Road surface natural color visible detection method based on deep learning

Abstract

The invention discloses a pavement natural color seeing detection method based on deep learning, and relates to the technical field of deep learning, and the monitoring method comprises the following steps: collecting the pavement material of each road, dividing each road according to the pavement material, and obtaining a plurality of types of road sets; a stored road surface shot picture is obtained, a color recognition model is constructed to recognize the road surface shot picture, and a color value is obtained; randomly selecting two detection records from the detection log, and analyzing the difference of the two detection records to obtain an influence coefficient of each influence factor on a color value; the influence coefficient of each influence factor on the color value is analyzed, the color value is calibrated, and the error range of the same kind of road set is obtained; when a road is detected, generating a detection record of the road; and comparing the actual characteristic value and the theoretical characteristic value of the detection record, and judging whether abnormity reminding is carried out on the road or not.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a method for detecting the true color of a road surface based on deep learning. Background Art

[0002] In the field of urban management and environmental sanitation, the See the Original Colors Campaign aims to restore the city to its original color and state through cleaning and maintenance, remove dust and dirt, and show the true face of the city, thereby improving the overall image of the city and the living environment of residents; Seeing the Original Color of the Road Surface is an important part of the See the Original Colors Campaign, which mainly refers to the thorough cleaning and maintenance of urban roads so that the roads can be restored to their original state and color.

[0003] Urban roads, due to long-term influence of various factors such as traffic, climate, pollution, etc., often have oil stains, dust, dirt and other garbage, which will obscure the original color of the road; if the various pollution on the road cannot be cleared in time, the impact will damage the image of the urban road at a minor level, and even cause damage or obstruction to the road in serious cases, posing a threat to people's driving safety. Summary of the invention

[0004] The purpose of the present invention is to provide a road surface true color detection method based on deep learning to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution: a road surface true color detection method based on deep learning, the monitoring method comprises the following steps:

[0006] Step S100: Establishing a road surface true color detection platform, generating a detection log for each road, storing the detection process of the road, and generating detection records in the detection log; collecting the road surface material of each road, dividing each road according to the road surface material, and obtaining a set of several types of roads;

[0007] Step S200: arbitrarily selecting a detection log of a road, arbitrarily extracting a detection record in the detection log; obtaining road surface photos stored in the detection record, constructing a color recognition model to recognize the road surface photos, and respectively obtaining color values ​​of the road surface photos;

[0008] Step S300: obtaining a set of influencing factors stored in a detection log of any road; selecting two detection records at random from the detection log, and collecting the influencing factors in the two detection records respectively; constructing an influencing factor evaluation model, analyzing the differences between the photos stored in the two detection records and the influencing factors, and obtaining the influence coefficient of each influencing factor on the color value;

[0009] Step S400: analyzing the influence coefficient of each influencing factor on the color value, calibrating the color value of the road surface photograph stored in each detection log in the detection log, obtaining the actual color value of the road surface photograph, and confirming the average color value of the road corresponding to the detection log;

[0010] Step S500: obtaining the average color value of each road in the same type of road set to obtain the actual color value range of the same type of road set; analyzing the distribution of each average color value in the actual color value range to obtain the error range of the same type of road set;

[0011] Step S600: Whenever a road is inspected, a detection record of the road is generated on the road surface detection platform; and a theoretical characteristic value of the detection record is obtained through the difference between the detection record and adjacent detection records;

[0012] Step S700: Calculate the color value of the road surface photograph stored in the detection record through a color recognition model to obtain an actual feature value; compare the actual feature value of the detection record with the theoretical feature value to determine whether to issue an abnormal reminder for the road.

[0013] Furthermore, step S100 includes the following steps:

[0014] Step S101: collecting the pavement materials of each road respectively and numbering each road; constructing a material database, wherein the material database contains several kinds of materials;

[0015] Step S102: randomly selecting a road pavement material and matching it with several materials in the material database. If there is a material in the material database that is the same as the pavement material, then feature marking is performed on the material.

[0016] Step S103: extracting materials containing characteristic marks from the material database, generating a corresponding road set for each material containing characteristic marks; and assigning the number of each road to the road set corresponding to the same material according to the pavement material.

