Tunnel distributed illumination compensation processing method and system

Abstract

The present invention provides a tunnel distributed illumination compensation processing method and system, which performs parameter learning on an illumination compensation guidance model by acquiring a template tunnel distributed illumination change data sequence, and combines the tunnel distributed illumination compensation guidance information to achieve linkage optimization of the illumination compensation guidance model and the guidance information. The generated iterative illumination compensation guidance model can provide accurate illumination compensation guidance for the illumination area in any tunnel distributed illumination change data according to the guidance of the iterative tunnel distributed illumination compensation guidance information, effectively improving the accuracy and efficiency of illumination compensation in the tunnel, ensuring the uniformity and comfort of illumination in the tunnel, and thus improving the safety of tunnel passage and driving experience.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a tunnel distributed illumination compensation processing method and system. Background Art

[0002] In tunnel lighting systems, the uniformity and suitability of illumination are crucial to ensuring driving safety and improving the driving experience. However, due to the complex and changing environment inside the tunnel, such as vehicle entry and exit, weather changes, and differences in tunnel structure, the illumination inside the tunnel is often uneven or insufficient. This unstable illumination not only affects the driver's vision, but may also cause traffic accidents, posing a serious threat to driving safety.

[0003] To solve this problem, traditional tunnel lighting systems usually use fixed illuminance settings, but this method cannot adapt to the dynamic changes in illuminance in the tunnel. With the development of technology, some advanced tunnel lighting systems have begun to use distributed illuminance compensation technology to achieve uniform illuminance distribution by real-time monitoring of the illuminance in the tunnel and dynamically adjusting it.

[0004] However, existing distributed illumination compensation technologies still have some problems. For example, these technologies often rely on a single illumination compensation model, which may not accurately adapt to the changes in illumination in different tunnels, different time periods, and different weather conditions. In addition, the existing illumination compensation process usually lacks effective guidance information, resulting in unsatisfactory compensation effects, and may even lead to over-compensation or under-compensation. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a tunnel distributed illumination compensation processing method, the method comprising:

[0006] Acquire a template tunnel distributed illumination change data sequence for learning parameter information of an illumination compensation guidance model; wherein the template tunnel distributed illumination change data sequence includes a plurality of template tunnel distributed illumination change data, the template tunnel distributed illumination change data includes illumination areas, and the illumination compensation guidance model is used to provide illumination compensation guidance for the illumination areas in the template tunnel distributed illumination change data;

[0007] Acquiring tunnel distributed illumination compensation guidance information; the tunnel distributed illumination compensation guidance information is used to guide the illumination compensation guidance process executed by the illumination compensation guidance model;

[0008] The template tunnel distributed illumination change data sequence is used to perform linkage optimization on the illumination compensation guidance model and the tunnel distributed illumination compensation guidance information to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information; wherein the iterative illumination compensation guidance model is used to provide illumination compensation guidance for an illumination area in any tunnel distributed illumination change data based on the guidance of the iterative tunnel distributed illumination compensation guidance information.

[0009] On the other hand, an embodiment of the present invention also provides a tunnel distributed light illumination compensation processing system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0010] Based on the above aspects, the embodiment of the present application learns the parameters of the illumination compensation guidance model by acquiring the template tunnel distributed illumination change data sequence, and combines the tunnel distributed illumination compensation guidance information to achieve the linkage optimization of the illumination compensation guidance model and the guidance information. The generated iterative illumination compensation guidance model can provide accurate illumination compensation guidance for the illumination area in any tunnel distributed illumination change data according to the guidance of the iterative tunnel distributed illumination compensation guidance information, effectively improving the accuracy and efficiency of illumination compensation in the tunnel, ensuring the uniformity and comfort of the illumination in the tunnel, thereby improving the safety of tunnel passage and driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the execution flow of the tunnel distributed illumination compensation processing method provided by an embodiment of the present invention.

[0012] Figure 2 It is a schematic diagram of the hardware architecture of a tunnel distributed illumination compensation processing system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 1 is a flow chart of a tunnel distributed illumination compensation processing method provided by an embodiment of the present invention. The tunnel distributed illumination compensation processing method is introduced in detail below.

[0014] Step S110, obtaining a template tunnel distributed illumination change data sequence for learning parameter information of an illumination compensation guidance model. The template tunnel distributed illumination change data sequence includes a plurality of template tunnel distributed illumination change data, the template tunnel distributed illumination change data includes illumination areas, and the illumination compensation guidance model is used to provide illumination compensation guidance for the illumination areas in the template tunnel distributed illumination change data.

[0015] In this embodiment, a tunnel lighting system includes multiple lighting areas, and the lighting conditions of each lighting area may change with time, vehicle flow, external weather and other factors.

[0016] The server obtains the template tunnel distributed illumination change data sequence from the database of the tunnel lighting system. For example, this sequence contains the tunnel illumination data for each hour in the past month. These data are stored in a specific format, and each data point (i.e., the template tunnel distributed illumination change data) corresponds to a specific illumination area in the tunnel, such as the entrance area, the driving lane area, the curve area, etc. For the entrance area, the data may contain information such as the illumination intensity value, illumination angle, and color temperature of the light source at different times. The server obtains this data to allow the illumination compensation guidance model to learn the illumination change patterns of different illumination areas under various conditions, so as to provide accurate illumination compensation guidance later.

[0017] Step S120, obtaining tunnel distributed illumination compensation guidance information. The tunnel distributed illumination compensation guidance information is used to guide the illumination compensation guidance process executed by the illumination compensation guidance model.

[0018] In this embodiment, in the same tunnel lighting management system scenario, the server obtains tunnel distributed illumination compensation guidance information.

[0019] For example, the text description in the basic guidance data may be the minimum requirements for the light intensity in different areas of the tunnel in the safety standard document (such as the light intensity in the driving lane area should be kept within a certain lux range to ensure clear vision for the driver). The area framing information may be the coordinate range that divides the tunnel into different functional areas (such as emergency parking strips, normal driving lanes, etc.), which helps the illumination compensation guidance model to clarify the focus of illumination compensation in different areas. The cursor point set can be the coordinate information of the pre-set light monitoring points at key locations in the tunnel (such as curve turning points, middle positions of the tunnel, etc.). The light intensity distribution diagram is a graphical description of the ideal tunnel light intensity distribution obtained based on historical data or design standards.

[0020] The server uses the second illumination analysis unit in the illumination compensation guidance model to feature encode these basic guidance data and generate the guidance encoding representation of the illumination compensation guidance model as the tunnel distributed illumination compensation guidance information. These guidance information will play a guiding role in the illumination compensation guidance process to ensure that the illumination compensation meets the actual needs and safety standards of tunnel lighting.

[0021] Step S130, using the template tunnel distributed illumination change data sequence to perform linkage optimization on the illumination compensation guidance model and the tunnel distributed illumination compensation guidance information, to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information. The iterative illumination compensation guidance model is used to provide illumination compensation guidance for an illumination area in any tunnel distributed illumination change data according to the guidance of the iterative tunnel distributed illumination compensation guidance information.

[0022] In this embodiment, it is assumed that the server divides the acquired template tunnel distributed illumination change data sequence into a first template subsequence and a second template subsequence in chronological order. The first template subsequence contains data for the first half of a month, and the second template subsequence contains data for the second half of a month.

[0023] Next, the first template subsequence is used to update the compensation guidance parameters of the illumination compensation guidance model to generate an iterative illumination compensation guidance model:

[0024] For each first template tunnel distributed illumination change data a in the first template subsequence (for example, taking the illumination data of the entrance area at a certain moment as an example), it is annotated with the first template illumination compensation guidance annotation data (such as the ideal illumination compensation value obtained based on the actual road conditions and lighting effect evaluation at that time).

