CN117974492B - Optimal control method and system for intelligent milling machine - Google Patents
Optimal control method and system for intelligent milling machine Download PDFInfo
- Publication number
- CN117974492B CN117974492B CN202410361430.XA CN202410361430A CN117974492B CN 117974492 B CN117974492 B CN 117974492B CN 202410361430 A CN202410361430 A CN 202410361430A CN 117974492 B CN117974492 B CN 117974492B
- Authority
- CN
- China
- Prior art keywords
- window
- pavement
- pixel point
- degree
- neighborhood
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000003801 milling Methods 0.000 title claims abstract description 63
- 238000001914 filtration Methods 0.000 claims abstract description 53
- 238000005457 optimization Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims description 35
- 239000011159 matrix material Substances 0.000 claims description 23
- 230000008569 process Effects 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 238000003064 k means clustering Methods 0.000 claims description 4
- 238000007621 cluster analysis Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 25
- 238000005259 measurement Methods 0.000 description 8
- 238000010276 construction Methods 0.000 description 6
- 238000009499 grossing Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to the technical field of filter image enhancement, in particular to an intelligent milling machine optimization control method and system, wherein each local image block is divided by combining the position distribution of point cloud data points corresponding to a pavement gray level image; combining the gray gradient distribution in the local image block and the road elevation value of the corresponding point cloud data point to obtain the noise interference degree; determining the length of a search window according to the area and texture characteristics of the local image block; combining the noise interference degree and the obvious texture characteristic degree to obtain the length of a neighborhood window; on the basis of the search window and the neighborhood window, the road surface denoising image with better denoising effect is obtained according to the noise interference degree, the obvious texture characteristic degree and the gray value distribution by a non-local mean value filtering method, so that the travelling speed of the milling machine is more accurately optimized and adjusted based on the road surface denoising image.
Description
Technical Field
The invention relates to the technical field of filter image enhancement, in particular to an intelligent milling machine optimization control method and system.
Background
A milling machine is one type of road maintenance equipment for removing a damaged layer of a road surface in order to repair and re-lay the road. When the road maintenance is carried out by the milling machine, the reasonable travelling speed can improve the construction efficiency while ensuring the construction quality; however, the road surface damage to different areas on the road is different, namely the road surface quality is different, so that different road surfaces need different construction time to ensure the construction quality; therefore, in order to ensure the construction quality and the construction efficiency, the milling machine needs to adjust different travelling speeds on the road surfaces with different road surface quality indexes; i.e. the milling machine needs to select the travelling speed according to the quality of the advancing road surface.
The specific condition of the advancing road surface can be directly and clearly reflected through the image, so that the road surface image is usually acquired in the advancing process of the milling machine; and inputting the road surface image into a trained deep learning model, outputting the required road surface quality, and controlling the travelling speed of the milling machine according to the road surface quality. However, due to the influence of the working environment of the milling machine and the acquisition equipment, the image acquired by the milling machine in the running process is interfered by noise, so that in order to ensure the accuracy of the road surface quality obtained by the deep learning model output, the acquired road surface image needs to be subjected to denoising pretreatment. In the prior art, a non-local mean value filtering method is generally adopted to carry out filtering treatment on the acquired pavement gray level image; however, the filtering effect of the non-local mean filtering method on the image is greatly influenced by the prior parameters selected subjectively, namely the length of the search window and the length of the neighborhood window; the pavement quality condition is complex, the denoising effect of different areas in the pavement image cannot be guaranteed due to the selection of a single priori parameter in the non-local mean value filtering method, and the denoising effect of the obtained pavement denoising image is poor, namely the denoising effect of the pavement gray level image of the advancing pavement in the advancing process of the milling machine in the prior art is poor through the non-local mean value filtering method, so that the accuracy of adjusting the advancing speed of the milling machine based on the denoised pavement denoising image is low.