[0017] Further, step S200 includes the following steps:

[0018] Step S201: arbitrarily obtain a road surface photograph stored in a detection record, convert the format of the road surface photograph, and generate an RGB image of the road surface photograph; obtain the value R of the red channel, the value G of the green channel, and the value B of the blue channel of the RGB image, and calculate the brightness of the RGB image as L=(R+G+B) / 3;

[0019] Step S202: convert the RGB image into Lab color space, set the red and green values ​​to a and the yellow and blue values ​​to b, and obtain a=(RG) / 2 and b=(GB) / 2;

[0020] Step S203: constructing a formula in the color recognition model: C=α×(L+a+b) / 3+β, where α and β are constant coefficients, and calculating the color value C of the road surface photograph;

[0021] Step S204: repeatedly input the road photograph into the color recognition model to calculate the color value, setting the number of times the road photograph is input into the color recognition model to N, and the number of times the same color value is counted to N. s , the recognition accuracy of the color recognition model is calculated to be Z = N s / N, set the accurate threshold to Z max , if Z ≥ Z max , then set α=β=1, if Z<Z max , the color recognition model is trained to determine α and β; repeatedly inputting the same road photograph into the color recognition model can help eliminate errors in the color recognition process, optimize the color recognition model, and obtain accurate constant coefficients; this is conducive to the accuracy of subsequent color recognition of each road photograph.

[0022] Further, step S300 includes the following steps:

[0023] Step S301: arbitrarily select the road photograph of the i-th storage record in the detection log, obtain the shooting time of the road photograph, and obtain that at the shooting time, the value of the j-th influencing factor in the influencing factor set stored in the i-th storage record is X j ;

[0024] Step S302: arbitrarily select two road photos of the detection record from the detection log, obtain the color values ​​of the two road photos respectively, and obtain the color difference of the two road photos as ΔC; obtain the numerical difference of the jth influencing factor in the influencing factor set stored in the two detection records as ΔX j ; Evaluate the model based on influencing factors:

[0025]

[0026] Among them, γ jis the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection log, and m is the number of influencing factors in the influencing factor set; by analyzing the color difference between two adjacent road surface photographs, the reason for the color difference is the influence of various environmental factors, because the two shooting environments are different, and due to the joint influence of multiple environmental factors, it is only necessary to obtain the comprehensive influence coefficient of each environmental factor; through the difference between any two photos, the comprehensive influence coefficient is determined numerically;

[0027] Step S303: Input the color difference of the road surface photos of any two detection records in the detection log and the numerical difference of each influencing factor into the influencing factor evaluation model and train the influencing factor evaluation model to determine the influence coefficient γ of the jth influencing factor on the color value of the road surface photos in the detection log j By photographing the road, the color presentation of the photographed photos can be obtained. According to the differences between different photos of the same road, the pollution changes of the road can be reflected, and the state of the road can be reflected more intuitively.

[0028] Further, step S400 includes the following steps:

[0029] Step S401: Setting the color value of the road photograph of the i-th storage record to C i , according to the formula:

[0030]

[0031] Calculate the actual color value T of the road surface photo of the i-th storage record i ; Because the color difference value of the road photo will be affected by various environmental factors, only by eliminating the influence of various environmental factors can the accurate actual color value be obtained; through the influence degree of each influencing factor obtained above, the environmental factors can be calibrated to obtain a more accurate actual color value;

[0032] Step S402: Obtain the actual color value of the road surface photograph of each stored record in the detection log, and obtain the actual color value range r of the road surface photograph in the detection log; and calculate the average value of the actual color value of the road surface photograph of each stored record to obtain the average color value of the road surface photograph in the detection log.

[0033] Further, step S500 includes the following steps:

[0034] Step S501: arbitrarily select a road set and set the actual color value range of the ath road in the road set to r a; According to the formula:

[0035]

[0036] Wherein, u is the number of roads in the road set; the actual color value range R of the road set is obtained by calculation;

[0037] Step S502: Obtain the average color value of the road surface photos in the detection log of each road in the road set, and calculate the average value to obtain the characteristic color value T of the road set. R ; Set the minimum actual color value in the actual color value range R of the road set to r min and the maximum actual color value is r max , according to the formula:

[0038]

[0039] The maximum deviation degree of the road set is calculated to be P; the error range of the road set is set to (0, P); according to the previous photo shooting conditions of each road in the road set, the color values ​​of the road in different time periods under normal circumstances can be obtained, and the color value range of the normal road can be obtained as a reference standard for judging road abnormalities.