[0025] Based on the tunnel distributed illumination compensation guidance information Y0 (the guidance code generated previously represents relevant information), the server performs illumination compensation guidance on the distributed illumination change data a of the first template tunnel through the illumination compensation guidance model, and generates an illumination compensation guidance estimation result (for example, the amount of illumination compensation that should be made to the entrance area calculated by the model based on the current guidance information and its own algorithm).

[0026] Then, based on the deviation between the illumination compensation guidance estimation result and the first template illumination compensation guidance annotation data, the model learning cost data of the first template tunnel distributed illumination change data a is calculated. For example, if the estimation result differs greatly from the annotation data, it means that the model's prediction at this data point is inaccurate, and the model learning cost will be high.

[0027] The above operation is performed on all X first template tunnel distributed illumination change data in the first template subsequence, and the illumination compensation guidance model is back-propagated and updated according to the determined model learning cost data of the X first template tunnel distributed illumination change data to generate an iterative illumination compensation guidance model.

[0028] Furthermore, the second template subsequence is used to update the illumination guidance strategy of the tunnel distributed illumination compensation guidance information to generate iterative tunnel distributed illumination compensation guidance information:

[0029] For each second template tunnel distributed illumination change data b in the second template subsequence (eg illumination data of a curved area at a certain moment), it is annotated with second template illumination compensation guidance annotation data (eg illumination adjustment value actually required for the curved area at that time).

[0030] The server calls the iterative illumination compensation guidance model (which has just been updated) to provide illumination compensation guidance for the distributed illumination change data b of the second template tunnel, and generates illumination compensation guidance estimation results (such as the model's new prediction value for illumination compensation in the curved area).

[0031] According to the deviation between the illumination compensation guidance estimation result and the second template illumination compensation guidance annotation data, the model learning cost data of the second template tunnel distributed illumination change data b is calculated.

[0032] According to the determined model learning cost data of the Z second template tunnel distributed illumination change data, the tunnel distributed illumination compensation guidance information Y0 is cyclically updated to generate iterative tunnel distributed illumination compensation guidance information Y1.

[0033] In another possible implementation, the first template subsequence and the second template subsequence are divided in the same manner as above.

[0034] Taking a first template tunnel distributed illumination change data a in the first template subsequence (such as data of a certain section of lane in the middle of the tunnel at a certain time) as an example, assume that the tunnel distributed illumination compensation guidance information before optimization is represented as Y0, and the parameter information to be updated in the illumination compensation guidance model is Q0.

[0035] The server updates the parameter information Q0 to be updated in the illumination compensation guidance model based on the tunnel distributed illumination compensation guidance information Y0 and the first template tunnel distributed illumination change data a, and generates iterative learning parameter information Q1 (for example, the parameters related to lane illumination compensation in the model are adjusted according to the illumination information and guidance information Y0 in the data a).

[0036] Next, based on the iterative learning parameter information Q1, a second template tunnel distributed illumination change data b in the second template subsequence (such as data of a lane at another time) is used to update the tunnel distributed illumination compensation guidance information Y0 to generate iterative tunnel distributed illumination compensation guidance information Y1 (for example, the strategy for lane illumination compensation in the guidance information is adjusted according to the new model parameter Q1 and the illumination information in the data b).

[0037] The iterative tunnel distributed illumination compensation guidance information Y1 is used as the new tunnel distributed illumination compensation guidance information, and the iterative learning parameter information Q1 is used as the new learning parameter information, and X first template tunnel distributed illumination change data and Z second template tunnel distributed illumination change data are used for cyclic update. For example, the above process is repeated until the iterative illumination compensation guidance model meets the training termination requirements (such as the error of the model on the validation set reaches the minimum value or no longer decreases significantly), and the iterative illumination compensation guidance model and the iterative tunnel distributed illumination compensation guidance information are generated.

[0038] Based on the above steps, the embodiment of the present application learns the parameters of the illumination compensation guidance model by acquiring the template tunnel distributed illumination change data sequence, and combines the tunnel distributed illumination compensation guidance information to achieve the linkage optimization of the illumination compensation guidance model and the guidance information. The generated iterative illumination compensation guidance model can provide accurate illumination compensation guidance for the illumination area in any tunnel distributed illumination change data according to the guidance of the iterative tunnel distributed illumination compensation guidance information, effectively improving the accuracy and efficiency of illumination compensation in the tunnel, ensuring the uniformity and comfort of the illumination in the tunnel, thereby improving the safety of tunnel passage and driving experience.

[0039] In a possible implementation, the template tunnel distributed illumination change data sequence is divided into a first template subsequence and a second template subsequence. Step S130 includes:

[0040] Step S131, using the first template subsequence to update the compensation guidance parameters of the illumination compensation guidance model to generate an iterative illumination compensation guidance model, and using the second template subsequence to update the illumination guidance strategy of the tunnel distributed illumination compensation guidance information to generate iterative tunnel distributed illumination compensation guidance information.

[0041] In detail, in the tunnel lighting management system scenario mentioned above, the first template subsequence contains lighting data of a specific time period (eg, the first half of a month) in the tunnel lighting system database.

[0042] For each first template tunnel distributed illumination change data a in the first template subsequence, take the data of a certain moment in the tunnel entrance area as an example. This data contains information such as the illumination intensity and angle of each lighting fixture in the entrance area, and is annotated with the first template illumination compensation guidance annotation data, which is the ideal illumination compensation value obtained based on the actual traffic flow, safety monitoring data and manual evaluation of the entrance area at that time.

[0043] The server performs illumination compensation guidance on the first template tunnel distributed illumination change data a through the illumination compensation guidance model based on the existing tunnel distributed illumination compensation guidance information Y0. For example, the first illumination analysis unit in the illumination compensation guidance model first analyzes the illumination data of the entrance area to obtain the characteristics of the current illumination distribution, and then the compensation guidance generation unit calculates the illumination compensation guidance estimation result based on the guidance information Y0, that is, the illumination compensation adjustment amount that the model believes should be made in the entrance area.

[0044] Next, the server calculates the model learning cost data of the first template tunnel distributed illumination change data a based on the deviation between the illumination compensation guidance estimation result and the first template illumination compensation guidance annotation data. If the estimation result differs greatly from the annotation data, it means that the model performs poorly at this data point, and the corresponding model learning cost data will be higher. For example, if the model estimates that the entrance area needs to increase the illumination intensity by 10%, but the annotation data shows that it should increase by 15%, then this deviation will be reflected in the calculation of the model learning cost data.

[0045] The server performs this operation on all X first template tunnel distributed illumination change data in the first template subsequence to obtain X model learning cost data. Then, the illumination compensation guidance model is back-propagated and updated based on these model learning cost data. During the back-propagation process, the model adjusts the parameters within the model (such as weights in the neural network, etc.) according to the learning cost of each data point, thereby generating an iterative illumination compensation guidance model. This iterative model can better adapt to the illumination compensation requirements of different illumination areas such as tunnel entrance areas.

[0046] The second template subsequence contains illumination data of another specific time period (eg, the second half of a month) in the tunnel lighting system database.

[0047] Take a second template tunnel distributed illumination change data b in the second template subsequence as an example, such as the data at a certain moment in the curve area, which is annotated with the second template illumination compensation guidance annotation data, which is the ideal illumination compensation value obtained based on the actual situation of the curve area at that time (such as curve curvature, vehicle speed and other factors).

[0048] The server calls the iterative illumination compensation guidance model just generated, performs illumination compensation guidance on the distributed illumination change data b of the second template tunnel, and generates an illumination compensation guidance estimation result. For example, the model calculates an estimated value of the illumination compensation amount of the curved area based on the illumination data of the curved area and its own algorithm.

[0049] Then, the server calculates the model learning cost data of the second template tunnel distributed illumination change data b according to the deviation between the illumination compensation guidance estimation result and the second template illumination compensation guidance annotation data.