Disclosure of Invention
In order to solve the technical problems that the denoising effect of a pavement gray image of a forward pavement in the advancing process of a milling machine is poor by a non-local mean filtering method in the prior art, so that the optimizing and adjusting effect of the advancing speed of the milling machine is poor based on the denoised pavement denoising image, the invention aims to provide an intelligent milling machine optimizing and controlling method and system, and the adopted technical scheme is as follows:
the invention provides an intelligent milling machine optimal control method, which comprises the following steps:
acquiring a pavement gray image and a corresponding point cloud data matrix of each advancing pavement in the advancing process of the milling machine; each point cloud data point in the point cloud data matrix corresponds to each pixel point in the pavement gray level image;
Dividing the pavement gray level image into at least two local image blocks according to the spatial position distribution condition of the point cloud data points in the point cloud data matrix; obtaining the obvious degree of the texture characteristics of each local image block according to the gray gradient distribution condition in each local image block; obtaining the noise interference degree of the pavement gray level image according to the pavement elevation value distribution concentration condition of the point cloud data points and the texture characteristic obvious degree;
obtaining the search window length of the pavement gray image according to the area of each local image block and the obvious degree of the texture characteristics; obtaining the neighborhood window length of each pixel point according to the search window length, the noise interference degree and the obvious degree of the texture characteristics of the local image block where each pixel point is located;
In a search window where each pixel point is located, based on a non-local mean value filtering method, combining the noise interference degree according to the neighborhood window of each pixel point and the similarity degree of the gray distribution condition, the obvious degree of texture features and the road surface elevation value distribution condition between each neighborhood window obtained by sliding traversal, and obtaining the final filtering gray value of each pixel point;
Obtaining a pavement denoising image according to the final filtering gray value of each pixel point in the pavement gray image; and optimally adjusting the travelling speed of the milling machine according to the pavement denoising image.
Further, the method for acquiring the local image block includes:
In the pavement gray level image, carrying out cluster analysis according to pixel coordinates of all point cloud data points in a point cloud data matrix through a k-means clustering algorithm to obtain at least two point cloud data point cluster clusters; and clustering each point cloud data point into a corresponding image area in the pavement gray level image to serve as a local image block.
Further, the method for obtaining the obviously degree of the texture features comprises the following steps:
in each local image block, taking the distance between each pixel point and the centroid of the local image block as the reference distance of each pixel point;
Taking the product of the normalized value of the gray gradient value of each pixel point and the reference distance as a local texture characteristic value of each pixel point;
And taking the normalized value of the average value of the local texture characteristic values of all pixel points in each local image block as the texture characteristic obvious degree of each local image block.
Further, the method for acquiring the noise interference degree comprises the following steps:
Taking the normalized value of the variance of the elevation values of the road surfaces of all the point cloud data points in each local image block as the road surface damage degree of each local image block; taking the product of the negative correlation mapping value of the pavement damage degree and the obvious degree of the texture feature as the texture feature parameter of each local image block; and taking the average value of the texture characteristic parameters of all the local image blocks as the noise interference degree of the pavement gray level image.
Further, the method for acquiring the search window length includes:
Taking the product of the arithmetic square root of the area of each local image block and the saliency of the texture feature as the reference window length of each local image block; and (3) taking the average value of the reference window lengths of all the local image blocks to be an odd integer upwards to obtain the search window length of the pavement gray level image.
Further, the method for obtaining the length of the neighborhood window comprises the following steps:
And taking the product of the search window length, the noise interference degree and the obvious degree of the texture feature of the local image block where each pixel point is positioned to be an odd integer upwards to obtain the neighborhood window length of each pixel point.
Further, the method for obtaining the final filtered gray value includes:
sequentially taking each pixel point as a target pixel point; taking the search window of the target pixel point as a target search window; taking a neighborhood window taking a target pixel point as a center as a target neighborhood window;
Traversing the target search window by a sliding window with the same shape and size as the target neighborhood window to obtain at least two comparison neighborhood windows; taking the pixel point corresponding to the central position of each contrast neighborhood window as a contrast pixel point; arranging all pixel points in the target neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence of the target neighborhood window; arranging all the pixel points in each comparison neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence corresponding to each comparison neighborhood window;
Taking the noise interference degree as a weight, and constructing a window similarity calculation model according to the pixel point reference sequence between each comparison neighborhood window and the target neighborhood window and the similarity on the gray level distribution condition, the texture feature obvious degree and the road surface elevation value distribution condition; obtaining the window similarity of each comparison neighborhood window according to the window similarity calculation model;
Taking the ratio between the window similarity of each comparison neighborhood window and the accumulated value of the window similarity of all the comparison neighborhood windows as the filtering weight of each comparison neighborhood window; taking the product of the filtering weight and the gray value of the contrast pixel point of each contrast neighborhood window as the reference weighted gray value of each contrast neighborhood window; and taking the accumulated value of the reference weighted gray values of all the comparison neighborhood windows as the final filtered gray value of the target pixel point.