[0040] Further, step S600 includes the following steps:

[0041] Step S601: Collect the pavement material of the road currently being tested to obtain the road set where the road is located; obtain the influencing factor set of the road, obtain the value of each influencing factor in the influencing factor set, and set the value of the jth influencing factor to

[0042] Step S602: Obtain the adjacent detection record with the shortest time interval from the current shooting time in the detection log of the road, and extract the color value of the adjacent detection record as C ’ ; Extract the value of the jth influencing factor in the influencing factor set stored in the adjacent detection record as X j ′, the numerical difference of the jth influencing factor between two detection records in the detection log is According to the formula:

[0043]

[0044] Among them, m is the number of influencing factors in the influencing factor set, γ j is the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection record; the theoretical characteristic value of the road surface photograph in the detection record is calculated to be Cnew ; According to the formula: Calculate the theoretical characteristic value T of the road surface photograph in the detection record new .

[0045] Furthermore, step S700 includes the following steps:

[0046] Step S701: Input the road photograph in the detection record into the color recognition model to obtain the color value of the road photograph as C′ new ; According to the formula: Calculate the actual characteristic value T′ of the road surface photograph in the detection record new ; The actual color value of the road is obtained by directly photographing the road and the theoretical color value of the road is obtained by observing the color value change of the road before. The difference between the two color values ​​is analyzed to determine whether the road is polluted;

[0047] Step S702: Set the error range of the road set to (0, P), and calculate the error deviation degree of the road surface photos in the detection record to be P new =|T new -T′ new | / T′ new ;like An abnormal reminder is then given for the road; a timely abnormal reminder is given for situations where the color value difference is large, so that the road situation can be checked in the first place to protect people's driving safety.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The present invention obtains the color presentation of the photographed photos by photographing the road. According to the differences between different photos of the same road, the pollution changes of the road can be reflected, and the state of the road can be reflected more intuitively;

[0050] 2. The present invention calculates the color value of the photographed photo, considers that the color presentation of the road is affected by various environmental factors, analyzes the degree of influence of various environmental factors, calibrates the color value of the photographed photo, eliminates the interference of environmental factors, and can obtain the color presentation of the road more accurately and reflect the pollution situation of the road;

[0051] 3. The present invention obtains the actual color value of the road by directly photographing the road and obtains the theoretical color value of the road by analyzing the change in the previous color value of the road, analyzes the difference between the two color values, determines whether the road is polluted, and issues timely abnormal reminders for situations where the color value difference is large. It can check the road condition in the first place, protect people's driving safety, and help the road restore its original color. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a structural diagram of a road surface true color detection method based on deep learning. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example: Figure 1 As shown, the present invention provides a road surface color detection method based on deep learning, and the monitoring method comprises the following steps:

[0055] Step S100: Establishing a road surface true color detection platform, generating a detection log for each road, storing the detection process of the road, and generating detection records in the detection log; collecting the road surface material of each road, dividing each road according to the road surface material, and obtaining a set of several types of roads;

[0056] Wherein, step S100 includes the following steps:

[0057] Step S101: collecting the pavement materials of each road respectively and numbering each road; constructing a material database, wherein the material database contains several kinds of materials;

[0058] Step S102: randomly selecting a road pavement material and matching it with several materials in the material database. If there is a material in the material database that is the same as the pavement material, then feature marking is performed on the material.

[0059] Step S103: extracting materials containing characteristic marks from the material database, generating a corresponding road set for each material containing the characteristic mark; and assigning the number of each road to the road set corresponding to the same material according to the pavement material;

[0060] Embodiment 1: Assume that the material database includes asphalt, cement, etc.; collect the pavement material of a road asphalt, and if it matches the material in the material database successfully, then feature mark the asphalt; classify all roads whose pavement material is asphalt into the same category.

[0061] Step S200: arbitrarily selecting a detection log of a road, arbitrarily extracting a detection record in the detection log; obtaining road surface photos stored in the detection record, constructing a color recognition model to recognize the road surface photos, and respectively obtaining color values ​​of the road surface photos;

[0062] Wherein, step S200 includes the following steps:

[0063] Step S201: arbitrarily obtain a road surface photograph stored in a detection record, convert the format of the road surface photograph, and generate an RGB image of the road surface photograph; obtain the value R of the red channel, the value G of the green channel, and the value B of the blue channel of the RGB image, and calculate the brightness of the RGB image as L=(R+G+B) / 3;

[0064] Step S202: convert the RGB image into Lab color space, set the red and green values ​​to a and the yellow and blue values ​​to b, and obtain a=(RG) / 2 and b=(GB) / 2;

[0065] Step S203: constructing a formula in the color recognition model: C=α×(L+a+b) / 3+β, where α and β are constant coefficients, and calculating the color value C of the road surface photograph;

