehicle model recognition process provided in an embodiment of the present invention. Detailed Implementation

[0010] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0011] This invention provides an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics. See also... Figure 1 and Figure 2 The method includes the following steps: S10. Obtain the target image of the target vehicle ahead, and preprocess the target image to obtain the target image tensor.

[0012] For example, the target image is scaled and pixel normalized to construct a target image tensor. ,in, These represent the batch size (the number of images processed simultaneously during a single training or inference process), image pixel height, image pixel width, and number of channels, respectively.

[0013] S20. Input the target image tensor into the pre-constructed brand recognition branch network and model recognition branch network respectively to obtain the brand prediction probability distribution and model prediction probability distribution; and perform random forward propagation based on the Monte Carlo random deactivation mechanism to obtain the random brand prediction probability distribution and random model prediction probability distribution.

[0014] For example, the brand identification branch network can use the DenseNet-201 network, and the model identification branch network can use the ResNet-50 network.

[0015] First, the target image tensor is input into the brand recognition branch network to obtain the original brand output value. Then, the original brand output value is normalized using Softmax to obtain the brand prediction probability distribution, as follows:

[0016] in, Indicates time The original output value of the brand, , Indicates the total number of brand categories. This indicates a brand identity branch network mapping. Indicates time Brand prediction probability distribution , This represents the Softmax function.

[0017] Similarly, the target image tensor is input into the model recognition branch network to obtain the original model output value. The original model output value is then Softmax normalized to obtain the model prediction probability distribution, as shown below:

[0018] in, Indicates time The original output value of the model. , Indicates the total number of model categories. This indicates the mapping of the model identification branch network. Indicates time Model prediction probability distribution .

[0019] Secondly, a Monte Carlo random deactivation mechanism is employed to perform operations on the target image tensor. After a random forward propagation, the predicted probability distributions for random brands and random models are obtained, as follows:

[0020] in, Indicates time No. The probability distribution of random brand predictions obtained from the next random forward propagation. Indicates time No. The probability distribution of random model prediction obtained by the next random forward propagation. This represents the total number of random forward propagations. and These represent the brand recognition branch network and the model recognition branch network in time, respectively. No. The Dropout random mask set during the second propagation.

[0021] S30. Based on the brand prediction probability distribution, model prediction probability distribution, random brand prediction probability distribution, and random model prediction probability distribution, determine the fusion model prediction probability distribution.

[0022] Optionally, step S30 may specifically include: S301. Based on the probability distribution of random brand prediction and the probability distribution of random model prediction, determine the stability of brand prediction and the stability of model prediction.

[0023] Optionally, the brand prediction stability and the model prediction stability are expressed as follows:

[0024] in, 、 They represent time respectively Brand forecast stability and model forecast stability, Indicates the total number of brand categories. Indicates the brand category index. Indicates the total number of model categories. Indicates the model category index. This represents variance calculation. This represents the probability distribution for predicting random brands. This represents the probability distribution for random model prediction. This represents the total number of random forward propagations.

[0025] S302. Determine the brand relative stability ratio based on brand predicted stability and model predicted stability.

[0026] Optionally, the brand relative stability ratio is expressed as:

[0027] in, Indicates time The brand's relative stability ratio This represents a constant used to prevent the denominator from being zero.

[0028] S303. Map the brand relative stability ratio to a gating weight, and perform exponential smoothing on the gating weight to obtain a smoothing weight.

[0029] Optionally, the gating weights are represented as follows:

[0030] in, Indicates time Gating weights, This represents the Sigmoid function. , The weight parameters represent the gating mapping. The bias parameter represents the gating mapping; The smoothing weights are represented as:

[0031] in, Indicates time Smoothing weights, This represents the smoothing coefficient.

[0032] S304. Based on the brand prediction probability distribution, model prediction probability distribution and smoothing weight, determine the fusion model prediction probability distribution.

[0033] Optionally, the model prediction probability distribution is fused and expressed as:

[0034] in, Indicates time The probability distribution of the fusion model This represents the intermediate vector obtained by mapping the brand prediction probability distribution to the model space. Represents a linear mapping matrix. Indicates time The brand prediction probability distribution , Indicates time Model prediction probability distribution , Indicates batch size.

[0035] S40. Based on the maximum a posteriori probability principle and the fusion model prediction probability distribution, determine the vehicle model identification result.

