Locomotive headlamp dynamic illumination detection method based on multi-sensor fusion

Abstract

The invention belongs to the technical field of locomotive illumination detection, and provides a locomotive headlamp dynamic illumination detection method based on multi-sensor fusion, which comprises the following steps: carrying out multiple transverse uniformity simulation tests on the illumination of each headlamp of a locomotive in a simulated nighttime rainfall environment to obtain a simulated transverse uniformity value; the difference degree between test data after each simulation test is reflected, if a transverse illumination uniform signal is received, an illumination region in multiple illumination simulation tests is analyzed to obtain a sub-region illumination value, and the difference degree between the test data after each simulation test is reflected in a rainfall environment at night of the locomotive headlamp; the difference degree between the reflected illuminance of the ponding area and the reflected illuminance of the non-ponding area in the sub-area is determined, the left illuminance sub-area and the right illuminance sub-area are analyzed respectively, different ponding distribution conditions can be adapted, the influence of the ponding reflected illuminance on the illuminance of the locomotive headlamp can be accurately analyzed, and the illuminance difference condition possibly causing the light curtain effect can be found in advance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of locomotive lighting detection, and particularly relates to a dynamic illumination detection method for locomotive headlamps based on multi-sensor fusion. BACKGROUND

[0002] In the scenario of locomotive driving at night, the illumination performance of the headlamps is directly related to driving safety, especially in the rain environment, the light curtain effect caused by the ground water becomes an important factor threatening driving safety.

[0003] In the prior art, the transverse illumination data is not analyzed in depth, the reflection illumination difference between the water accumulation area and the non-water accumulation area cannot be identified through sub-area division, and the difference cannot be processed according to the asymmetry of the water accumulation distribution of the left and right areas, so that the influence of water reflection on the illumination is difficult to be accurately evaluated, and the early warning mechanism of the light curtain effect is missing. Secondly, the identification and analysis ability of the headlamp illumination mutation sub-area is lacking, the probability of the light curtain effect in different areas cannot be quantified, and the lighting system parameter adjustment lacks pertinence. At the same time, the existing scheme does not combine the light curtain effect sub-area with the locomotive driving state, and the expected occurrence time of the light curtain effect cannot be dynamically calculated, so that the illumination adjustment measures are often lagging behind the risk changes. In addition, in the illumination adjustment process, the traditional method lacks a dynamic gradient adjustment strategy based on the time dimension, and it is difficult to real-time respond to the illumination difference caused by water reflection, so that the problem that the driver's field of vision is disturbed by the light curtain effect cannot be effectively solved.

[0004] Therefore, the application provides a dynamic illumination detection method for locomotive headlamps based on multi-sensor fusion. SUMMARY

[0005] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background art.

[0006] The technical scheme adopted by the application to solve the technical problems is: A dynamic illumination detection method for locomotive headlamps based on multi-sensor fusion, comprising: Under the simulation of the night rain environment, the illumination of each headlamp of the locomotive is tested multiple times in a transverse uniform manner, and whether the illumination is transverse uniform is evaluated; If a transverse illumination uniformity signal is received, the illumination regions in the multiple illumination simulation tests are analyzed, and the illumination mutation sub-area is screened out; The illumination mutation sub-area is analyzed, the probability of the light curtain effect in the illumination mutation sub-area is evaluated, and the light curtain effect sub-area is screened out; The light curtain efficiency sub-region distance and the locomotive running speed are acquired for the light curtain effect sub-region, the light curtain effect predicted time is obtained, the illumination dynamic gradient adjustment value is acquired within the light curtain effect predicted time, and the light curtain effect occurrence probability is reduced.