[0066] Step S204: repeatedly input the road photograph into the color recognition model to calculate the color value, setting the number of times the road photograph is input into the color recognition model to N, and the number of times the same color value is counted to N. s , the recognition accuracy of the color recognition model is calculated to be Z = N s / N, set the accurate threshold to Z max , if Z ≥ Z max , then set α=β=1, if Z<Z max , then the color recognition model is trained to determine α and β;

[0067] Example 2: Set the number of times a road photo is input into the color recognition model to 10 times, the number of times the same color value is obtained is 9 times, and the recognition accuracy of the color recognition model is 90%; set the accuracy threshold to 80%, because 90%>80%, indicating that the color recognition of the color recognition model is relatively accurate, and obtain α=β=1.

[0068] Step S300: obtaining a set of influencing factors stored in a detection log of any road; selecting two detection records at random from the detection log, and collecting the influencing factors in the two detection records respectively; constructing an influencing factor evaluation model, analyzing the differences between the photos stored in the two detection records and the influencing factors, and obtaining the influence coefficient of each influencing factor on the color value;

[0069] Wherein, step S300 includes the following steps:

[0070] Step S301: arbitrarily select the road photograph of the i-th storage record in the detection log, obtain the shooting time of the road photograph, and obtain that at the shooting time, the value of the j-th influencing factor in the influencing factor set stored in the i-th storage record is X j ;

[0071] Step S302: arbitrarily select two road photos of the detection record from the detection log, obtain the color values ​​of the two road photos respectively, and obtain the color difference of the two road photos as ΔC; obtain the numerical difference of the jth influencing factor in the influencing factor set stored in the two detection records as ΔX j ; Evaluate the model based on influencing factors:

[0072]

[0073] Among them, γ j is the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection log, and m is the number of influencing factors in the influencing factor set;

[0074] Example 3: The influencing factor set includes temperature and humidity; the color difference between photo A and photo B is set to be 6.6, where the temperature difference is 2 and the humidity difference is 8; the color difference between photo A and photo C is set to be 4.8, where the temperature difference is 1 and the humidity difference is 8; and the following is obtained:

[0075] 6.6=2×v+10×w, 4.8=1×v+8×w, we get v=4, w=0.5; the influence of temperature is 0.8, and the influence of humidity is 0.5;

[0076] Step S303: Input the color difference of the road surface photos of any two detection records in the detection log and the numerical difference of each influencing factor into the influencing factor evaluation model and train the influencing factor evaluation model to determine the influence coefficient γ of the jth influencing factor on the color value of the road surface photos in the detection log j The value of .

[0077] Step S400: analyzing the influence coefficient of each influencing factor on the color value, calibrating the color value of the road surface photograph stored in each detection log in the detection log, obtaining the actual color value of the road surface photograph, and confirming the average color value of the road corresponding to the detection log;

[0078] Wherein, step S400 includes the following steps:

[0079] Step S401: Setting the color value of the road photograph of the i-th storage record to C i , according to the formula:

[0080]

[0081] Calculate the actual color value T of the road surface photo of the i-th storage record i ;

[0082] Example 3: The influence degree of temperature is set to 0.8, the influence degree of humidity is set to 0.5; the color value is 10, and the actual color value is 10×(1-0.8)×(1-0.5)=1;

[0083] Step S402: Obtain the actual color value of the road surface photograph of each stored record in the detection log, and obtain the actual color value range r of the road surface photograph in the detection log; and calculate the average value of the actual color value of the road surface photograph of each stored record to obtain the average color value of the road surface photograph in the detection log.

[0084] Step S500: obtaining the average color value of each road in the same type of road set to obtain the actual color value range of the same type of road set; analyzing the distribution of each average color value in the actual color value range to obtain the error range of the same type of road set;

[0085] Wherein, step S500 includes the following steps:

[0086] Step S501: arbitrarily select a road set and set the actual color value range of the ath road in the road set to r a ; According to the formula:

[0087]

[0088] Wherein, u is the number of roads in the road set; the actual color value range R of the road set is obtained by calculation;

[0089] Step S502: Obtain the average color value of the road surface photos in the detection log of each road in the road set, and calculate the average value to obtain the characteristic color value T of the road set.R ; Set the minimum actual color value in the actual color value range R of the road set to r min and the maximum actual color value is r max , according to the formula:

[0090]

[0091] The maximum deviation degree of the road set is calculated to be P; and the error range of the road set is set to (0, P).

