Method and system for determining dirtiness of mud truck compartment based on artificial intelligence

Abstract

The present invention provides a method and system for determining the dirtiness of a mud truck compartment based on artificial intelligence, the method comprising: the color hue of the mud truck compartment and the soil color hue constitute the hue of a dirty pixel in the compartment image, the dirtiness of the dirty pixel is obtained according to the proportion of the soil color hue in the hue of the dirty pixel; the distribution concentration of the dirty pixel is obtained based on the distribution of pixels with the same dirtiness around each dirty pixel; the overall dirtiness of the mud truck compartment is determined based on the dirtiness and distribution concentration of each dirty pixel in the compartment image. The present invention determines the overall dirtiness of the mud truck compartment according to the dirtiness and distribution concentration of each pixel, comprehensively considers the influencing factors of the compartment dirtiness and quantifies them, and improves the rationality and accuracy of the judgment of the overall dirtiness of the compartment.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an artificial intelligence-based method and system for determining the dirtiness of a dump truck compartment. Background Art

[0002] At present, there are several methods for detecting the degree of dirtiness of vehicles, such as light reflection method and image subtraction combined with contour analysis. Among them, the light reflection method is greatly affected by external light, and requires the configuration of light source equipment and receiving equipment other than the camera, which is costly. The image subtraction combined with contour analysis method requires obtaining a clean vehicle image that is consistent with the model of the vehicle to be detected, which has great restrictions on the vehicle model and the implementation of the scheme. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a method for determining the dirtiness of a mud truck compartment based on artificial intelligence, the method comprising:

[0004] The color hue of the mud truck compartment and the soil color hue constitute the hue of the dirty pixel in the compartment image, and the degree of dirtiness of the dirty pixel is obtained according to the proportion of the soil color hue in the hue of the dirty pixel;

[0005] The distribution concentration of the dirty pixel points is obtained based on the distribution of the pixel points with the same degree of dirtiness around each dirty pixel point;

[0006] The overall dirtiness of the mud truck compartment is determined based on the dirtiness and distribution concentration of each dirty pixel in the compartment image.

[0007] Furthermore, the method for obtaining the proportion of the soil color hue is:

[0008] The proportions of the car color hue and the soil color hue in the hue of the dirty pixels are , ,Change , , obtain the gradient hue card, each position in the gradient hue card corresponds to a set of proportion values ;

[0009] Find the position of the dirty pixel hue in the gradient hue card, and the corresponding The value is the ratio of the soil color hue to the hue of the dirty pixel.

[0010] Further, and The value range is [0, 1].

[0011] Furthermore, the method for obtaining the distribution concentration degree of the dirty pixel points is: extracting the dirty pixel points of different dirt degrees respectively to obtain a pixel distribution map of each dirt degree;

[0012] Dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map, and the distribution concentration of dirty pixels is obtained according to the degree of dirtiness of the dirty pixels and the size of the dirty sub-region to which they belong.

[0013] Furthermore, the dirtiness level and distribution concentration level corresponding to each dirty pixel are multiplied and then summed to obtain the dirtiness level of the mud truck compartment.

[0014] The present invention also proposes a system for determining the dirtiness of a mud truck compartment based on artificial intelligence, the system comprising:

[0015] The first module is used to obtain the degree of dirtiness of the dirty pixel points according to the proportion of the soil color hue in the hue of the dirty pixel points; wherein the mud truck compartment color hue and the soil color hue constitute the hue of the dirty pixel points in the compartment image;

[0016] The second module is used to obtain the distribution concentration of dirty pixels based on the distribution of pixels with the same degree of dirtiness around each dirty pixel;

[0017] The third module is used to determine the overall dirtiness of the mud truck compartment based on the dirtiness and distribution concentration of each dirty pixel in the compartment image.

[0018] Furthermore, the steps for obtaining the proportion of the soil color hue are:

[0019] The proportions of the car color hue and the soil color hue in the hue of the dirty pixels are , ,Change , , obtain the gradient hue card, each position in the gradient hue card corresponds to a set of proportion values ;

[0020] Find the position of the dirty pixel hue in the gradient hue card, and the corresponding The value is the ratio of the soil color hue to the hue of the dirty pixel.

[0021] Further, and The value range is [0, 1].

[0022] Furthermore, the step of obtaining the distribution concentration degree of the dirty pixel points is: extracting the dirty pixel points of different dirt degrees respectively to obtain a pixel distribution map of each dirt degree;

[0023] Dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map, and the distribution concentration of dirty pixels is obtained according to the degree of dirtiness of the dirty pixels and the size of the dirty sub-region to which they belong.