[0007] As a further scheme of the present application, the illumination of each headlamp of the locomotive is subjected to multiple lateral uniformity simulation tests under simulated night rainfall environment, and whether the illumination is laterally uniform is evaluated. If the lateral illumination uniformity signal is received, the illumination regions in the multiple illumination simulation tests are analyzed, and illumination mutation sub-regions are screened out. The illumination mutation sub-regions are analyzed, the light curtain effect occurrence probability degree in the illumination mutation sub-regions is evaluated, and the light curtain effect sub-regions are screened out. The light curtain efficiency sub-region distance and the locomotive running speed are acquired for the light curtain effect sub-region, the light curtain effect predicted time is obtained, the illumination dynamic gradient adjustment value is acquired within the light curtain effect predicted time, and the light curtain effect occurrence probability is reduced.

[0008] As a further scheme of the present application, the unit lateral uniformity value is acquired in each lateral uniformity simulation test of the illumination of the headlamp of the locomotive, and the process is as follows: The left illumination region is equally divided into a plurality of left illumination sub-regions, and the right illumination region is equally divided into a plurality of right illumination sub-regions. The lamp illumination of each left illumination sub-region and the lamp illumination of each right illumination sub-region are acquired, and left illumination sequence and right illumination sequence are constructed. The same ordered elements are combined by being extracted from the left illumination sequence and the right illumination sequence, a plurality of left-right illumination analysis groups are obtained, and are input into the Euclidean distance formula, and the unit lateral uniformity value is output.

[0009] As a further scheme of the present application, whether the illumination is laterally uniform is evaluated, and the process is as follows: The standard deviation of the lateral uniformity value corresponding to each lateral uniformity simulation test is calculated, and the simulation lateral uniformity value is obtained. If the simulation lateral uniformity value is less than or equal to the simulation lateral uniformity threshold value, the lateral illumination uniformity signal is displayed.

[0010] As a further scheme of the present application, the illumination regions in the multiple illumination simulation tests are analyzed, and the sub-region illumination mutation value is obtained, and the process is as follows: The left illumination region and the right illumination region are taken as target objects, and the local water accumulation sub-region and the local non-water accumulation sub-region are extracted. The water accumulation reflection illumination and the non-water accumulation reflection illumination corresponding to the local water accumulation sub-region and the local non-water accumulation sub-region are acquired. The water accumulation reflection illuminance corresponding to the water accumulation area and the non-water accumulation reflection illuminance are input into a convolution layer in a convolution neural network to construct a reflection illuminance convolution neural network, the water accumulation reflection illuminance and the non-water accumulation reflection illuminance in each reflection illuminance convolution layer are subjected to ratio calculation, and mean value processing is performed to output a sub-area illumination sudden value.

[0011] As a further scheme of the present application, the screening process of the illumination sudden change sub-area is as follows: If the sub-area illumination sudden value is greater than or equal to a sub-area illumination sudden threshold value, the analyzed illumination sub-area is identified as an illumination sudden change sub-area.

[0012] As a further scheme of the present application, the illumination sudden change sub-area is analyzed to obtain a convolution comparison distribution value, and the process is as follows: Each reflection illuminance convolution layer in the reflection illuminance convolution neural network is extracted, and the water accumulation reflection illuminance and the non-water accumulation reflection illuminance in each reflection illuminance convolution layer are subjected to ratio calculation to obtain a unit convolution comparison degree; Each unit convolution comparison degree is input into a Euclidean distance formula to output a convolution comparison distribution value.

[0013] As a further scheme of the present application, the probability degree of the appearance of a light curtain effect in the illumination sudden change sub-area is evaluated, and a light curtain effect sub-area is screened out, and the process is as follows: The convolution comparison distribution value and the sub-area illumination sudden value are input into a geometric product model to output a light curtain probability value; If the light curtain probability value is greater than a light curtain probability threshold value, the analyzed illumination sudden change sub-area is identified as a light curtain effect sub-area; If the light curtain probability value is less than or equal to the light curtain probability threshold value, the analyzed illumination sudden change sub-area is represented as a non-light curtain effect sub-area.