[0050] The server performs such an operation on all Z second template tunnel distributed illumination change data in the second template subsequence to obtain Z model learning cost data. Based on these model learning cost data, the tunnel distributed illumination compensation guidance information Y0 is cyclically updated. For example, if the data of a certain curve area shows that the model has a large deviation in illumination compensation guidance, then when updating the guidance information, the part related to the illumination guidance strategy of the curve area will be adjusted to generate iterative tunnel distributed illumination compensation guidance information Y1. This iterative guidance information can more accurately guide the illumination compensation guidance model to perform illumination compensation on illumination areas such as tunnel curves.

[0051] Step S132, using the first template subsequence to update the compensation guidance parameters of the illumination compensation guidance model, and in the process of updating the compensation guidance parameters, using the second template subsequence to update the illumination guidance strategy of the tunnel distributed illumination compensation guidance information, to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information.

[0052] Continuing with the example of a first template tunnel distributed illumination change data a in the first template subsequence, such as the data of the middle lane of the tunnel at a certain moment, let the tunnel distributed illumination compensation guidance information before optimization be Y0, and the parameter information to be updated in the illumination compensation guidance model be Q0.

[0053] The server updates the parameter information Q0 to be updated in the illumination compensation guidance model based on the tunnel distributed illumination compensation guidance information Y0 and the first template tunnel distributed illumination change data a. For example, according to the illumination intensity, color temperature and other information in the illumination data of the lane and the guidance information Y0, the parameters related to the illumination compensation of the lane in the model (such as the parameters related to the illumination intensity calculation and illumination distribution adjustment) are adjusted to generate iterative learning parameter information Q1.

[0054] Next, according to the iterative learning parameter information Q1, a second template tunnel distributed illumination change data b in the second template subsequence, such as data of the same lane at another time, is used to update the tunnel distributed illumination compensation guidance information Y0. According to the new model parameter Q1 and the illumination information in the data b (such as the illumination change of different lamps), the strategy for lane illumination compensation in the guidance information is adjusted to generate iterative tunnel distributed illumination compensation guidance information Y1.

[0055] The server uses the iterative tunnel distributed illumination compensation guidance information Y1 as new tunnel distributed illumination compensation guidance information, and uses the iterative learning parameter information Q1 as new learning parameter information.

[0056] Then, X pieces of distributed illumination change data of the first template tunnel and Z pieces of distributed illumination change data of the second template tunnel are used for cyclic updating. For example, for each piece of distributed illumination change data of the first template tunnel, the model parameters are updated using the guidance information in the above manner, and then the updated parameters and the distributed illumination change data of the second template tunnel are used to update the guidance information, and this is repeated.

[0057] In this process, as the number of cycles increases, the illumination compensation guidance model and the tunnel distributed illumination compensation guidance information are continuously optimized. When the iterative illumination compensation guidance model meets the training termination requirements, for example, the prediction error of the model on the validation data set reaches the preset minimum value, or the model performance is no longer significantly improved after multiple consecutive cycles, the final iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information are generated. These final models and guidance information can better meet the illumination compensation needs of different lighting areas in the tunnel and improve the overall performance of the tunnel lighting system.

[0058] In a possible implementation, the first template subsequence includes X first template tunnel distributed illumination change data, any one of the first template tunnel distributed illumination change data is represented as first template tunnel distributed illumination change data a, and the first template tunnel distributed illumination change data a is annotated with first template illumination compensation guidance annotation data, X and a are positive integers, 1≤a≤X. Let the tunnel distributed illumination compensation guidance information before optimization be represented as Y0, and the iterative tunnel distributed illumination compensation guidance information be represented as Y1.

[0059] Step S131 includes:

[0060] Step S1311, according to the tunnel distributed illumination compensation guidance information Y0, illumination compensation guidance is performed on the first template tunnel distributed illumination change data a through the illumination compensation guidance model to generate an illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a.

[0061] Step S1312, calculating the model learning cost data of the first template tunnel distributed illumination change data a according to the deviation between the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data.

[0062] Step S1313, based on the determined model learning cost data of the X first template tunnel distributed illumination change data, back-propagation update is performed on the illumination compensation guidance model to generate an iterative illumination compensation guidance model.

[0063] In a possible implementation, step S1312 includes:

[0064] Step S1312 - 1 , obtaining a target model cost function for optimizing the illumination compensation guidance model, wherein the target model cost function includes a first model cost function and a second model cost function.

[0065] Step S1312-2, based on the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data, the first model cost function is used to determine the first model cost parameter of the first template tunnel distributed illumination change data a.

[0066] Step S1312-3, based on the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data, use the second model cost function to determine the second model cost parameter of the first template tunnel distributed illumination change data a.

[0067] Step S1312-4: perform a fusion calculation on the first model cost parameter and the second model cost parameter to generate model learning cost data of the first template tunnel distributed illumination change data a.

[0068] In a possible implementation, the second template subsequence includes Z second template tunnel distributed illumination change data, any one of the second template tunnel distributed illumination change data is represented as second template tunnel distributed illumination change data b, and the second template tunnel distributed illumination change data b is annotated with second template illumination compensation guidance annotation data, Z and b are positive integers, 1≤b≤Z.

[0069] Step S131 further includes:

[0070] Step S1314, calling the iterative illumination compensation guidance model, performing illumination compensation guidance on the second template tunnel distributed illumination change data b, and generating an illumination compensation guidance estimation result of the second template tunnel distributed illumination change data b.

[0071] Step S1315, calculating the model learning cost data of the second template tunnel distributed illumination change data b according to the deviation between the illumination compensation guidance estimation result of the second template tunnel distributed illumination change data b and the second template illumination compensation guidance annotation data.

[0072] Step S1316, based on the determined model learning cost data of the Z second template tunnel distributed illumination change data, the tunnel distributed illumination compensation guidance information Y0 is cyclically updated to generate iterative tunnel distributed illumination compensation guidance information Y1.

[0073] In this embodiment, in the tunnel lighting management system scenario, the first template subsequence contains X first template tunnel distributed illumination change data. Taking the lighting condition of the tunnel entrance area as an example, it is assumed that the first template tunnel distributed illumination change data a in the first template subsequence represents the illumination data of the entrance area at a specific moment. This data includes the illumination intensity of each lighting fixture in the entrance area (such as the lux value of the lamps at different positions), the illumination angle (information such as the angle between the illumination direction of the lamp and the vertical direction), etc. At the same time, this data a is annotated with the first template illumination compensation guidance annotation data. These annotation data are based on the actual traffic flow of the entrance area at that time (the number of vehicles, vehicle models, etc. affect the illumination demand), safety monitoring data (such as whether there are accident potential areas that require special illumination), and manual evaluation (such as the ideal illumination conditions judged by maintenance personnel based on experience). The ideal illumination compensation value is obtained, for example, some areas need to increase or decrease the illumination intensity by a certain percentage.

[0074] The server provides illumination compensation guidance for the first template tunnel distributed illumination change data a through the illumination compensation guidance model based on the tunnel distributed illumination compensation guidance information Y0 before optimization. The first illumination analysis unit in the illumination compensation guidance model first performs a detailed analysis of the illumination data of the entrance area, such as whether the distribution of illumination intensity is uniform, whether the illumination intensity in different areas meets safety standards, etc. Then, the compensation guidance generation unit calculates the illumination compensation guidance estimation result based on the guidance information Y0 (which may include the target intensity range of illumination in the entrance area, illumination uniformity requirements, etc.). For example, based on the current illumination conditions in the entrance area and the guidance information Y0, the model estimates that some lamps need to increase the illumination intensity by 5%, and the illumination angles in some areas need to be adjusted by a certain degree.

[0075] Next, the server obtains the target model cost function for optimizing the illumination compensation guidance model, which includes a first model cost function and a second model cost function. For example, the first model cost function may focus on measuring the difference in illumination intensity between the illumination compensation guidance estimation result and the labeled data, which can be a variation of the mean square error (MSE) function. The second model cost function may focus more on the difference in uniformity of illumination distribution, such as calculating the difference between the standard deviation of illumination intensity and the ideal standard deviation.