Further, the method for optimally adjusting the traveling speed of the milling machine according to the pavement denoising image comprises the following steps:
Inputting the pavement denoising image into a trained convolutional neural network, and outputting pavement quality indexes of each advancing pavement in the advancing process of the milling machine; and adjusting the travelling speed of the milling machine according to the road surface quality index.
Further, the window similarity calculation model includes:
Wherein, For target neighborhood window/>Corresponding/>Window similarity of each comparison neighborhood window; /(I)Noise interference degree of the pavement gray level image; /(I)For target neighborhood window/>And/>The number of pixels in the pixel reference sequence corresponding to the neighborhood windows is compared; /(I)For target neighborhood window/>Gray values and the/>, of all pixels in a pixel reference sequenceMean square error of gray values of all pixels in the pixel point reference sequence of each contrast neighborhood window; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>The obvious degree of the texture characteristics of the local image block where each pixel point is positioned; for/> The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowThe obvious degree of the texture characteristics of the local image block where each pixel point is positioned; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>Road elevation values of point cloud data points corresponding to the pixel points; /(I)For/>The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowRoad elevation values of point cloud data points corresponding to the pixel points; /(I)A function is selected for the maximum.
The invention also provides an intelligent milling machine optimization control system, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes any one of the steps of the intelligent milling machine optimization control method when executing the computer program.
The invention has the following beneficial effects:
According to the filtering principle of the non-local mean filtering method, the difference of the size of the search window and the size of the neighborhood window can influence the smoothing effect and the denoising effect of the filtered image; the complexity of the road surface quality can cause different texture details of different areas of the road surface gray level image, and the different road surface gray level images are influenced by noise, so that the required smoothing effect and denoising effect are different, and therefore, if the effect of non-local mean value filtering is to be ensured, when the filtering gray level value calculation is carried out on the pixel points of different areas, search windows and neighborhood windows with different sizes are required to be set according to the texture details of corresponding image areas and the influence degree of noise on the whole image; i.e. the noise disturbance level of the road surface grey scale image needs to be calculated first and the texture detail characteristics of different image areas.
For a normal road surface without road surface damage, the corresponding road surface gray level image is relatively uniform in low frequency, and under the condition of noise interference, a part of high frequency area appears in the image; therefore, for a normal road surface, the degree of noise interference of the image can be measured by counting the duty ratio of a high-frequency part in the image; that is, for the local image block belonging to the normal road surface, the corresponding apparent degree of the texture features is generated due to the interference of the noise, and the calculation of the noise interference degree can be performed through the corresponding apparent degree of the texture features. However, in the case of a road surface gray-scale image in which there is a road surface damage, since the road surface corresponding to the road surface damaged area is uneven, the area in which shadow textures appear is relatively large, and the corresponding road surface damaged area also appears as a high frequency; therefore, for the local image blocks belonging to the road surface damage area, the more serious the corresponding road surface damage is, the lower the reliability of noise interference degree representation is through the obvious degree of texture characteristics; therefore, when the influence degree of noise interference on the image is measured, the obvious degree of texture features of different areas needs to be distinguished for analysis.
To calculate texture detail features of different image areas, firstly, dividing the image areas is needed; because the road surface areas with different damage degrees are subjected to different noise interference measurement, and road surface damage is reflected in the point cloud data matrix and is expressed as the spatial position distribution of the point cloud data points, for example, the height of different point cloud data points can be influenced by the road surface with different concave-convex shapes, if clustering is performed based on the spatial positions of the point cloud data points, and because each point cloud data point corresponds to the characteristic of one pixel point, namely, the image blocks are divided according to the clustering result, the divided image blocks have different texture detail characteristics, namely, the damage degrees are different. According to the space position distribution condition of the point cloud data points in the point cloud data matrix, the pavement gray level image is divided into at least two local image blocks; texture detail features for each local image block are further computed.