[0014] As a further scheme of the present application, the light curtain efficiency sub-area distance and the driving speed prediction value are obtained in the following manner: The distance sensor is used to obtain the distance to the light curtain efficiency sub-area; The simulation detection period is set, and the simulation detection period is equally divided into simulation detection time periods; The speed sensor is used to obtain the locomotive driving speed in each simulation detection time period, and the standard deviation is calculated to obtain a driving speed standard deviation; If the driving speed standard deviation is greater than a driving speed standard deviation threshold value, the locomotive driving speeds in all simulation detection time periods are compared in size, and the maximum locomotive driving speed is selected as the driving speed prediction value; If the standard deviation of the running speed is less than or equal to the standard deviation of the running speed threshold, it is indicated that the locomotive running speed difference is small in the simulation detection period, and the locomotive running speed in all simulation detection periods is averaged to obtain a running speed prediction value.

[0015] As a further aspect of the application, the light curtain effect prediction time is obtained in the following manner: The running speed prediction value and the distance to the light curtain effect sub-region are combined and input into a distance-speed formula to obtain the light curtain effect prediction time.

[0016] As a further aspect of the application, the illumination dynamic gradient adjustment value is obtained in the following manner: In the light curtain effect prediction time, the light curtain probability value is subtracted from the light curtain probability threshold, and the ratio of the light curtain effect prediction time is calculated to obtain the illumination dynamic gradient adjustment value.

[0017] The beneficial effects of the application are as follows: The application simulates the illumination of each headlamp of the locomotive multiple times in a transverse uniform manner under a simulated night rainfall environment to obtain a simulated transverse uniform value, which reflects the degree of difference between the test data after each simulation test in the process of simulating the illumination of the headlamp of the locomotive multiple times in a transverse uniform manner. The present application analyzes the illumination mutation sub-region, evaluates the probability degree of the light curtain effect in the illumination mutation sub-region, thereby reflecting the probability of the light curtain effect in the illumination mutation sub-region, and helps to identify which illumination mutation sub-region is more prone to the light curtain effect. According to the light curtain probability value of different illumination mutation sub-regions, the parameters of the lighting system can be adjusted accordingly. Moreover, for the light curtain effect sub-region, the light curtain efficiency sub-region distance and the locomotive running speed are obtained, and the light curtain effect prediction time is obtained. In the light curtain effect prediction time, the illumination dynamic gradient adjustment value is obtained, which is beneficial to improve the accuracy and timeliness of subsequent gradient adjustment of the illumination of the locomotive headlamp, reduce the illumination difference caused by the ground water reflection illumination, avoid the occurrence of the light curtain effect, provide a clear field of view for the driver, thereby effectively improving the driving safety of the locomotive and reducing traffic accidents caused by the light curtain effect. BRIEF DESCRIPTION OF DRAWINGS

[0018] The present application will be further described below in conjunction with the accompanying drawings.

[0019] Figure 1 is the step flow chart of embodiment 1 of the present application. DETAILED DESCRIPTION

[0020] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application will be further described below in conjunction with specific embodiments.