[0076] According to the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data, the first model cost function is used to determine the first model cost parameter of the first template tunnel distributed illumination change data a. For example, if the first model cost function is a mean square error function, then the estimation result (such as the estimation that the illumination intensity of the lamp needs to be increased by 5% as mentioned above) and the annotation data (the actual illumination intensity value that needs to be increased) are substituted into the function calculation to obtain the first model cost parameter.

[0077] Similarly, according to the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data, the second model cost function is used to determine the second model cost parameter of the first template tunnel distributed illumination change data a. For example, the second model cost parameter is calculated according to the illumination distribution uniformity estimation result and the ideal uniformity of the annotation.

[0078] The server performs a fusion calculation on the first model cost parameter and the second model cost parameter to generate the model learning cost data of the first template tunnel distributed illumination change data a. For example, the fusion can be performed in a weighted average manner, and different weights can be assigned to the first model cost parameter and the second model cost parameter according to actual needs (for example, if the numerical accuracy of the illumination intensity is considered to be more important, a higher weight can be assigned to the first model cost parameter), and then the model learning cost data is calculated. This model learning cost data reflects the prediction error of the illumination compensation guidance model on the data a, and the higher the value, the greater the error.

[0079] Then, the server performs the above operation on all X first template tunnel distributed illumination change data in the first template subsequence to obtain X model learning cost data. For example, for illumination data at different times of the tunnel entrance area (i.e., different first template tunnel distributed illumination change data), their model learning cost data are calculated respectively.

[0080] Based on the determined model learning cost data of the X first template tunnel distributed illumination change data, the illumination compensation guidance model is back-propagated and updated. In the illumination compensation guidance model (assuming it is a neural network model), the back-propagation algorithm adjusts the parameters within the model (such as the weights and biases of the neural network, etc.) according to the model learning cost data of each data. For example, if the model has a large prediction error for the illumination intensity at a certain data point (the model learning cost data is high), then during the back-propagation process, the weight parameters related to the illumination intensity calculation will be adjusted so that the model can more accurately predict the illumination compensation value in subsequent calculations. After learning and adjusting the parameters of all X data, an iterative illumination compensation guidance model is generated. This iterative model can be more accurate than the original model when dealing with illumination compensation problems in illuminated areas such as tunnel entrance areas.

[0081] Furthermore, in the tunnel lighting management system, the second template subsequence contains Z second template tunnel distributed lighting change data. Taking the lighting condition of the tunnel curve area as an example, it is assumed that the second template tunnel distributed lighting change data b represents the lighting data of the curve area at a specific moment. This data contains information such as the lighting intensity and lighting angle of the curve area, and is annotated with the second template lighting compensation guidance annotation data. These annotation data are ideal lighting compensation values ​​derived based on the actual conditions of the curve area at the time (such as the curvature of the curve, the vehicle speed, safety facilities near the curve, etc.). For example, a higher lighting intensity may be required on the inside of the curve to ensure that the driver can see the curve clearly.

[0082] The server calls the previously generated iterative illumination compensation guidance model to provide illumination compensation guidance for the distributed illumination change data b of the second template tunnel. The first illumination analysis unit in the iterative illumination compensation guidance model analyzes the illumination data of the curve area, for example, analyzing whether the difference in illumination intensity at different positions in the curve area meets the requirements for safe driving. Then, the compensation guidance generation unit calculates the illumination compensation guidance estimation result according to the algorithm of the model. For example, based on the current illumination conditions in the curve area and its own algorithm, the model estimates that the illumination intensity of some lamps on the inside of the curve needs to be increased by 8%, and the illumination angle on the outside of the curve needs to be adjusted by a certain degree, etc.

[0083] The server calculates the model learning cost data of the second template tunnel distributed illumination change data b based on the deviation between the illumination compensation guidance estimation result of the second template tunnel distributed illumination change data b and the second template illumination compensation guidance annotation data. For example, if the model estimates that the illumination intensity of the lamps on the inside of the curve needs to be increased by 8%, and the annotation data shows that it should be increased by 10%, then based on this deviation, the model learning cost data is calculated using a model cost function similar to the previous one (which can be a function adjusted for special circumstances in the curve area). This data reflects the prediction error of the iterative illumination compensation guidance model on the curve area data b.

[0084] Next, the server performs the above operation on all Z second template tunnel distributed illumination change data in the second template subsequence to obtain Z model learning cost data. For example, for illumination data at different times in the tunnel curve area (i.e., different second template tunnel distributed illumination change data), their model learning cost data are calculated respectively.

[0085] Based on the determined model learning cost data of the Z second template tunnel distributed illumination change data, the tunnel distributed illumination compensation guidance information Y0 is cyclically updated. For example, if the prediction error of the model at a data point in a certain curve area is large (the model learning cost data is high), then when updating the guidance information Y0, the part related to the illumination guidance strategy in the curve area will be adjusted. The target range of illumination intensity in the curve area, illumination uniformity requirements, etc. may be adjusted. After learning all Z data and adjusting the guidance information, the iterative tunnel distributed illumination compensation guidance information Y1 is generated. This iterative guidance information can more accurately guide the illumination compensation guidance model to perform illumination compensation on illumination areas such as tunnel curves.

[0086] In a possible implementation, step S132 includes:

[0087] Step S1321, based on the tunnel distributed illumination compensation guidance information Y0, the first template tunnel distributed illumination change data a is used to update the parameter information Q0 to be updated in the illumination compensation guidance model to generate iterative learning parameter information Q1.

[0088] Step S1322: Based on the iterative learning parameter information Q1, the second template tunnel distributed illumination change data b is used to update the tunnel distributed illumination compensation guidance information Y0 to generate iterative tunnel distributed illumination compensation guidance information Y1.

[0089] Step S1323, using the iterative tunnel distributed illumination compensation guidance information Y1 as new tunnel distributed illumination compensation guidance information, and using the iterative learning parameter information Q1 as new learning parameter information, using X first template tunnel distributed illumination change data and Z second template tunnel distributed illumination change data for cyclic update. When the iterative illumination compensation guidance model meets the training termination requirements, an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information are generated.

[0090] In this embodiment,

[0091] In the scenario of tunnel lighting management system, there is the tunnel distributed illumination compensation guidance information Y0 before optimization, which includes various guidance strategies for illumination compensation in different areas of the tunnel, such as the target range of illumination intensity in the entrance area, the requirements for illumination uniformity in the curve area, etc. At the same time, the illumination compensation guidance model has its initial parameter information Q0 to be updated, which may be related to functions such as illumination intensity calculation and illumination distribution adjustment, such as parameters such as weights and biases in neural networks.

[0092] The first template subsequence contains multiple first template tunnel distributed illumination change data. Take one of the first template tunnel distributed illumination change data a as an example, assuming that this data a is the illumination data of the middle lane of the tunnel at a certain moment. This data contains information such as the illumination intensity and illumination angle of each lighting fixture in the lane, and is annotated with the first template illumination compensation guidance annotation data. These annotation data are ideal illumination compensation values ​​derived from the actual traffic flow, vehicle type distribution, and safety monitoring data of the lane at that time.

[0093] The server updates the parameter information Q0 to be updated in the illumination compensation guidance model based on the tunnel distributed illumination compensation guidance information Y0 and the first template tunnel distributed illumination change data a. For example, the guidance information about the lane in Y0 may include the target illumination intensity distribution pattern of the lane (such as slightly higher illumination intensity in the middle area and slightly lower illumination intensity on both sides). The server combines the actual illumination intensity, illumination angle and other information in the data a with the guidance information in Y0 to adjust the parameters related to illumination compensation of the lane in the model. If the data a shows that the illumination intensity of a certain lamp in the middle lane is lower than the target value in Y0, and the relevant weights and biases in the model parameters Q0 cause the model to predict the illumination intensity of the lamp inaccurately under the current situation, then these parameters will be adjusted. By processing the data a in this way, iterative learning parameter information Q1 is generated. This Q1 is an optimization of the original model parameters Q0 based on the specific illumination data a and the guidance information Y0, so that the model can better adapt to the illumination compensation requirements of the lane.