In the pavement gray level image, the distribution of noise pixels is random, but normal image textures generally represent pavement textures or pavement damage areas, and the positions of the pixels are concentrated; therefore, the more discrete the position distribution of the pixel points with gradients in the image blocks is, and the more obvious the gradient characteristics are, the more obvious the block texture characteristics aiming at noise analysis are; therefore, for each local image block, the measurement of the texture detail characteristics can be carried out according to the gray gradient distribution condition in each image block, and the subsequent calculation of the noise influence degree is convenient. Therefore, the invention obtains the obvious degree of the texture feature of each local image block according to the gray gradient distribution condition in each local image block; and thus the noise interference degree is calculated by combining the texture feature obvious degree of the local image block.
For each local image block, the corresponding road surface damage degree is different, and the lower the road surface damage degree is, the lower the corresponding real texture is in the region without considering the influence of noise, the higher the reliability of the calculated texture feature obvious degree representation of the interference degree of noise is; the more the corresponding texture is, the lower the reliability of the calculated characteristic obvious degree representation of the texture is, which is affected by noise, in the region with higher pavement damage degree; the noise disturbance level can be further calculated by weighting the apparent level of the texture feature by the road surface damage level of each partial image block. For each local image block, the more gentle the corresponding road surface is, the smaller the damage degree is, namely the smaller the difference of the road surface elevation values of the whole point cloud data points in the road surface elevation values is; conversely, the more uneven the road surface is, the greater the damage degree is generally, namely the greater the difference of the whole road surface elevation values of the whole point cloud data points in the road surface is; therefore, the weight can be carried out according to the reliability of the overall difference of the road elevation values as the obvious degree of the texture features, and the noise interference degree is calculated; therefore, the noise interference degree of the pavement gray level image is obtained according to the pavement elevation value distribution concentration condition and the texture characteristic obvious degree of the point cloud data points.
Further, the length of the neighborhood window and the length of the search window are calculated according to the noise interference degree and the obvious texture characteristic degree; because the same local image block represents the area with similar pavement elevation values, if the search window contains the same local image block of the image as far as possible, the neighborhood window in the search window has more referential property in the process of comparison; the present invention performs calculation of the search window length by combining the areas of the respective partial image blocks. In addition, as the neighborhood window with more complicated textures needs more image information for comparison, the accuracy of filtering is improved; therefore, the invention combines the obvious degree of the texture characteristic on the basis of the area of the local image block to calculate the length of the search window. Further, when the length of the neighborhood window is calculated, the comparison calculation of the neighborhood window with more details is needed for the corresponding pixel point when the noise interference is stronger and the texture characteristics are more obvious, so that the more ideal noise reduction effect is obtained; and considering that the neighborhood window needs to be limited in the search window, the neighborhood window length of each pixel point is obtained according to the search window length, the noise interference degree and the obvious degree of the texture characteristics of the local image block where each pixel point is positioned.