[0021] Embodiment 1: Please refer to Figure 1 The multi-sensor fusion locomotive headlamp dynamic illumination detection method described in the embodiment of the present application comprises the following steps: Step 1: Under the simulated night rainfall environment, the illumination of each headlamp of the locomotive is tested multiple times in a horizontal uniform manner, and whether the illumination is horizontal uniform is evaluated. It should be noted that the locomotive has two headlamps, which are distributed in the left and right sides of the locomotive, respectively, as the left headlamp and the right headlamp. In some embodiments, the horizontal uniform value of the unit is obtained during each horizontal uniform simulation test of the illumination of the locomotive headlamp, and the process is as follows: For example, when the left headlamp of the locomotive is tested, the left illumination region is equally divided into a plurality of left illumination sub-regions. The light intensity of each left illumination sub-region is obtained by a photodiode type sensor, and the left illumination sequence is obtained by sorting according to the distance between each left illumination sub-region and the locomotive. Similarly, when the right headlamp of the locomotive is tested, the right illumination region is equally divided into a plurality of right illumination sub-regions. The illuminance of each right illuminance sub-region is obtained by using a photodiode-type sensor, and then sorted according to the distance between each right illuminance sub-region and the locomotive to obtain the right illuminance sequence. It should be noted that the total number of elements in the right illuminance sub-region is the same as the total number of elements in the left illuminance sub-region. Therefore, the total number of elements in the left illuminance sequence is the same as the total number of elements in the right illuminance sequence. Elements with the same order are extracted from the left and right illuminance sequences respectively and combined to obtain multiple left and right illuminance analysis groups; For example, to extract the illuminance corresponding to the first-ranked left illuminance sub-region from the left illuminance sequence, it is necessary to extract the illuminance corresponding to the first-ranked right illuminance sub-region from the right illuminance sequence and combine them to obtain a set of left and right illuminance analysis groups; to extract the illuminance corresponding to the second-ranked left illuminance sub-region from the left illuminance sequence, it is necessary to extract the illuminance corresponding to the second-ranked right illuminance sub-region from the right illuminance sequence and combine them to obtain a set of left and right illuminance analysis groups; to extract the illuminance corresponding to the third-ranked left illuminance sub-region from the left illuminance sequence, it is necessary to extract the illuminance corresponding to the third-ranked right illuminance sub-region from the right illuminance sequence and combine them to obtain a set of left and right illuminance analysis groups. Multiple left and right illuminance analysis groups are input into the Euclidean distance formula, and the lateral uniformity value of the cell is output. ; Specifically, the Euclidean distance formula is: , This represents the total number of left and right illuminance analysis groups. Represented as the first in the left illuminance sequence The lamp illuminance corresponding to each left illuminance sub-region Represented as the th in the right illuminance sequence The lamp illuminance corresponding to each right illuminance sub-region; The standard deviation is calculated based on the horizontal uniform value corresponding to each horizontal uniform simulation test to obtain the simulated horizontal uniform value. If the simulated lateral uniform value is less than or equal to the simulated lateral uniform threshold, it indicates that during the multiple lateral uniform simulation tests of the locomotive headlight illuminance, the difference between the test data after each simulation test is small, and it is displayed as a lateral illuminance uniform signal. If the simulated lateral uniform value is greater than the simulated lateral uniform threshold, it indicates that during the multiple lateral uniform simulation tests of the locomotive headlight illuminance, the test data after each simulation test has a large degree of difference, which is displayed as a lateral illuminance non-uniform signal. Step 2: If a uniform lateral illuminance signal is received, analyze the illuminance region within the multiple illuminance simulation tests and screen out the illuminance mutation sub-regions. In some embodiments, the analysis is performed with the left-side illuminance area as the target object, and the process is as follows: extracting a local water accumulation sub-region in the left illumination sub-region, and obtaining the lamp illumination reflected by the local water accumulation sub-region through the silicon photodiode sensor as water accumulation reflected illumination; Similarly, a local non-water accumulation sub-region in the left illumination sub-region is extracted, and the lamp illumination reflected by the local non-water accumulation sub-region is obtained through the silicon photodiode sensor as non-water accumulation reflected illumination; arbitrarily inputting the water accumulation reflected illumination and the non-water accumulation reflected illumination corresponding to the water accumulation region into the convolution layer in the convolutional neural network to construct a reflected illumination convolutional neural network; It should be noted that the object corresponding to the reflected illumination convolutional neural network is a local water accumulation sub-region; The reflected illumination convolutional neural network is composed of a plurality of reflected illumination convolutional layers, and each reflected illumination convolutional layer is composed of a water accumulation reflected illumination and a non-water accumulation reflected illumination; Further, the water accumulation reflected illumination and the non-water accumulation reflected illumination corresponding to the water accumulation region are input into the convolution layer in the