[0094] Now we have iterative learning parameter information Q1, which is an optimization result of the illumination compensation guidance model parameters. The second template subsequence contains multiple second template tunnel distributed illumination change data. Take one of the second template tunnel distributed illumination change data b as an example. Assume that this data b is the illumination data of the same lane at another time. It is annotated with the second template illumination compensation guidance annotation data. These annotation data are ideal illumination compensation values ​​obtained based on the actual traffic conditions of the lane at that time (such as changes in vehicle speed, differences in traffic flow in different time periods, etc.).

[0095] Next, the server updates the tunnel distributed illumination compensation guidance information Y0 based on the iterative learning parameter information Q1 and the second template tunnel distributed illumination change data b. Since Q1 has optimized the illumination compensation guidance model once, the model will have different performances when analyzing data b. For example, the model adjusted by Q1 can more accurately evaluate the illumination conditions of the lane when analyzing information such as illumination intensity and illumination angle in data b. According to the actual illumination conditions in data b and the illumination compensation guidance estimation results made by the model based on Q1 (this result has a certain relationship with the labeled data), the part of Y0 related to the illumination guidance strategy for the lane is adjusted. If the model finds that the target range of illumination intensity for the lane specified in Y0 is not suitable under certain traffic conditions based on the analysis of data b by Q1, then the target range and other contents will be adjusted to generate iterative tunnel distributed illumination compensation guidance information Y1. This Y1 is an optimization of the original guidance information Y0 based on the optimized model parameters Q1 and the new illumination data b, which can more accurately guide the illumination compensation guidance model to the illumination compensation process of the lane.

[0096] Further, the server uses the iterative tunnel distributed illumination compensation guidance information Y1 as new tunnel distributed illumination compensation guidance information, and uses the iterative learning parameter information Q1 as new learning parameter information. Then, the X first template tunnel distributed illumination change data and the Z second template tunnel distributed illumination change data are used for cyclic update.

[0097] For each first template tunnel distributed illumination change data (taking the data of different areas such as the tunnel entrance area and the curve area as an example), according to the previous method, based on the new guidance information Y1, use these data to update the current model parameters (i.e., Q1) to obtain new iterative learning parameter information. Then, for each second template tunnel distributed illumination change data (also covering different tunnel areas), based on the new iterative learning parameter information, use these data to update the current guidance information (i.e., Y1) to obtain new iterative tunnel distributed illumination compensation guidance information. This process is repeated continuously.

[0098] During the cyclic update process, the server will continuously check whether the iterative illumination compensation guidance model meets the training termination requirements. For example, the training termination requirement may be that the prediction error of the model on the validation data set (this data set contains some tunnel lighting data, which is used to test the generalization ability of the model) reaches a pre-set minimum value. When the prediction error of the model on the validation data set no longer decreases significantly, or has reached the set minimum value, the model is considered to have converged and meets the training termination requirements.

[0099] At this point, the server stops the cyclic update process and generates an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information. This iterative illumination compensation guidance model can more accurately provide illumination compensation guidance for the illumination areas in the illumination change data of various areas in the tunnel (such as the entrance area, driving lane, curve area, etc.), and the iterative tunnel distributed illumination compensation guidance information can more effectively guide the work of the illumination compensation guidance model, thereby improving the illumination compensation effect and management efficiency of the entire tunnel lighting system.

[0100] In a possible implementation, the illumination compensation guidance model includes: a first illumination analysis unit, a second illumination analysis unit, and a compensation guidance generation unit. The compensation guidance generation unit is composed of: an illumination fine-tuning subunit, a deconvolution subunit, and a multimodal mapping subunit. The illumination fine-tuning subunit refers to an illumination processing mechanism combining a multi-source fusion model in an adaptive adjustment unit. The illumination fine-tuning subunit provides illumination compensation prior knowledge for the illumination compensation guidance model based on the illumination processing mechanism. The method also includes:

[0101] Step A110: Obtain a first function layer to be updated from the illumination fine-tuning subunit.

[0102] Step A120: Obtain the second function layer to be updated from the deconvolution subunit.

[0103] Step A130: Obtain the third function layer to be updated from the multimodal mapping subunit.

[0104] Step A140: output the first function layer, the second function layer, and the third function layer to be updated as parameter information to be updated in the illumination compensation guidance model.

[0105] In this embodiment, in the tunnel lighting management system scenario, the illumination compensation guidance model is responsible for analyzing the illumination conditions in the tunnel and providing illumination compensation guidance. The first illumination analysis unit mainly performs a preliminary analysis on the input illumination data (such as tunnel distributed illumination change data). For example, for the illumination data of the tunnel entrance area, the first illumination analysis unit will analyze the distribution of illumination intensity, illumination angle, and illumination coordination between different lamps.

[0106] In addition to participating in generating the tunnel distributed illumination compensation guidance information, the second illumination analysis unit may also perform other feature extraction operations on the illumination data. For example, it can analyze the color temperature information in the illumination data, as well as the changing trend of illumination in different time periods.

[0107] The compensation guidance generation unit generates illumination compensation guidance based on the analysis results of the previous two units. The illumination fine-tuning subunit in this unit combines the illumination processing mechanism of the multi-source fusion model in the adaptive adjustment unit to provide illumination compensation prior knowledge for the entire illumination compensation guidance model. For example, in the illumination compensation of the tunnel curve area, the illumination fine-tuning subunit knows that the inner side of the curve usually requires a higher illumination intensity based on the illumination processing mechanism of the multi-source fusion model (which may integrate multiple aspects of information such as the tunnel's structural information and vehicle traffic patterns), and it will provide this prior knowledge to the entire model.

[0108] As the lighting conditions in the tunnel change and more lighting data is accumulated, some function layers in the light fine-tuning subunit need to be updated. For example, there may be a function layer in this subunit that specifically handles the relationship between light intensity and vehicle speed (different vehicle speeds in the tunnel may require different light intensity adjustment strategies). As the types of vehicles and traffic flow patterns change, the parameters of this function layer may no longer be applicable. The server obtains this function layer to be updated from the light fine-tuning subunit, which contains the parameters and calculation logic related to the light intensity-vehicle speed relationship.

[0109] The deconvolution subunit may be responsible for converting a low-resolution illumination feature map into a high-resolution illumination compensation guidance map in the illumination compensation guidance model. For example, when analyzing a large area in a tunnel (such as the illumination of an entire lane), a low-resolution illumination feature representation may be obtained first, and then converted into a high-resolution, more detailed illumination compensation guidance by the deconvolution subunit. If the lighting layout of the tunnel changes (such as increasing or decreasing the number of lamps), the function layer related to the feature map resolution conversion in the deconvolution subunit needs to be updated. The server obtains this function layer to be updated from the deconvolution subunit, which contains parameters related to resolution conversion, such as convolution kernel size, step size, etc.

[0110] The multimodal mapping subunit is responsible for mapping information of different modalities (such as light intensity, color temperature, light angle, etc.) to a unified illumination compensation guidance space. For example, in tunnel lighting, different lamps may have different color temperatures. The multimodal mapping subunit combines the color temperature information with information such as light intensity and angle to generate comprehensive illumination compensation guidance. If new illumination modal information is introduced (such as special illumination characteristics brought by new lighting technology), the function layer that processes modal mapping in the multimodal mapping subunit needs to be updated. The server obtains the function layer to be updated from the multimodal mapping subunit, which contains parameters related to modal mapping, such as the allocation parameters of different modal weights.