After the search window length and the neighborhood window length are obtained, further carrying out self-adaptive calculation of final filtering gray values on each pixel point by combining non-local mean filtering; according to the principle of a non-local mean value filtering algorithm, similarity calculation is needed to be performed between a neighborhood window of each pixel point and each neighborhood window in a corresponding search window; the traditional similarity calculation method is to calculate the mean square error between the neighborhood windows in gray scale; however, for the damaged road image processed by the milling machine, different damaged areas usually show different image features, for example, a normal road area has less texture information compared with the damaged road area, but due to random distribution of noise, the noise points in the normal road are compared with the dense areas in a similarity manner, so that the weight of the damaged area is increased, and the filtering result is inaccurate. When the image is less interfered by noise, the traditional similarity measurement method based on the mean square error can truly obtain a relatively ideal calculation result; however, when the image is greatly interfered by noise, according to the calculation method of the length of the neighborhood window, the pixel points of different local image blocks can appear in the corresponding neighborhood window; and different image blocks are influenced by larger noise, so that the image characteristics of the neighborhood windows of the calculated pixel points are possibly more similar to those of other image blocks, and the similarity weight of the neighborhood windows which do not belong to the same local image block is indirectly improved, so that the accuracy of similarity measurement is influenced; because the elevation values in the same local image block are relatively close, the similarity weight of the neighborhood windows which do not belong to the same local image block is reduced; under the condition of large noise interference, the measurement index of the similarity should be more biased to the area with similar elevation values; therefore, according to the characteristics, in the search window where each pixel is located, based on a non-local mean value filtering method, the more accurate final filtering gray value corresponding to each pixel is obtained according to the gray distribution condition, the obvious degree of texture features and the similarity degree of the road surface elevation value distribution condition between the neighborhood window of each pixel and each neighborhood window obtained by sliding traversal by combining the noise interference degree. Therefore, a pavement denoising image with better denoising effect is further obtained according to the final filtering gray value, and the travelling speed of the milling machine is more accurately optimized and adjusted based on the pavement denoising image.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of an optimization control method for an intelligent milling machine according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a pavement gray scale image according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following description refers to the specific implementation, structure, characteristics and effects of an intelligent milling machine optimization control method and system according to the invention in combination with the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the intelligent milling machine optimization control method and system provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an optimization control method of an intelligent milling machine according to an embodiment of the invention is shown, where the method includes:
step S1: acquiring a pavement gray image and a corresponding point cloud data matrix of each advancing pavement in the advancing process of the milling machine; each point cloud data point in the point cloud data matrix corresponds to each pixel point in the pavement gray scale image.
The embodiment of the invention aims to provide an intelligent milling machine optimization control method which is used for carrying out denoising treatment on a pavement gray level image according to the pavement gray level image of each advancing pavement and a corresponding point cloud data matrix in the advancing process of the milling machine, so as to obtain a pavement denoising image with better denoising effect, namely more accurate pavement denoising effect; therefore, the travelling speed of the milling machine is more accurately optimally adjusted according to the pavement denoising image.
Therefore, the embodiment of the invention acquires the pavement gray level image and the corresponding point cloud data matrix of each advancing pavement in the advancing process of the milling machine; each point cloud data point in the point cloud data matrix corresponds to each pixel point in the pavement gray scale image. According to the embodiment of the invention, a CCD camera is carried on a milling machine to collect road surface images, and the collected images are grayed to obtain road surface gray level images required by the embodiment of the invention; simultaneously carrying an ultrasonic sensor to acquire three-dimensional point cloud data corresponding to the pavement gray level image; in order to enable each point cloud data point to correspond to each pixel point in the road surface image, interpolation processing is performed when the point cloud data precision is smaller than the resolution of the acquired road surface image, voxel downsampling is performed when the point cloud data precision is larger than the resolution of the acquired road surface image, and each point cloud data point can only correspond to one pixel point in the obtained point cloud data matrix, namely the point cloud data matrix can be perfectly mapped to the road surface image.
For point cloud data, each point cloud data corresponds to a relative height; that is, with respect to the height of the lowest point of the road surface, the embodiment of the invention uses the height difference of each point cloud data point from the lowest point of the advancing road surface corresponding to the point cloud data matrix as the road surface elevation value of each point cloud data point, thereby facilitating the subsequent analysis of the damaged road surface area. It should be noted that, the implementer may also collect the road surface image through other image collecting devices according to the specific implementation environment; and the road surface elevation value can be represented by other methods, for example, the elevation difference of each point cloud data point from the highest point of the advancing road surface corresponding to the point cloud data matrix is used as the road surface elevation value of each point cloud data point; or the distance between each point cloud data point and the ultrasonic sensor is used as a road elevation value and the like; however, the heights of the point cloud data points need to be represented by the road elevation values, so that the road elevation values corresponding to the point cloud data points on the same road elevation are the same, and further description is omitted here. Referring to fig. 2, a schematic diagram of a pavement gray image according to an embodiment of the present invention is shown, in fig. 2, noise pixels are widely distributed and have obvious texture salient features in a pavement normal area; in the damaged road surface region, the texture features of the noise pixels are mixed by the damaged road surface texture itself.