convolutional neural network, and the corresponding rule is that: based on the water accumulation region and the non-water accumulation region, all non-water accumulation regions adjacent to any one water accumulation region are extracted and combined with the water accumulation reflected illumination corresponding to the water accumulation region, and then input into the convolution layer in the convolutional neural network to construct a reflected illumination convolutional layer; In the reflected illumination convolutional neural network, the water accumulation reflected illumination and the non-water accumulation reflected illumination in each reflected illumination convolutional layer are subjected to ratio calculation and mean value processing, and a sub-region illumination value is output; Similarly, the right illumination region is taken as the target object for analysis, and the process is as follows: extracting a local water accumulation sub-region in the right illumination sub-region, and obtaining the lamp illumination reflected by the local water accumulation sub-region through the silicon photodiode sensor as water accumulation reflected illumination; Similarly, a local non-water accumulation sub-region in the right illumination sub-region is extracted, and the lamp illumination reflected by the local non-water accumulation sub-region is obtained through the silicon photodiode sensor as non-water accumulation reflected illumination; arbitrarily inputting the water accumulation reflected illumination and the non-water accumulation reflected illumination corresponding to the water accumulation region into the convolution layer in the convolutional neural network to construct a reflected illumination convolutional neural network; It should be noted that the object corresponding to the reflected illumination convolutional neural network is a water accumulation region; The reflected illumination convolutional neural network is composed of a plurality of reflected illumination convolutional layers, and each reflected illumination convolutional layer is composed of a water accumulation reflected illumination and a non-water accumulation reflected illumination; Further, the water-reflectance and the non-water-reflectance corresponding to any water area are input into a convolution layer of the convolution neural network according to the following rule: the right reflectance sub-area is divided according to the water area and the non-water area, all non-water areas adjacent to any water area are extracted and combined with the water-reflectance corresponding to the water area respectively, and then input into the convolution layer of the convolution neural network to construct a reflectance convolution layer. In the reflectance convolution neural network, the water-reflectance and the non-water-reflectance in each reflectance convolution layer are calculated by ratio and then processed by mean value to obtain a sub-area reflectance value. As understood by those skilled in the art, the water-reflectance and the non-water-reflectance in each reflectance convolution layer are calculated by ratio because the brightness distribution of the water area and the surrounding area is measured, the local contrast is calculated, and the light curtain effect is warned when the local contrast is too large. It can be understood that the sub-area reflectance value represents the difference between the water area and the non-water area in the sub-area under the influence of the reflectance of the locomotive headlamp in the night rain environment, and identifies the area of local reflectance mutation in the sub-area. On the other hand, the sub-area reflectance value can be used as an important index to determine whether to trigger the light curtain effect warning. By analyzing the left and right reflectance sub-areas respectively, the influence of the water reflectance on the reflectance of the locomotive headlamp can be accurately analyzed under different water distribution conditions, and the reflectance difference that may cause the light curtain effect can be found in advance to issue a warning in time, so that the driver has enough time to take measures to avoid traffic accidents caused by the light curtain effect, and the technical problem of difficult early warning of the light curtain effect is solved. The sub-area reflectance value is compared with the sub-area reflectance threshold value as follows: If the sub-area reflectance value is greater than or equal to the sub-area reflectance threshold value, it means that the analyzed reflectance sub-area is greatly affected by the ground water reflectance, and the analyzed reflectance sub-area is identified as a reflectance mutation sub-area. If the sub-area reflectance value is less than the sub-area reflectance threshold value, it means that the analyzed reflectance sub-area is less affected by the ground water reflectance. The technical solution of this embodiment is as follows: Under simulated nighttime rainy conditions, multiple lateral uniformity simulation tests are conducted on the illuminance of each headlight of the locomotive to obtain simulated lateral uniformity values. These simulated lateral uniformity values ​​reflect the degree of difference between the test data after each simulation test during the multiple lateral uniformity simulation tests of the locomotive headlight illuminance. Furthermore, if a lateral illuminance uniformity signal is received, the illuminance area within the multiple illuminance simulation tests is analyzed to obtain sub-region illuminance spike values. These sub-region illuminance spike values ​​reflect the degree of difference between the reflected illuminance of the locomotive headlights in the sub-region under nighttime rainy conditions, between water-filled and non-water-filled areas. The left and right illuminance sub-regions are analyzed separately to adapt to different water distribution conditions, accurately analyze the impact of water-reflected illuminance on the locomotive headlight illuminance, and detect illuminance differences that may cause a light curtain effect in advance, issuing timely warnings and providing drivers with sufficient time to take measures to avoid traffic accidents caused by the light curtain effect. This solves the technical problem of the difficulty in providing early warnings for the light curtain effect.