[0111] The server combines the first function layer obtained from the illumination fine-tuning subunit, the second function layer obtained from the deconvolution subunit, and the third function layer obtained from the multimodal mapping subunit, and outputs the parameter information to be updated in the illumination compensation guidance model. These parameter information will be used in the subsequent model update process to adapt to changes in the tunnel lighting system.

[0112] In a possible implementation manner, the tunnel distributed illumination compensation guidance information includes a guidance coding representation. Step S120 includes:

[0113] Step S121, obtaining basic guidance data loaded into the illumination compensation guidance model, wherein the basic guidance data includes at least one of a text description, an area frame, a cursor point set, and a light intensity distribution map.

[0114] Step S122: feature-encoding the basic guidance data through the second illumination analysis unit in the illumination compensation guidance model to generate a guidance encoding representation of the illumination compensation guidance model.

[0115] In the tunnel lighting management system, the server first obtains the basic guidance data loaded into the illumination compensation guidance model. For example, the text description may be a document from the tunnel lighting design specification, which clearly stipulates the lower and upper limits of the illumination intensity and the requirements for illumination uniformity in different areas of the tunnel (such as the entrance area, the driving lane, the curve area, the exit area, etc.). The area framing data may be the coordinate information that divides the tunnel into different areas in space, such as the coordinate range of the driving lane area, which helps the model to clarify the illumination compensation range of different areas. The cursor point set can be the coordinate information of the illumination monitoring points pre-set at key positions in the tunnel (such as the tangent point of the curve, the middle position of the tunnel, etc.). The illumination data of these monitoring points is very important for evaluating the illumination conditions of the entire tunnel. The light intensity distribution diagram is a graphical description of the ideal tunnel illumination intensity distribution obtained based on historical data or design standards. It shows the illumination intensity that should be achieved in each area of ​​the tunnel.

[0116] The server performs feature encoding on these basic guidance data through the second illumination analysis unit in the illumination compensation guidance model to generate a guidance coding representation of the illumination compensation guidance model. For example, for the illumination intensity requirements in the text description, the second illumination analysis unit may convert it into a numerical feature vector. For the area framing data, it may convert the coordinate information into a coding form that matches the input data format of the illumination compensation model. For the cursor point set, its coordinate information and related illumination requirements are encoded into a specific vector. For the light intensity distribution map, the key features are extracted and encoded into a representation that the model can understand. These encoded information are combined to form a guidance coding representation of the tunnel distributed illumination compensation guidance information, which will play a guiding role in the operation of the illumination compensation guidance model to ensure that the illumination compensation meets the actual needs and safety standards of tunnel lighting.

[0117] In a possible implementation, the method further includes:

[0118] Step B110, obtaining target distributed illumination data to be analyzed in an application scenario, wherein the application scenario includes at least one of a tunnel monitoring scenario, a night driving scenario, a lighting experiment scenario, and a light show scenario.

[0119] Step B120, based on the iterative tunnel distributed illumination compensation guidance information, illumination compensation guidance is performed on the target distributed illumination data through the iterative illumination compensation guidance model to generate an illumination compensation result of the target distributed illumination data.

[0120] Step B130, based on the illumination compensation result of the target distributed illumination data, continue to perform illumination analysis operation on the target distributed illumination data in the application scenario. The illumination analysis operation includes: any one of tunnel illumination safety analysis, night driving sight optimization analysis, illumination experiment effect evaluation, and light show effect presentation.

[0121] In a possible implementation, the illumination compensation guidance model includes: a first illumination analysis unit, a second illumination analysis unit, and a compensation guidance generation unit. The illumination compensation result includes an illumination compensation parameter array, the illumination compensation parameter array includes an illumination compensation parameter of each illumination node in the target distributed illumination data, and each illumination compensation parameter represents an illumination adjustment value corresponding to the corresponding illumination node.

[0122] Step B120 includes:

[0123] Step B121, performing illumination analysis on the target distributed illumination data by the first illumination analysis unit to generate an illumination feature vector of the target distributed illumination data, wherein the illumination analysis includes: any one or more of illumination intensity distribution analysis, color temperature distribution analysis, and illumination uniformity analysis.

[0124] Step B122: The compensation guidance generating unit generates compensation guidance for the illumination feature vector and the iterative tunnel distributed illumination compensation guidance information to generate an illumination compensation parameter array of the target distributed illumination data.

[0125] In this embodiment, in the tunnel monitoring scenario, the server obtains target distributed lighting data from a network of numerous lighting sensors installed in the tunnel. These sensors are distributed in different locations of the tunnel, such as the entrance area, the driving lane, the curve area, the exit area, etc. The sensor collects lighting data at regular intervals (for example, every second or every few minutes), including information such as light intensity (in lux) and color temperature (in Kelvin). These data are transmitted to the server to form target distributed lighting data. For example, the sensor in the entrance area may collect a light intensity of 500 lux and a color temperature of 4000K; the sensor in the middle part of the driving lane may collect a light intensity of 300 lux and a color temperature of 3500K, etc. These data reflect the actual light distribution in the current tunnel.

[0126] For night driving scenarios, the server may obtain target distributed lighting data from the lighting monitoring equipment installed on the vehicle and the lighting facility monitoring system along the road. The equipment on the vehicle can monitor the lighting conditions of the road ahead of the vehicle, including the light intensity and color temperature at different distances. The lighting facility monitoring system along the road provides lighting parameters of lighting equipment such as street lamps. For example, the light intensity 10 meters in front of the vehicle is 100 lux and the color temperature is 3000K, while the light intensity of roadside lamps varies at different locations, which may be 200 lux near the street lamps and 50 lux far away from the street lamps. These data are combined to form the target distributed lighting data in the night driving scenario, which is used for subsequent illumination compensation guidance and analysis.

[0127] In the lighting experiment scenario, the server obtains the target distributed lighting data from the experimental equipment. For example, in an indoor lighting experiment, the experimental area is divided into multiple small areas, each of which has dedicated lighting adjustment equipment and sensors. The server obtains data such as light intensity and color temperature in each area from these sensors. Assume that the experimental area is divided into 9 small areas, the light intensity of area 1 is 800 lux and the color temperature is 5000K; the light intensity of area 2 is 600 lux and the color temperature is 4500K, etc. These detailed lighting data constitute the target distributed lighting data in the lighting experiment scenario, so as to evaluate and optimize the lighting conditions in the experiment.

[0128] In the light show scene, the server obtains the target distributed lighting data from each lighting device in the light show control system. The lighting devices in the light show have different colors, brightness, and flashing modes. The server obtains the current lighting parameters of each lighting device, such as the light intensity of a certain color spotlight is 1000 lux, the color temperature corresponding to the color is 6000K, and its flashing frequency is 2 times per second. These data from many lighting devices are integrated together to form the target distributed lighting data in the light show scene, which is used to ensure that the light show effect meets the expectations.

[0129] Furthermore, in a tunnel monitoring scenario, the server performs illumination analysis on target distributed illumination data through the first illumination analysis unit in the illumination compensation guidance model.

[0130] The first light analysis unit calculates the difference in light intensity in different areas of the tunnel. For example, at the entrance of the tunnel, the light intensity collected by the sensor is 500 lux, while the light intensity in a certain area deep in the tunnel is 300 lux. The first light analysis unit will analyze whether this intensity difference meets the requirements of safety and comfort. If the difference is too large, it may affect the driver's vision adaptation and light intensity compensation is required.

[0131] At the same time, the first lighting analysis unit will also analyze the distribution of color temperature. For example, the color temperature of the entrance area is 4000K, while the color temperature of the lane is 3500K. Whether the transition between areas with different color temperatures is smooth will also affect the visual effect and safety. If the color temperature transition is unnatural, it may be necessary to adjust the lighting parameters of certain areas.