Step S2: dividing a pavement gray level image into at least two local image blocks according to the spatial position distribution condition of point cloud data points in a point cloud data matrix; obtaining the obvious degree of the texture characteristics of each local image block according to the gray gradient distribution condition in each local image block; and obtaining the noise interference degree of the pavement gray level image according to the pavement elevation value distribution concentration condition and the texture characteristic obvious degree of the point cloud data points.
According to the filtering principle of the non-local mean filtering method, the difference of the size of the search window and the size of the neighborhood window can influence the smoothing effect and the denoising effect of the filtered image; the complexity of the road surface quality can cause different texture details of different areas of the road surface gray level image, and the different road surface gray level images are influenced by noise, so that the required smoothing effect and denoising effect are different, and therefore, if the effect of non-local mean value filtering is to be ensured, when the filtering gray level value calculation is carried out on the pixel points of different areas, search windows and neighborhood windows with different sizes are required to be set according to the texture details of corresponding image areas and the influence degree of noise on the whole image; i.e. the noise disturbance level of the road surface grey scale image needs to be calculated first and the texture detail characteristics of different image areas.
For a normal road surface without road surface damage, the corresponding road surface gray level image is relatively uniform in low frequency, and under the condition of noise interference, a part of high frequency area appears in the image; therefore, for a normal road surface, the degree of noise interference of the image can be measured by counting the duty ratio of a high-frequency part in the image; that is, for the local image block belonging to the normal road surface, the corresponding apparent degree of the texture features is generated due to the interference of the noise, and the calculation of the noise interference degree can be performed through the corresponding apparent degree of the texture features. However, in the case of a road surface gray-scale image in which there is a road surface damage, since the road surface corresponding to the road surface damaged area is uneven, the area in which shadow textures appear is relatively large, and the corresponding road surface damaged area also appears as a high frequency; therefore, for the local image blocks belonging to the road surface damage area, the more serious the corresponding road surface damage is, the lower the reliability of noise interference degree representation is through the obvious degree of texture characteristics; therefore, when the influence degree of noise interference on the image is measured, the obvious degree of texture features of different areas needs to be distinguished for analysis.
To calculate texture detail features of different image areas, firstly, dividing the image areas is needed; because the road surface areas with different damage degrees are subjected to different noise interference measurement, and road surface damage is reflected in the point cloud data matrix and is expressed as the spatial position distribution of the point cloud data points, for example, the height of different point cloud data points can be influenced by the road surface with different concave-convex shapes, if clustering is performed based on the spatial positions of the point cloud data points, and because each point cloud data point corresponds to the characteristic of one pixel point, namely, the image blocks are divided according to the clustering result, the divided image blocks have different texture detail characteristics, namely, the damage degrees are different. Therefore, the embodiment of the invention divides the pavement gray level image into at least two local image blocks according to the spatial position distribution condition of the point cloud data points in the point cloud data matrix.
Preferably, the method for acquiring the local image block includes:
When the clustering is carried out based on the space positions of the point cloud data points, the divided image blocks can have different texture detail characteristics, so that in the pavement gray level image, the clustering analysis is carried out according to the pixel coordinates of all the point cloud data points in the point cloud data matrix through a k-means clustering algorithm to obtain at least two point cloud data point clustering clusters; and clustering each point cloud data point into a corresponding image area in the pavement gray level image to serve as a local image block. Because each point cloud data point corresponds to one pixel point, each point cloud data point cluster corresponds to one pixel point set in the corresponding pavement gray level image, and the image area corresponding to the pixel point set is a local image block. In the embodiment of the present invention, the number of clusters corresponding to the k-means clustering algorithm is obtained by an elbow method, and the elbow method is a prior art well known to those skilled in the art, and is not further limited and described herein.
In the pavement gray level image, the distribution of noise pixels is random, but normal image textures generally represent pavement textures or pavement damage areas, and the positions of the pixels are concentrated; therefore, the more discrete the position distribution of the pixel points with gradients in the image blocks is, and the more obvious the gradient characteristics are, the more obvious the block texture characteristics aiming at noise analysis are; therefore, for each local image block, the measurement of the texture detail characteristics can be carried out according to the gray gradient distribution condition in each image block, and the subsequent calculation of the noise influence degree is convenient. Therefore, the embodiment of the invention obtains the obvious degree of the texture characteristics of each local image block according to the gray gradient distribution condition in each local image block.