[0036] Optionally, the vehicle model identification result is represented as:

[0037] in, Indicates time Vehicle model identification results Indicates time Predicted vehicle model Indicates the model category index. express In Model Category Index The probability components on, Indicates the confidence level.

[0038] S50. Based on the preset vehicle model brake light response characteristic lookup table and vehicle model identification results, determine the brake light response characteristics of the target vehicle ahead.

[0039] For example, a preset vehicle model brake light response characteristic lookup table is represented as follows:

[0040] in, This indicates a preset vehicle model brake light response characteristic lookup table, i.e., a vehicle model brake light response characteristic lookup table. This indicates the brake light response indicator for the corresponding vehicle model under energy recovery deceleration conditions. This indicates that the vehicle's brake taillights will not illuminate during the regenerative braking phase. This indicates that the brake taillights will always illuminate when the vehicle is decelerating.

[0041] Based on the vehicle model identification results, the brake light response characteristics corresponding to the vehicle model identification results can be determined by querying the preset vehicle model brake light response characteristic lookup table.

[0042] S60. If, based on the brake light response characteristics, it is determined that the target vehicle ahead is exhibiting deceleration behavior without explicit brake light warning, then the collision time threshold of the automatic emergency braking system is corrected.

[0043] Optionally, step S60 may specifically include: S601, Obtain the longitudinal acceleration of the target vehicle ahead.

[0044] For example, the longitudinal acceleration of a target vehicle ahead can be obtained using millimeter-wave radar.

[0045] S602. Based on the brake light response characteristics, longitudinal acceleration, and preset deceleration judgment threshold, determine whether the target vehicle ahead is decelerating without a visible brake light warning.

[0046] Optionally, step S602 may specifically include: like

[0047] in, This indicates the brake light response characteristics of the target vehicle ahead. This indicates that the vehicle model corresponding to the target vehicle ahead will not illuminate its brake taillights during the deceleration and recovery phase. This indicates that the vehicle model corresponding to the target vehicle ahead will illuminate its brake taillights when decelerating. This represents the longitudinal acceleration of the target vehicle ahead. This indicates the preset deceleration threshold. Then determine whether the target vehicle ahead is slowing down without a visible brake light warning.

[0048] S603. If so, the collision time threshold of the automatic emergency braking system shall be corrected.

[0049] For example, after determining that the target vehicle ahead is decelerating without explicit brake light warning, based on the probability components of step S40... The collision time threshold of the Automatic Emergency Braking (AEB) system is corrected.

[0050] Optionally, the correction calculation process can be expressed as follows:

[0051] in, This represents the corrected collision time threshold. This represents the standard AEB collision time threshold. This represents the maximum advance prediction time constant obtained from the calibration of the existing AEB system. This represents the acceleration threshold corresponding to natural gliding or slight resistance. This represents the normalization coefficient for deceleration intensity.

[0052] This invention presents an automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics. It constructs a dual-branch vehicle recognition architecture that integrates brand semantics and fine-grained model features, and introduces a prediction stability-driven gating fusion mechanism to achieve high-confidence identification of the preceding vehicle's model. Furthermore, it introduces a vehicle model brake light response characteristic lookup table, incorporating differences in braking signal response at the vehicle design level as prior information into the risk assessment process. When deceleration behavior without explicit brake light warning is detected, the collision time threshold of the automatic emergency braking system is dynamically corrected. Thus, this invention achieves a shift from a passive response relying on explicit braking signals to an active predictive control integrating vehicle model priors and dynamic characteristics, effectively improving longitudinal driving safety in mixed-traffic scenarios involving new energy vehicles.

[0053] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.

[0054] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0055] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0056] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. An automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics, characterized in that, include: Acquire a target image of the target vehicle ahead, and preprocess the target image to obtain a target image tensor; The target image tensor is input into the pre-constructed brand recognition branch network and model recognition branch network respectively to obtain the brand prediction probability distribution and model prediction probability distribution; and random forward propagation is performed based on the Monte Carlo random deactivation mechanism to obtain random brand prediction probability distribution and random model prediction probability distribution. Based on the brand prediction probability distribution, the model prediction probability distribution, the random brand prediction probability distribution, and the random model prediction probability distribution, the fusion model prediction probability distribution is determined. Based on the maximum a posteriori probability principle and the fused model prediction probability distribution, the vehicle model identification result is determined; Based on the preset vehicle model brake light response characteristic lookup table and the vehicle model identification result, the brake light response characteristics of the target vehicle ahead are determined; If, based on the brake light response characteristics, it is determined that the target vehicle ahead exhibits deceleration behavior without explicit brake light warning, then the collision time threshold of the automatic emergency braking system is corrected.

2. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 1, characterized in that, The step of determining the fused model prediction probability distribution based on the brand prediction probability distribution, the model prediction probability distribution, the random brand prediction probability distribution, and the random model prediction probability distribution includes: Based on the random brand prediction probability distribution and the random model prediction probability distribution, the brand prediction stability and model prediction stability are determined. Based on the predicted stability of the brand and the predicted stability of the model, determine the brand relative stability ratio; The brand relative stability ratio is mapped to a gating weight, and the gating weight is exponentially smoothed to obtain a smoothing weight. Based on the brand prediction probability distribution, the model prediction probability distribution, and the smoothing weight, the fused model prediction probability distribution is determined.

3. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 2, characterized in that, If, based on the brake light response characteristics, it is determined that the target vehicle ahead exhibits deceleration behavior without explicit brake light warning, then the collision time threshold of the automatic emergency braking system is corrected, including: Obtain the longitudinal acceleration of the target vehicle ahead; Based on the brake light response characteristics, longitudinal acceleration, and preset deceleration determination threshold, it is determined whether the target vehicle ahead exhibits deceleration behavior without explicit brake light warning. If so, the collision time threshold of the automatic emergency braking system will be corrected.

4. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 2, characterized in that, The brand prediction stability and the model prediction stability are respectively expressed as follows: in, 、 They represent time respectively Brand forecast stability and model forecast stability, Indicates the total number of brand categories. Indicates the brand category index. Indicates the total number of model categories. Indicates the model category index. This represents variance calculation. This represents the probability distribution for predicting random brands. This represents the probability distribution for random model prediction. This represents the total number of random forward propagations.

5. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 4, characterized in that, The brand relative stability ratio is expressed as: in, Indicates time The brand's relative stability ratio This represents a constant used to prevent the denominator from being zero.

6. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 5, characterized in that, The gating weight is represented as follows: in, Indicates time Gating weights, This represents the Sigmoid function. , The weight parameters represent the gating mapping. The bias parameter represents the gating mapping. The smoothing weights are expressed as follows: in, Indicates time Smoothing weights, This represents the smoothing coefficient.

7. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 6, characterized in that, The predicted probability distribution of the fused model is expressed as follows: in, Indicates time The probability distribution of the fusion model This represents the intermediate vector obtained by mapping the brand prediction probability distribution to the model space. Represents a linear mapping matrix. Indicates time The brand prediction probability distribution , Indicates time Model prediction probability distribution , Indicates batch size.

8. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 7, characterized in that, The vehicle model identification result is represented as follows: in, Indicates time Vehicle model identification results Indicates time Predicted vehicle model Indicates the model category index. express In Model Category Index The probability components on, Indicates the confidence level.

9. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 3, characterized in that, The step of determining whether the target vehicle ahead exhibits deceleration behavior without explicit brake light warning, based on the brake light response characteristics, longitudinal acceleration, and a preset deceleration threshold, includes: like in, This indicates the brake light response characteristics of the target vehicle ahead. This indicates that the vehicle model corresponding to the target vehicle ahead will not illuminate its brake taillights during the deceleration and recovery phase. This indicates that the vehicle model corresponding to the target vehicle ahead will illuminate its brake taillights when decelerating. This represents the longitudinal acceleration of the target vehicle ahead. This indicates the preset deceleration threshold. Then determine whether the target vehicle ahead is decelerating without a visible brake light warning.

10. The automatic emergency braking prediction method based on vehicle model recognition and brake light response characteristics according to claim 9, characterized in that, The process for correcting the collision time threshold of the automatic emergency braking system is as follows: in, This represents the corrected collision time threshold. This represents the standard AEB collision time threshold. This represents the maximum advance prediction time constant. This represents the acceleration threshold corresponding to natural gliding or slight resistance. This represents the normalization coefficient for deceleration intensity.

Images