[0092] Step S600: Whenever a road is inspected, a detection record of the road is generated on the road surface detection platform; and a theoretical characteristic value of the detection record is obtained through the difference between the detection record and adjacent detection records;

[0093] Wherein, step S600 includes the following steps:

[0094] Step S601: Collect the pavement material of the road currently being tested to obtain the road set where the road is located; obtain the influencing factor set of the road, obtain the value of each influencing factor in the influencing factor set, and set the value of the jth influencing factor to

[0095] Step S602: Obtain the adjacent detection record with the shortest time interval from the current shooting time in the detection log of the road, and extract the color value of the adjacent detection record as C ’ ; Extract the value of the jth influencing factor in the influencing factor set stored in the adjacent detection record as X j ′, the numerical difference of the jth influencing factor between two detection records in the detection log is According to the formula:

[0096]

[0097] Among them, m is the number of influencing factors in the influencing factor set, γ j is the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection record; the theoretical characteristic value of the road surface photograph in the detection record is calculated to be C new ; According to the formula: Calculate the theoretical characteristic value T of the road surface photograph in the detection record new .

[0098] Step S700: Calculate the color value of the road surface photograph stored in the detection record by using a color recognition model to obtain an actual feature value; compare the actual feature value of the detection record with the theoretical feature value to determine whether to issue an abnormality reminder for the road;

[0099] Wherein, step S700 includes the following steps:

[0100] Step S701: Input the road photograph in the detection record into the color recognition model to obtain the color value of the road photograph as C′ new ; According to the formula: Calculate the actual characteristic value T′ of the road surface photograph in the detection record new ;

[0101] Step S702: Set the error range of the road set to (0, P), and calculate the error deviation degree of the road surface photos in the detection record to be P new =|T new -T′ new | / T′ new ;like Then an abnormal reminder is given for the road.

[0102] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A road surface true color detection method based on deep learning, characterized by: The monitoring method comprises the following steps: Step S100: Establishing a road surface true color detection platform, generating a detection log for each road, storing the detection process of the road, and generating detection records in the detection log; collecting the road surface material of each road, dividing each road according to the road surface material, and obtaining a set of several types of roads; Step S200: arbitrarily selecting a detection log of a road, arbitrarily extracting a detection record in the detection log; obtaining road surface photos stored in the detection record, constructing a color recognition model to recognize the road surface photos, and respectively obtaining color values ​​of the road surface photos; Step S300: obtaining a set of influencing factors stored in a detection log of any road; selecting two detection records at random from the detection log, and collecting the influencing factors in the two detection records respectively; constructing an influencing factor evaluation model, analyzing the differences between the photos stored in the two detection records and the influencing factors, and obtaining the influence coefficient of each influencing factor on the color value; Step S400: analyzing the influence coefficient of each influencing factor on the color value, calibrating the color value of the road surface photograph stored in each detection log in the detection log, obtaining the actual color value of the road surface photograph, and confirming the average color value of the road corresponding to the detection log; Step S500: obtaining the average color value of each road in the same type of road set to obtain the actual color value range of the same type of road set; analyzing the distribution of each average color value in the actual color value range to obtain the error range of the same type of road set; Step S600: Whenever a road is inspected, a detection record of the road is generated on the road surface detection platform; and a theoretical characteristic value of the detection record is obtained through the difference between the detection record and adjacent detection records; Step S700: Calculate the color value of the road surface photograph stored in the detection record through a color recognition model to obtain an actual feature value; compare the actual feature value of the detection record with the theoretical feature value to determine whether to issue an abnormal reminder for the road.

2. The method for detecting the road surface's true color based on deep learning according to claim 1, characterized in that: The step S100 includes the following steps: Step S101: collecting the pavement materials of each road respectively and numbering each road; constructing a material database, wherein the material database contains several kinds of materials; Step S102: randomly selecting a road pavement material and matching it with several materials in the material database. If there is a material in the material database that is the same as the pavement material, then feature marking is performed on the material. Step S103: extracting materials containing characteristic marks from the material database, generating a corresponding road set for each material containing characteristic marks; and assigning the number of each road to the road set corresponding to the same material according to the pavement material.