[0132] For the illumination uniformity of the entire tunnel, the first illumination analysis unit will calculate statistical indicators such as the standard deviation of illumination intensity in different areas. If the standard deviation is too large, it means that the illumination is uneven, and some areas may be too bright or too dark, which also requires illumination compensation adjustment. Through these analyses, the first illumination analysis unit generates an illumination feature vector of the target distributed illumination data, which contains a quantitative representation of information such as illumination intensity distribution, color temperature distribution, and illumination uniformity.

[0133] In the night driving scenario, the first lighting analysis unit will also perform a similar analysis on the lighting data of the road ahead of the vehicle. For example, the changing trend of the light intensity at different distances in front of the vehicle and the impact of color temperature on the driver's line of sight are analyzed to generate the corresponding lighting feature vector. In the lighting experiment scenario, the difference in light intensity and color temperature between different small areas, as well as the overall illumination uniformity, are analyzed to generate a lighting feature vector that meets the experimental requirements. In the light show scenario, the coordination between the light intensity, color temperature and flashing mode of different lighting equipment is analyzed to generate the lighting feature vector in the light show scenario.

[0134] In the tunnel monitoring scenario, the server generates compensation guidance for the illumination feature vector and iterative tunnel distributed illumination compensation guidance information through the compensation guidance generation unit. The iterative tunnel distributed illumination compensation guidance information contains information such as the ideal illumination parameter requirements for different areas of the tunnel. For example, the guidance information stipulates that the illumination intensity in the entrance area should be between 450 and 550 lux, and the standard deviation of illumination uniformity should be less than a certain value. The compensation guidance generation unit calculates the illumination compensation parameters of each illumination node (i.e., the area corresponding to each sensor) based on the actual illumination conditions in the illumination feature vector (such as the current illumination intensity in the entrance area is 500 lux, and the standard deviation of illumination uniformity is large) and the guidance information. Assuming that the illumination node corresponding to a sensor in the entrance area needs to increase the illumination intensity by 50 lux according to the calculation, then this illumination compensation parameter is + 50 lux. By performing such calculations on all illumination nodes in the tunnel, an illumination compensation parameter array of the target distributed illumination data is generated.

[0135] In the night driving scenario, the compensation guidance generation unit calculates the illumination compensation parameters at different positions on the road ahead of the vehicle based on the special requirements of night driving (such as ensuring the driver's clear vision and avoiding glare, etc.) and the illumination feature vector. For example, for the area with low illumination intensity 10 meters in front of the vehicle, it is calculated that a certain amount of illumination intensity needs to be increased to optimize the driver's vision. In the illumination experiment scenario, according to the expected goals of the experiment (such as achieving a specific illumination intensity and color temperature distribution pattern) and the illumination feature vector, the illumination compensation parameters of each small area are calculated to adjust the illumination conditions of the experimental area to meet the experimental requirements. In the light show scenario, according to the creative design requirements of the light show (such as brightness and color coordination between different lighting devices) and the illumination feature vector, the illumination compensation parameters of each lighting device are calculated, such as adjusting the illumination intensity or flashing frequency of a certain color spotlight to achieve a better light show effect.

[0136] The server performs tunnel lighting safety analysis in the tunnel monitoring scenario based on the lighting compensation results of the target distributed lighting data. For example, if the lighting compensation results show that the lighting intensity in certain curved areas needs to be increased, this is because the original lighting intensity may cause the driver's vision to be obstructed when turning, and increasing the lighting intensity can improve safety. The server will further analyze the effect of increasing the lighting intensity, such as by calculating the driver's visible distance, viewing angle and other parameters under the new lighting intensity, to evaluate whether the tunnel safety standards are met. At the same time, the impact of lighting compensation on other aspects will also be considered, such as whether glare and other problems will occur. If the visible distance meets the safety standard requirements after increasing the light intensity and no adverse effects such as glare are generated, then it can be considered that the lighting compensation is effective in terms of safety.

[0137] In the night driving scenario, the server performs night driving vision optimization analysis based on the illumination compensation results. For example, the illumination compensation parameter array shows that the illumination intensity and color temperature of a certain section of road ahead of the vehicle need to be adjusted. The server analyzes the impact of the adjusted illumination conditions on the driver's vision. If the adjusted illumination intensity enables the driver to see obstacles on the road more clearly, and the adjustment of the color temperature makes the road signs more eye-catching, then the illumination compensation effectively optimizes the night driving vision. At the same time, the server also considers the effect of illumination compensation under different weather conditions (such as rainy days and foggy days) to ensure that the safety and comfort of night driving can be improved in all situations.

[0138] In the lighting experiment scenario, the server evaluates the lighting experiment effect based on the lighting compensation results. If the lighting compensation parameter array adjusts the light intensity and color temperature of each small area, the server will analyze whether these adjustments meet the expected goals of the experiment. For example, the purpose of the experiment is to study the effects of different combinations of light intensity and color temperature on plant growth. The server will observe whether the growth indicators of the plants (such as leaf color, growth rate, etc.) change in the expected direction based on the actual lighting conditions after lighting compensation. If the growth indicators of the plants are improved under the adjusted lighting environment, it means that the lighting compensation has played a positive role in the experimental results. Otherwise, the lighting parameters need to be further adjusted.

[0139] In the light show scenario, the server analyzes the light show effect presentation based on the light compensation results. For example, the light compensation parameters adjust the light intensity, color, and flickering frequency of certain lighting equipment. The server evaluates the overall effect of the light show after these adjustments, such as the coordination of colors, the rhythm of light flickering, and the coordination with the music or performance. If the adjusted light show is more visually attractive, the color transition is more natural, and it matches the rhythm of the music or performance better, then it means that the light compensation has successfully improved the effect of the light show.

[0140] Figure 2The hardware structure of the tunnel distributed illumination compensation processing system 100 provided in the embodiment of the present invention for implementing the above-mentioned tunnel distributed illumination compensation processing method is shown as follows: Figure 2 As shown, the tunnel distributed illumination compensation processing system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0141] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the tunnel distributed light illumination compensation processing system 100 uses to execute or use to complete the exemplary method described in the present invention.

[0142] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the tunnel distributed light illumination compensation processing method of the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0143] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned tunnel distributed illumination compensation processing system 100. The implementation principles and technical effects are similar and will not be repeated in this embodiment.

[0144] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above tunnel distributed illumination compensation processing method is implemented.

[0145] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A tunnel distributed illumination compensation processing method, characterized in that: The method comprises: Acquire a template tunnel distributed illumination change data sequence for learning parameter information of an illumination compensation guidance model; wherein the template tunnel distributed illumination change data sequence includes a plurality of template tunnel distributed illumination change data, the template tunnel distributed illumination change data includes illumination areas, and the illumination compensation guidance model is used to provide illumination compensation guidance for the illumination areas in the template tunnel distributed illumination change data; Acquiring tunnel distributed illumination compensation guidance information; the tunnel distributed illumination compensation guidance information is used to guide the illumination compensation guidance process executed by the illumination compensation guidance model; The template tunnel distributed illumination change data sequence is used to perform linkage optimization on the illumination compensation guidance model and the tunnel distributed illumination compensation guidance information to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information; wherein the iterative illumination compensation guidance model is used to provide illumination compensation guidance for an illumination area in any tunnel distributed illumination change data according to the guidance of the iterative tunnel distributed illumination compensation guidance information; The template tunnel distributed illumination change data sequence is divided into a first template subsequence and a second template subsequence; the template tunnel distributed illumination change data sequence is used to perform linkage optimization on the illumination compensation guidance model and the tunnel distributed illumination compensation guidance information to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information, including: Using the first template subsequence to update the compensation guidance parameters of the illumination compensation guidance model to generate an iterative illumination compensation guidance model, and using the second template subsequence to update the illumination guidance strategy of the tunnel distributed illumination compensation guidance information to generate iterative tunnel distributed illumination compensation guidance information; or, The first template subsequence is used to update the compensation guidance parameters of the illumination compensation guidance model, and in the process of updating the compensation guidance parameters, the second template subsequence is used to update the illumination guidance strategy of the tunnel distributed illumination compensation guidance information to generate an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information; The first template subsequence includes X first template tunnel distributed illumination change data, any one of the first template tunnel distributed illumination change data is represented as first template tunnel distributed illumination change data a, and the first template tunnel distributed illumination change data a is annotated with first template illumination compensation guidance annotation data, X and a are positive integers, 1≤a≤X; the tunnel distributed illumination compensation guidance information before optimization is represented as Y0, and the iterative tunnel distributed illumination compensation guidance information is represented as Y1; The step of updating the compensation guidance parameters of the illumination compensation guidance model by using the first template subsequence to generate an iterative illumination compensation guidance model includes: According to the tunnel distributed illumination compensation guidance information Y0, the illumination compensation guidance model is used to perform illumination compensation guidance on the first template tunnel distributed illumination change data a, and an illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a is generated; Calculating the model learning cost data of the first template tunnel distributed illumination change data a according to the deviation between the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data; According to the determined model learning cost data of the X first template tunnel distributed illumination change data, the illumination compensation guidance model is back-propagated and updated to generate an iterative illumination compensation guidance model.