[0022] Example 2: Please refer to Figure 1 As shown in the embodiment of the present invention, a multi-sensor fusion method for dynamic illuminance detection of locomotive headlights includes the following steps: Step 3: Analyze the illuminance mutation sub-regions, assess the probability of light curtain effect occurring in the illuminance mutation sub-regions, and screen out the light curtain effect sub-regions; In some embodiments, each reflectance convolutional layer in the reflectance convolutional neural network is extracted, and the ratio of the reflectance of water accumulation to the reflectance of non-water accumulation in each reflectance convolutional layer is calculated to obtain the unit convolutional comparison degree. The convolutional alignment score of each unit is input into the Euclidean distance formula, and the output is the convolutional alignment distribution value. ; Specifically, the Euclidean distance formula is: ,in, This represents the total number of unit convolutional alignment degrees. Represented as the first Unit convolutional alignment degree Represented as the first Unit convolutional alignment; The convolutional alignment distribution value and the sub-region illumination burst value are input into the geometric product model, and the output is the light curtain probability value. The skilled in the art can understand that the meaning represented by the light curtain probability value is that the unit convolution comparison degree distribution in the reflected illumination convolutional neural network (reflected by the convolution comparison distribution value) and the difference degree of the water accumulation and non-water accumulation area in the sub-area (reflected by the sub-area illumination sudden value) are integrated, reflecting the probability of the light curtain effect in the illumination sudden change sub-area, which is helpful to identify which illumination sudden change sub-area is more prone to light curtain effect. According to the light curtain probability value of different illumination sudden change sub-areas, the parameters of the lighting system can be adjusted accordingly, so as to effectively avoid the line of sight obstruction caused by the light curtain effect and improve the driving safety of the locomotive in the night rainfall environment. The light curtain probability value is compared with the light curtain probability threshold value, and the process is as follows: If the light curtain probability value is greater than the light curtain probability threshold value, it means that the probability of light curtain effect in the analyzed illumination sudden change sub-area is relatively high, and the analyzed illumination sudden change sub-area is identified as a light curtain effect sub-area. If the light curtain probability value is less than or equal to the light curtain probability threshold value, it means that the probability of light curtain effect in the analyzed illumination sudden change sub-area is relatively low, and the analyzed illumination sudden change sub-area is identified as a non-light curtain effect sub-area. Step four: for the light curtain effect sub-area, the distance to the light curtain efficiency sub-area and the driving speed prediction value are obtained, and the light curtain effect prediction time is obtained. In the light curtain effect prediction time, the illumination dynamic gradient adjustment value is obtained to reduce the probability of light curtain effect. In some embodiments, the distance to the light curtain efficiency sub-area is obtained by using a distance sensor. Set the simulation detection period, and equally divide the simulation detection period into simulation detection time periods. The locomotive driving speed in each simulation detection time period is obtained by using a speed sensor, and the standard deviation is calculated to obtain the driving speed standard deviation. If the driving speed standard deviation is greater than the driving speed standard deviation threshold value, it means that the locomotive driving speed difference is large in the simulation detection period, then the locomotive driving speed in all simulation detection time periods is compared, and the maximum locomotive driving speed is selected as the driving speed prediction value, which is combined with the distance to the light curtain effect sub-area and input into the distance-speed formula to output the light curtain effect prediction time. If the driving speed standard deviation is less than or equal to the driving speed standard deviation threshold value, it means that the locomotive driving speed difference is small in the simulation detection period, then the locomotive driving speed in all simulation detection time periods is calculated by mean value, which is combined with the distance to the light curtain effect sub-area and input into the distance-speed formula to output the light curtain effect prediction time. In the light curtain effect prediction time, the light curtain probability value is subtracted from the light curtain probability threshold value, and the ratio is calculated to obtain the illumination dynamic gradient adjustment value. It should be noted that the purpose of obtaining the illumination dynamic gradient adjustment value is to consider the running speed of the locomotive, obtain the more accurate time of reaching the light curtain effect sub-region according to the change of the running speed of the locomotive combined with the distance of the light curtain effect sub-region, thereby improving the accuracy and timeliness of subsequent gradient adjustment of the illumination of the front headlamp of the locomotive, reducing the illumination difference caused by ground area water reflection illumination, avoiding the occurrence of light curtain effect, providing a clear field of view for the driver, thereby effectively improving the driving safety of the locomotive and reducing traffic accidents caused by light curtain effect. The technical scheme of the embodiment is: analyzing the illumination mutation sub-region, evaluating the probability degree of the occurrence of light curtain effect in the illumination mutation sub-region, thereby reflecting the probability of the occurrence of light curtain effect in the illumination mutation sub-region, helping to identify which illumination mutation sub-region is more prone to light curtain effect, according to the light curtain probability value of different illumination mutation sub-regions, the parameters of the lighting system can be adjusted accordingly, and for the light curtain effect sub-region, the distance of the light curtain efficiency sub-region and the running speed of the locomotive are obtained to obtain the light curtain effect expected time, and the illumination dynamic gradient adjustment value is obtained within the light curtain effect expected time, which is conducive to improving the accuracy and timeliness of subsequent gradient adjustment of the illumination of the front headlamp of the locomotive, reducing the illumination difference caused by ground area water reflection illumination, avoiding the occurrence of light curtain effect, providing a clear field of view for the driver, thereby effectively improving the driving safety of the locomotive and reducing traffic accidents caused by light curtain effect.