[0024] Furthermore, the dirtiness level and distribution concentration level corresponding to each dirty pixel are multiplied and then summed to obtain the dirtiness level of the mud truck compartment.

[0025] The beneficial effects of the present invention are:

[0026] 1. The carriage color and the soil color are mixed together to present a dirty carriage effect. Therefore, the present invention obtains the degree of dirtiness of the dirty pixel based on the proportion of the soil color hue in the hue of the dirty pixel, removes the influence of the carriage color, and can obtain the accurate degree of dirtiness of the pixel.

[0027] 2. A corresponding gradient hue card can be constructed for vehicles of any color based on the present invention, thereby obtaining the degree of dirtiness of each dirty pixel point. Therefore, the present invention has high generalization ability. In addition, the present invention only needs to perform image acquisition based on an ordinary camera to obtain the overall degree of dirtiness of the vehicle compartment, which has low cost, few implementation restrictions, and strong practicality.

[0028] 3. The present invention determines the overall dirtiness of the dump truck compartment according to the dirtiness and distribution concentration of each pixel point, comprehensively considers the factors affecting the dirtiness of the compartment and quantifies them, thereby improving the rationality and accuracy of the judgment of the overall dirtiness of the compartment. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The present invention is a flowchart for implementing the method. DETAILED DESCRIPTION

[0030] In order to allow those skilled in the art to better understand the present invention, the present invention is described in detail below in conjunction with embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0031] Embodiment 1:

[0032] This embodiment provides a method for determining the dirtiness of a mud truck compartment based on artificial intelligence. The implementation process of the method is as follows: Figure 1 As shown, specifically, the method includes:

[0033] Step S1, the color hue of the mud truck compartment and the soil color hue constitute the hue of the dirty pixel in the compartment image, and the dirtiness of the dirty pixel is obtained according to the proportion of the soil color hue in the hue of the dirty pixel.

[0034] 1) Extract dirty pixels:

[0035] To obtain the image of the dump truck compartment, in order to reduce the interference caused by the viewing angle and the acquisition method during the image acquisition process and improve the accuracy of judgment, the present invention draws on the line scanning method to process the front view image of the side of the dump truck compartment. Specifically, the acquired image is pixel selected and image stitched to obtain a complete front view image of the side of the compartment. The main process is: the present invention places a camera on both sides of the road that the dump truck passes before entering the automatic car wash to collect the compartment image, and according to the vehicle's driving speed and the camera's sampling frequency, adjusts the number of columns of pixels selected in each frame of the image, so that the pixel selection of the compartment image is continuous and uninterrupted, and the information is not repeatedly stacked, and then the pixel information is spliced ​​column by column to obtain a complete compartment image.

[0036] In the embodiment, the pixels can be divided into two categories according to their colors. One category is the pixels of the carriage, reflective strips on the carriage, slogans, etc., which belong to the color of the vehicle itself and are free of dirt and mud; the other category is the dirty pixels on the carriage with different degrees of dirt and mud. Preferably, in the embodiment, a DNN network is used to classify the color of the pixels, that is, to extract the dirty pixels. The DNN network is an Encoder-Decoder structure, and its specific training process is as follows: the front view image of the side of the carriage collected constitutes a data set; the labels are 2 categories, non-dirty carriage pixels and dirty carriage pixels. This method is pixel-level classification, that is, all pixels in the carriage image are labeled with corresponding labels, and the value of the non-dirty carriage pixels is marked as 0, and the value of the dirty carriage pixels is marked as 1; the cross entropy loss function is used.

[0037] It should be noted that in order to reduce interference from other colors and improve the accuracy of the present invention, in the embodiment, the interfering colors such as reflective strips and slogans in the vehicle pixels that are different from the color of the car body are replaced with the color of the car body itself.

[0038] 2) Get the proportion of soil color hue in the hue of dirty pixels:

[0039] a) The proportions of the car color hue and the soil color hue in the hue of the dirty pixels are , ,Change , , obtain the gradient hue card, each position in the gradient hue card corresponds to a set of proportion values , and The value range of is [0, 1]; the gradient hue card can be a hue circle, a hue ring, or a hue belt. Preferably, in the embodiment, the hue circle is taken as an example to illustrate the method for obtaining the hue circle:

[0040] and The starting values ​​of are 0 and 1, that is, only the car color hue H1 is included, and the soil color, i.e. the khaki hue H2, is not included; change and The value of , that is, the hue of the dirty pixel includes both the car color hue and the soil color hue. Specifically, it gradually increases in a clockwise direction. , reduce , and always maintain , we get a gradient hue circle with a gradient hue. Specifically, and The value of and the clockwise angle about, that is , .