Preferably, the method for obtaining the saliency of the texture features comprises the following steps:
In the local image block, larger gray gradient values generally correspond to more obvious textures, and if the more obvious texture distribution is also more discrete, the texture characteristics of the corresponding local image block are relatively more obvious; therefore, if the pixel point with larger gray gradient value in the local image block is deviated from the centroid, the corresponding texture feature obviously degree is larger; therefore, in each local image block, the embodiment of the invention takes the distance between each pixel point and the mass center of the local image block as the reference distance of each pixel point; measuring the deviation degree of each pixel point relative to the centroid by a reference distance; taking the product of the normalized value of the gray gradient value of each pixel and the reference distance as the local texture characteristic value of each pixel; and taking the normalized value of the average value of the local texture characteristic values of all pixel points in each local image block as the texture characteristic obvious degree of each local image block.
In the embodiment of the invention, each local image block is taken as the first image block in turnLocal image block, thenThe method for calculating the texture feature significance of each local image block is expressed as the following formula:
Wherein, For/>Texture feature clarity of individual local image blocks,/>For/>Number of pixels in each local image block,/>For/>First/>, in a local image blockGray gradient values of the individual pixels; /(I)For/>Maximum value of gray gradient values in each local image block; /(I)For/>First/>, in a local image blockA reference distance of the individual pixel points; Is a normalization function; /(I) For/>First/>, in a local image blockNormalized values of gray gradient values of the individual pixel points; /(I)For/>First/>, in a local image blockLocal texture feature values for individual pixels.
For each local image block, the corresponding road surface damage degree is different, and the lower the road surface damage degree is, the lower the corresponding real texture is in the region without considering the influence of noise, the higher the reliability of the calculated texture feature obvious degree representation of the interference degree of noise is; the more the corresponding texture is, the lower the reliability of the calculated characteristic obvious degree representation of the texture is, which is affected by noise, in the region with higher pavement damage degree; that is, the lower the damage degree of the pavement is, the more the calculated obvious degree of the texture features can be used for representing the interference degree of noise; the noise disturbance level can be further calculated by weighting the apparent level of the texture feature by the road surface damage level of each partial image block. For each local image block, the more gentle the corresponding road surface is, the smaller the damage degree is, namely the smaller the difference of the road surface elevation values of the whole point cloud data points in the road surface elevation values is; conversely, the more uneven the road surface is, the greater the damage degree is generally, namely the greater the difference of the whole road surface elevation values of the whole point cloud data points in the road surface is; therefore, the weight can be carried out according to the reliability of the overall difference of the road elevation values as the obvious degree of the texture features, and the noise interference degree is calculated; therefore, according to the road elevation value distribution concentration condition and the texture characteristic obvious degree of the point cloud data points, the noise interference degree of the road gray image is obtained.
Preferably, the method for acquiring the noise interference level includes:
The more gentle road surface, namely the road surface with lower damage degree, the smaller the difference of the whole road surface elevation values of the whole corresponding point cloud data points; the more uneven road surface, namely the road surface with higher damage degree, the larger the difference of the road surface elevation values of the whole corresponding point cloud data points, so the embodiment of the invention takes the normalized value of the variance of the road surface elevation values of all the point cloud data points in each local image block as the road surface damage degree of each local image block.
Because the reliability of the corresponding texture feature obvious degree representing the noise interference degree is higher when the pavement damage degree is lower, the embodiment of the invention takes the product of the negative correlation mapping value of the pavement damage degree and the texture feature obvious degree as the texture feature parameter of each local image block. And further combining the texture characteristic parameters of all the local image blocks to obtain the noise interference degree of the whole pavement gray image, so that the embodiment of the invention takes the average value of the texture characteristic parameters of all the local image blocks as the noise interference degree of the pavement gray image.