[0023] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-sensor fusion locomotive headlamp dynamic illumination detection method, characterized in that: The application relates to a light curtain effect prediction method and device for a locomotive. The application comprises: In a simulated night rainfall environment, the illuminance of each headlamp of the locomotive is tested multiple times in a lateral uniform manner to evaluate whether the illuminance is laterally uniform; If a lateral illuminance uniformity signal is received, the illuminance regions in the multiple illuminance simulation tests are analyzed to screen out illuminance mutation sub-regions; The illuminance mutation sub-regions are analyzed to evaluate the probability of light curtain effect in the illuminance mutation sub-regions and screen out light curtain effect sub-regions; 2. The multi-sensor fusion based dynamic illuminance detection method for locomotive headlamps as claimed in claim 1 wherein: For the light curtain effect sub-regions, the light curtain efficiency sub-region distance and the locomotive running speed are obtained to obtain the light curtain effect prediction time, and the illuminance dynamic gradient adjustment value is obtained within the light curtain effect prediction time to reduce the probability of light curtain effect. In each lateral uniformity simulation test of the locomotive headlamp illuminance, the unit lateral uniformity value is obtained, and the process is as follows: The left illuminance region is equally divided into a plurality of left illuminance sub-regions, and the right illuminance region is equally divided into a plurality of right illuminance sub-regions; The lamp illuminance of each left illuminance sub-region and the lamp illuminance of each right illuminance sub-region are obtained to construct a left illuminance sequence and a right illuminance sequence; 3. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: The same ordered elements are combined from the left illuminance sequence and the right illuminance sequence to obtain a plurality of left and right illuminance analysis groups, which are input into the Euclidean distance formula to output the unit lateral uniformity value. The process of evaluating whether the illuminance is laterally uniform is as follows: The standard deviation of the lateral uniformity value corresponding to each lateral uniformity simulation test is calculated to obtain a simulation lateral uniformity value; 4. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: If the simulation lateral uniformity value is less than or equal to the simulation lateral uniformity threshold, a lateral illuminance uniformity signal is displayed. The process of analyzing the illuminance regions in the multiple illuminance simulation tests to obtain a sub-region illuminance sudden value is as follows: The left illuminance region and the right illuminance region are taken as target objects, and local water accumulation sub-regions and local non-water accumulation sub-regions are extracted; The water accumulation reflection illuminance and the non-water accumulation reflection illuminance corresponding to the local water accumulation sub-regions and the local non-water accumulation sub-regions are obtained; 5. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: The water accumulation reflection illuminance and the non-water accumulation reflection illuminance corresponding to the water accumulation region are input into the convolution layer of the convolutional neural network to construct a reflection illuminance convolutional neural network, the water accumulation reflection illuminance and the non-water accumulation reflection illuminance in each reflection illuminance convolution layer are subjected to ratio calculation, and mean value processing is performed to output the sub-region illuminance sudden value. The process of screening the illuminance mutation sub-regions is as follows:

6. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 4 wherein: If the sub-region illuminance sudden value is greater than or equal to the sub-region illuminance sudden threshold, the analyzed illuminance sub-region is identified as an illuminance mutation sub-region. The process of analyzing the illuminance mutation sub-regions to obtain a convolution comparison distribution value is as follows: Each reflection illuminance convolution layer in the reflection illuminance convolutional neural network is extracted, and the water accumulation reflection illuminance and the non-water accumulation reflection illuminance in each reflection illuminance convolution layer are subjected to ratio calculation to obtain a unit convolution comparison degree; 7. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 5 wherein: Each unit convolution comparison degree is input into the Euclidean distance formula to output the convolution comparison distribution value. The process of evaluating the probability of light curtain effect in the illuminance mutation sub-regions to screen out light curtain effect sub-regions is as follows: The convolution comparison distribution value and the sub-region illuminance sudden value are input into a geometric product model to output a light curtain probability value; If the light curtain probability value is greater than the light curtain probability threshold value, the analyzed illumination mutation sub-region is identified as a light curtain effect sub-region; If the light curtain probability value is less than or equal to the light curtain probability threshold value, the analyzed illumination mutation sub-region is represented as a non-light curtain effect sub-region.

8. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: The acquisition method of the light curtain efficiency sub-region distance and the travel speed prediction value is: The distance sensor is used to obtain the distance to the light curtain efficiency sub-region; The simulation detection period is set, and the simulation detection period is equally divided into simulation detection time periods; The speed sensor is used to obtain the locomotive travel speed in each simulation detection time period, and the standard deviation is calculated to obtain the travel speed standard deviation; If the travel speed standard deviation is greater than the travel speed standard deviation threshold value, the locomotive travel speeds in all simulation detection time periods are compared in size, and the maximum locomotive travel speed is selected as the travel speed prediction value; If the travel speed standard deviation is less than or equal to the travel speed standard deviation threshold value, it indicates that the locomotive travel speed difference is small in the simulation detection period, and the locomotive travel speeds in all simulation detection time periods are calculated by mean value, as the travel speed prediction value.

9. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: The acquisition method of the light curtain effect prediction time is: The travel speed prediction value and the distance to the light curtain effect sub-region are combined and input into the distance-speed formula to output the light curtain effect prediction time.

10. The multi-sensor fusion based dynamic headlamp illumination detection method for locomotive as claimed in claim 1 wherein: The acquisition process of the illumination dynamic gradient adjustment value is as follows: In the light curtain effect prediction time, the light curtain probability value is subtracted from the light curtain probability threshold value, and the ratio calculation is performed with the light curtain effect prediction time to obtain the illumination dynamic gradient adjustment value.

Images