[0041] Therefore, each position point on the gradient hue circle corresponds to a set of proportion values , that is, the hue value corresponding to each position point is different, and the difference in hue value between position points comes from the proportion of the hue of the car color and the hue of the soil color and are different.

[0042] b) Find the position of the dirty pixel hue in the gradient hue card, and the corresponding The value is the proportion of the soil color hue in the hue of the dirty pixel. Specifically, for each dirty pixel extracted, its corresponding position in the gradient hue circle can be found according to its hue. and The value is the ratio of the car color hue and the soil color hue in the dirty pixel hue.

[0043] 3) In the embodiment The value indicates the degree of dirtiness of the dirty pixel. The larger the value, the greater the proportion of the soil color hue in the hue of the dirty pixel, the thicker the thickness of the soil at the dirty pixel, and the dirtier the pixel is.

[0044] It should be noted that the application of the present invention to obtain The premise is that the color of the dirty carriage can at least be vaguely seen in the dirty area of ​​the dirty carriage.

[0045] Step S2, obtaining the distribution concentration of the dirty pixel points based on the distribution of the pixel points with the same dirtiness level around each dirty pixel point.

[0046] The method for obtaining the distribution concentration of dirty pixels is specifically as follows: dirty pixels of different dirtiness levels are extracted respectively to obtain a pixel distribution map at each dirtiness level; dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map, and the distribution concentration of dirty pixels is obtained according to the dirtiness of the dirty pixels and the size of the dirty sub-region to which they belong.

[0047] The method for obtaining the pixel distribution map at each dirtiness level in the embodiment is as follows:

[0048] Will All dirty pixels are considered as one class, with a dirtiness level of 1, and the pixel values ​​that do not belong to this level in the original carriage image are set to 0 to obtain a pixel distribution map ;Will All dirty pixels are considered as one class, with a dirtiness level of 2, and the pixel values ​​that do not belong to this level in the original carriage image are set to 0 to obtain a pixel distribution map ; And so on, we get ten pixel distribution maps.

[0049] In the embodiment, the dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map. Specifically, the pixel distribution map middle, The value range of is [1,10], and the method of dividing the dirty sub-area is:

[0050] In the pixel distribution map In the method, each dirty pixel is taken as the center point of a window of preset size, preferably, the window size is 3*3, the number of dirty pixels in the window is obtained, and the window with the largest number of dirty pixels is the original window. If there are multiple windows with the largest number of dirty pixels, each window with the largest number of dirty pixels is a pre-selected window, and the sum of the distances between the center point of the pre-selected window and the center points of other windows is obtained. The pre-selected window with the smallest sum of distances is the final original window.

[0051] Take the original window as the center window, find the four neighboring windows of the center window, and get the number of dirty pixels in each neighboring window. ,like , then The neighborhood window of is used as the new center window, and the four neighborhood windows of the new center window are obtained again. If the number of dirty pixels in the neighborhood window is , then The neighborhood window of the new center window is used as the new center window, and the above steps are repeated until the number of dirty pixels in all the neighborhood windows of the new center window is less than or equal to 5. The window and the original window constitute a dirty sub-area.

[0052] In the pixel distribution map After removing the dirty pixels included in all the obtained dirty sub-regions, the original window is re-determined, and a new dirty sub-region is obtained according to the above steps; if the number of dirty pixels in the four neighborhood windows of the original window is less than or equal to 5, the area corresponding to the original window is a dirty sub-region.

[0053] In the embodiment, the distribution concentration of the dirty pixel points is obtained according to the degree of dirtiness of the dirty pixel points and the size of the dirty sub-region to which they belong, specifically:

[0054] For each dirty pixel, based on the degree of dirtiness corresponding to the dirty pixel , get the pixel distribution map corresponding to the dirty pixel , and pixel distribution map The size of the dirty sub-region including the dirty pixel point is preferably the number of windows in the dirty sub-region. Characterize the size of the dirty sub-region ,but is the distribution concentration of the dirty pixel .

[0055] Step S3, based on the dirtiness and distribution concentration of each dirty pixel in the compartment image, determine the overall dirtiness of the dump truck compartment.

[0056] right After normalization, we get Specifically, the dirtiness level and distribution concentration level corresponding to each dirty pixel are multiplied and then summed to obtain the overall dirtiness level of the mud truck compartment: , The overall dirtiness of the mud truck compartment. , They are the first The degree of dirtiness and distribution concentration of dirty pixels. is the total number of dirty pixels in the carriage image, is the total number of pixels in the carriage image.

[0057] Step S4, determining the cleaning intensity according to the overall dirtiness of the dump truck compartment.