2. The tunnel distributed illumination compensation processing method according to claim 1, characterized in that: The calculating the model learning cost data of the first template tunnel distributed illumination change data a according to the deviation between the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data includes: Acquire a target model cost function for optimizing the illumination compensation guidance model, wherein the target model cost function includes a first model cost function and a second model cost function; Determine a first model cost parameter of the first template tunnel distributed illumination change data a using the first model cost function according to the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data; Determine a second model cost parameter of the first template tunnel distributed illumination change data a using the second model cost function according to the illumination compensation guidance estimation result of the first template tunnel distributed illumination change data a and the first template illumination compensation guidance annotation data; The first model cost parameter and the second model cost parameter are fused and calculated to generate model learning cost data of the first template tunnel distributed illumination change data a.

3. The tunnel distributed illumination compensation processing method according to claim 2, characterized in that: The second template subsequence includes Z second template tunnel distributed illumination change data, any one of the second template tunnel distributed illumination change data is represented as second template tunnel distributed illumination change data b, and the second template tunnel distributed illumination change data b is annotated with second template illumination compensation guidance annotation data, Z and b are positive integers, 1≤b≤Z; The step of updating the tunnel distributed illumination compensation guidance information with the second template subsequence to generate iterative tunnel distributed illumination compensation guidance information includes: Calling the iterative illumination compensation guidance model to perform illumination compensation guidance on the second template tunnel distributed illumination change data b, and generating an illumination compensation guidance estimation result of the second template tunnel distributed illumination change data b; Calculate the model learning cost data of the second template tunnel distributed illumination change data b according to the deviation between the illumination compensation guidance estimation result of the second template tunnel distributed illumination change data b and the second template illumination compensation guidance annotation data; According to the determined model learning cost data of the Z second template tunnel distributed illumination change data, the tunnel distributed illumination compensation guidance information Y0 is cyclically updated to generate iterative tunnel distributed illumination compensation guidance information Y1.

4. The tunnel distributed illumination compensation processing method according to claim 1, characterized in that: The method of updating the compensation guidance parameters of the illumination compensation guidance model by using the first template subsequence, and updating the illumination guidance strategy of the tunnel distributed illumination compensation guidance information by using the second template subsequence during the updating of the compensation guidance parameters, and generating an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information, includes: According to the tunnel distributed illumination compensation guidance information Y0, the first template tunnel distributed illumination change data a is used to update the parameter information Q0 to be updated in the illumination compensation guidance model to generate iterative learning parameter information Q1; According to the iterative learning parameter information Q1, the second template tunnel distributed illumination change data b is used to update the tunnel distributed illumination compensation guidance information Y0 to generate iterative tunnel distributed illumination compensation guidance information Y1; The iterative tunnel distributed illumination compensation guidance information Y1 is used as new tunnel distributed illumination compensation guidance information, and the iterative learning parameter information Q1 is used as new learning parameter information, and X first template tunnel distributed illumination change data and Z second template tunnel distributed illumination change data are used for cyclic update; until the iterative illumination compensation guidance model meets the training termination requirements, an iterative illumination compensation guidance model and iterative tunnel distributed illumination compensation guidance information are generated.

5. The tunnel distributed illumination compensation processing method according to any one of claims 1 to 4, characterized in that: The illumination compensation guidance model includes: a first illumination analysis unit, a second illumination analysis unit and a compensation guidance generation unit; wherein the compensation guidance generation unit is composed of: an illumination fine-tuning subunit, a deconvolution subunit, and a multimodal mapping subunit, the illumination fine-tuning subunit refers to an illumination processing mechanism combining a multi-source fusion model in an adaptive adjustment unit, and the illumination fine-tuning subunit provides illumination compensation prior knowledge for the illumination compensation guidance model according to the illumination processing mechanism; the method also includes: Obtaining a first function layer to be updated from the illumination fine-tuning subunit; Obtaining a second function layer to be updated from the deconvolution subunit; Obtaining a third function layer to be updated from the multimodal mapping subunit; The first function layer, the second function layer, and the third function layer to be updated are output as parameter information to be updated in the illumination compensation guidance model.

6. The tunnel distributed illumination compensation processing method according to claim 5, characterized in that: The tunnel distributed illumination compensation guidance information includes a guidance code representation; and obtaining the tunnel distributed illumination compensation guidance information includes: Obtaining basic guidance data loaded into the illumination compensation guidance model; wherein the basic guidance data includes at least one of a text description, an area frame, a cursor point set, and a light intensity distribution map; The basic guidance data is feature-encoded by a second illumination analysis unit in the illumination compensation guidance model to generate a guidance coding representation of the illumination compensation guidance model.

7. The tunnel distributed illumination compensation processing method according to claim 1, characterized in that: The method further comprises: Obtaining target distributed illumination data to be analyzed in an application scenario, wherein the application scenario includes at least one of a tunnel monitoring scenario, a night driving scenario, a lighting experiment scenario, and a light show scenario; According to the iterative tunnel distributed illumination compensation guidance information, illumination compensation guidance is performed on the target distributed illumination data through the iterative illumination compensation guidance model to generate an illumination compensation result for the target distributed illumination data; According to the illumination compensation result of the target distributed illumination data, continue to perform illumination analysis operation on the target distributed illumination data in the application scenario; wherein the illumination analysis operation includes: any one of tunnel illumination safety analysis, night driving sight optimization analysis, illumination experiment effect evaluation, and light show effect presentation; The illumination compensation guidance model comprises: a first illumination analysis unit, a second illumination analysis unit and a compensation guidance generation unit; the illumination compensation result comprises an illumination compensation parameter array, the illumination compensation parameter array comprises an illumination compensation parameter of each illumination node in the target distributed illumination data, and each illumination compensation parameter represents an illumination adjustment value corresponding to the corresponding illumination node; The step of performing illumination compensation guidance on the target distributed illumination data through the iterative illumination compensation guidance model according to the iterative tunnel distributed illumination compensation guidance information to generate an illumination compensation result of the target distributed illumination data includes: Performing illumination analysis on the target distributed illumination data through the first illumination analysis unit to generate an illumination feature vector of the target distributed illumination data; wherein the illumination analysis includes: any one or more of illumination intensity distribution analysis, color temperature distribution analysis, and illumination uniformity analysis; The compensation guidance generating unit performs compensation guidance generation on the illumination feature vector and the iterative tunnel distributed illumination compensation guidance information to generate an illumination compensation parameter array of the target distributed illumination data.

8. A tunnel distributed illumination compensation processing system, characterized in that: The tunnel distributed light illumination compensation processing system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the tunnel distributed light illumination compensation processing method described in any one of claims 1 to 7 above.

Images