3. The method for detecting the road surface's true color based on deep learning according to claim 2, characterized in that: The step S200 includes the following steps: Step S201: arbitrarily obtain a road surface photograph stored in a detection record, convert the format of the road surface photograph, and generate an RGB image of the road surface photograph; obtain the value R of the red channel, the value G of the green channel, and the value B of the blue channel of the RGB image, and calculate the brightness of the RGB image as L=(R+G+B) / 3; Step S202: convert the RGB image into Lab color space, set the red and green values ​​to a and the yellow and blue values ​​to b, and obtain a=(RG) / 2 and b=(GB) / 2; Step S203: constructing a formula in the color recognition model: C=α×(L+a+b) / 3+β, where α and β are constant coefficients, and calculating the color value C of the road surface photograph; Step S204: repeatedly input the road photograph into the color recognition model to calculate the color value, setting the number of times the road photograph is input into the color recognition model to N, and the number of times the same color value is counted to N. s , the recognition accuracy of the color recognition model is calculated to be Z = N s / N, set the accurate threshold to Z max , if Z ≥ Z max , then set α=β=1, if Z<Z max , then the color recognition model is trained to determine α and β.

4. The method for detecting the road surface's true color based on deep learning according to claim 3, characterized in that: The step S300 includes the following steps: Step S301: arbitrarily select the road photograph of the i-th storage record in the detection log, obtain the shooting time of the road photograph, and obtain that at the shooting time, the value of the j-th influencing factor in the influencing factor set stored in the i-th storage record is X j ; Step S302: arbitrarily select two road photos of the detection record from the detection log, obtain the color values ​​of the two road photos respectively, and obtain the color difference of the two road photos as ΔC; obtain the numerical difference of the jth influencing factor in the influencing factor set stored in the two detection records as ΔX j ; Evaluate the model based on influencing factors: Among them, γ j is the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection log, and m is the number of influencing factors in the influencing factor set; Step S303: Input the color difference between the road surface photos of any two detection records in the detection log and the numerical difference of each influencing factor into the influencing factor evaluation model and train the influencing factor evaluation model to determine the influence coefficient γ of the jth influencing factor on the color value of the road surface photos in the detection log j The value of .

5. The method for detecting the road surface's true color based on deep learning according to claim 4, characterized in that: The step S400 includes the following steps: Step S401: Setting the color value of the road photograph of the i-th storage record to C i , according to the formula: Calculate the actual color value T of the road surface photo of the i-th storage record i ; Step S402: Obtain the actual color value of the road surface photograph of each stored record in the detection log, and obtain the actual color value range r of the road surface photograph in the detection log; and calculate the average value of the actual color value of the road surface photograph of each stored record to obtain the average color value of the road surface photograph in the detection log.

6. The method for detecting the road surface's true color based on deep learning according to claim 5, characterized in that: The step S500 includes the following steps: Step S501: arbitrarily select a road set and set the actual color value range of the ath road in the road set to r a ; According to the formula: Wherein, u is the number of roads in the road set; the actual color value range R of the road set is obtained by calculation; Step S502: Obtain the average color value of the road surface photos in the detection log of each road in the road set, and calculate the average value to obtain the characteristic color value T of the road set. R ; Set the minimum actual color value in the actual color value range R of the road set to r min and the maximum actual color value is r max , according to the formula: The maximum deviation degree of the road set is calculated to be P; and the error range of the road set is set to (0, P).

7. The method for detecting the road surface's true color based on deep learning according to claim 6, characterized in that: The step S600 includes the following steps: Step S601: Collect the pavement material of the road currently being tested to obtain the road set where the road is located; obtain the influencing factor set of the road, obtain the value of each influencing factor in the influencing factor set, and set the value of the jth influencing factor to Step S602: Obtain the adjacent detection record with the shortest time interval from the current shooting time in the detection log of the road, and extract the color value of the adjacent detection record as C ’ ; Extract the value of the jth influencing factor in the influencing factor set stored in the adjacent detection record as X j ′, the numerical difference of the jth influencing factor between two detection records in the detection log is According to the formula: Among them, m is the number of influencing factors in the influencing factor set, γ j is the influence coefficient of the jth influencing factor on the color value of the road surface photograph in the detection record; the theoretical characteristic value of the road surface photograph in the detection record is calculated to be C new ; According to the formula: Calculate the theoretical characteristic value T of the road surface photograph in the detection record new .

8. The method for detecting the road surface's true color based on deep learning according to claim 7, characterized in that: The step S700 includes the following steps: Step S701: Input the road photograph in the detection record into the color recognition model to obtain the color value of the road photograph as C′ new ; According to the formula: Calculate the actual characteristic value T' of the road surface photograph in the detection record new ; Step S702: Set the error range of the road set to (0, P), and calculate the error deviation degree of the road surface photos in the detection record to be P new =|T new -T′ new | / T′ new ;like Then an abnormal reminder is given for the road.

Images