[0058] For an automatic washing room, the maximum water output and maximum water gun pressure is constant, so it is used as the basis for adjusting the cleaning intensity. The value range of is (0,1). The larger the value, the higher the overall dirtiness of the carriage, and the greater the total amount of water and water gun pressure required. Specifically, the total amount of water required to clean the mud truck carriage is , the water gun pressure is , to achieve appropriate intensity cleaning according to the overall dirtiness of the mud truck compartment.

[0059] Embodiment 2:

[0060] Based on the same inventive concept as the above method embodiment, this embodiment provides a system for determining the dirtiness of a mud truck compartment based on artificial intelligence. Specifically, the system includes:

[0061] The first module is used to obtain the degree of dirtiness of the dirty pixel points according to the proportion of the soil color hue in the hue of the dirty pixel points; wherein the mud truck compartment color hue and the soil color hue constitute the hue of the dirty pixel points in the compartment image;

[0062] The second module is used to obtain the distribution concentration of dirty pixels based on the distribution of pixels with the same degree of dirtiness around each dirty pixel;

[0063] The third module is used to determine the overall dirtiness of the mud truck compartment based on the dirtiness and distribution concentration of each dirty pixel in the compartment image.

[0064] Regarding the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment; the above description is intended to enable those skilled in the art to better understand the present invention, and is not intended to limit the present invention. All modifications and changes made within the spirit and principles of the present invention are within the protection scope of the present invention.

Claims

1. A method for determining the dirtiness of a mud truck compartment based on artificial intelligence, characterized in that: The method includes: The color hue of the mud truck compartment and the soil color hue constitute the hue of the dirty pixel in the compartment image, and the degree of dirtiness of the dirty pixel is obtained according to the proportion of the soil color hue in the hue of the dirty pixel; wherein, the method for obtaining the proportion of the soil color hue is: The proportions of the car color hue and the soil color hue in the hue of the dirty pixel are α and β respectively. The values ​​of α and β are changed to obtain a gradient hue card, in which each position in the gradient hue card corresponds to a set of proportion values ​​(α, β); Find the position of the dirty pixel hue in the gradient hue card. The β value corresponding to this position is the proportion of the soil color hue in the dirty pixel hue. The distribution concentration of the dirty pixel points is obtained based on the distribution of the pixel points with the same degree of dirtiness around each dirty pixel point; Based on the dirtiness and distribution concentration of each dirty pixel in the carriage image, the overall dirtiness of the mud truck carriage is determined; The value range of α and β is [0, 1].

2. The method according to claim 1, characterized in that The method for obtaining the distribution concentration degree of the dirty pixel points is: respectively extracting the dirty pixel points of different dirtiness degrees to obtain a pixel distribution map of each dirtiness degree; Dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map, and the distribution concentration of dirty pixels is obtained according to the degree of dirtiness of the dirty pixels and the size of the dirty sub-region to which they belong.

3. The method according to claim 2, characterized in that The dirtiness level of the dump truck compartment is obtained by multiplying the dirtiness level and distribution concentration level corresponding to each dirty pixel point and then summing them up.

4. A system for determining the dirtiness of a mud truck compartment based on artificial intelligence, characterized in that: The system includes: The first module is used to obtain the degree of dirtiness of the dirty pixel points according to the proportion of the soil color hue in the hue of the dirty pixel points; wherein the mud truck compartment color hue and the soil color hue constitute the hue of the dirty pixel points in the compartment image; wherein the proportion of the soil color hue is obtained specifically as follows: The proportions of the car color hue and the soil color hue in the hue of the dirty pixel are α and β respectively. The values ​​of α and β are changed to obtain a gradient hue card, in which each position in the gradient hue card corresponds to a set of proportion values ​​(α, β); Find the position of the dirty pixel hue in the gradient hue card. The β value corresponding to this position is the proportion of the soil color hue in the dirty pixel hue. The second module is used to obtain the distribution concentration of dirty pixels based on the distribution of pixels with the same degree of dirtiness around each dirty pixel; The third module is used to determine the overall dirtiness of the mud truck compartment based on the dirtiness and distribution concentration of each dirty pixel in the compartment image; The value range of α and β is [0, 1].

5. The system according to claim 4, characterized in that The step of obtaining the distribution concentration degree of the dirty pixel points is: extracting the dirty pixel points of different dirt degrees respectively to obtain a pixel distribution map of each dirt degree; Dirty sub-regions are divided according to the distribution of dirty pixels in the pixel distribution map, and the distribution concentration of dirty pixels is obtained according to the degree of dirtiness of the dirty pixels and the size of the dirty sub-region to which they belong.

6. The system according to claim 5, characterized in that The dirtiness level of the dump truck compartment is obtained by multiplying the dirtiness level and distribution concentration level corresponding to each dirty pixel point and then summing them up.

Images