In the embodiment of the invention, the method for acquiring the noise interference degree of the pavement gray image is expressed as the following formula:
Wherein, Noise interference degree of the pavement gray level image; /(I)The number of the local image blocks in the pavement gray level image is the number of the local image blocks in the pavement gray level image; /(I)Is the/>, in the pavement gray level imageThe degree of texture feature clarity of the individual local image blocks; /(I)Is the/>, in the pavement gray level imageVariance of road elevation values of all point cloud data points in each local image block,/>Is the/>, in the pavement gray level imageMaximum variance of road elevation values of all point cloud data points in each local image block,/>In order to preset the adjustment parameters, the embodiment of the invention sets the preset adjustment parameters to 1 for preventing the denominator from being 0, and an implementer can adjust the preset adjustment parameters according to the specific implementation environment; /(I)Is the/>, in the pavement gray level imageThe pavement damage degree of each local image block, namely the normalization is carried out by a maximum value method in the embodiment of the invention, and an implementer can also adopt other normalization methods, such as linear normalization and the like; /(I)Is the/>, in the pavement gray level imageNegative correlation mapping values of pavement damage degree of individual local image blocks can be mapped by other methods, but the values after the negative correlation mapping need to be in a range of 0 to 1, for example, by an exponential function/>, which is based on natureNegative correlation mapping is performed.
Step S3: obtaining the search window length of the pavement gray image according to the area of each local image block and the obvious degree of the texture characteristics; and obtaining the neighborhood window length of each pixel point according to the search window length, the noise interference degree and the obvious degree of the texture characteristics of the local image block where each pixel point is located.
Further according to the purpose of the embodiment of the invention, the length of the neighborhood window and the length of the search window are calculated according to the noise interference degree and the obvious degree of the texture characteristics; because the same local image block represents the area with similar pavement elevation values, if the search window contains the same local image block of the image as far as possible, the neighborhood window in the search window has more referential property in the process of comparison; the present invention performs calculation of the search window length by combining the areas of the respective partial image blocks. In addition, as the neighborhood window with more complicated textures needs more image information for comparison, the accuracy of filtering is improved; therefore, the embodiment of the invention calculates the length of the search window by combining the obvious degree of the texture features on the basis of the area of the local image block. According to the embodiment of the invention, the search window length of the pavement gray image is obtained according to the area of each local image block and the obvious degree of the texture characteristics.
Preferably, the method for acquiring the length of the search window includes:
Taking the product of the arithmetic square root of the area of each local image block and the obvious degree of the texture characteristic as the reference window length of each local image block; and (3) taking the average value of the reference window lengths of all the local image blocks to be an odd integer upwards to obtain the search window length of the pavement gray level image. In the embodiment of the invention, the search window is a square window, so the area of the search window is the square of the length of the search window; the area of the search window needs to be approximate to the area of the local image block as much as possible when the search window needs to comprise the whole local image block, so that when the reference window length corresponding to the local image block is calculated, the calculation of the reference window length is performed on the basis of the arithmetic square root of the area of the local image block; further, considering that more complicated neighborhood windows need more image information for comparison, when the obvious degree of the texture features of the local image block is larger, the corresponding reference window length should be larger, and therefore the calculation of the reference window length is carried out by taking the obvious degree of the texture features as weight; and finally, combining the reference window lengths of all the local image blocks in the pavement gray level image, and obtaining the search window length of the pavement gray level image required by the embodiment of the invention in a mean value solving mode. The reason for taking the odd integer upward is that the length of the search window is typically odd to enable the filtered pixel to be in the center of the window; and since the search window is square, when the calculated search window length is 21, the corresponding search window is a 21×21 square window.
In the embodiment of the invention, the method for acquiring the length of the search window is expressed as the following formula:
Wherein, The length of the search window is the pavement gray level image; /(I)The number of the local image blocks in the pavement gray level image is the number of the local image blocks in the pavement gray level image; /(I)Is the/>, in the pavement gray level imageThe degree of texture feature clarity of the individual local image blocks; /(I)Is the first in the pavement gray level imageAreas of the local image blocks; /(I)Is an arithmetic square root function; /(I)To take an odd integer function upwards.
Further, when the length of the neighborhood window is calculated, the comparison calculation of the neighborhood window with more details is needed for the corresponding pixel point when the noise interference is stronger and the texture characteristics are more obvious, so that the more ideal noise reduction effect is obtained; and considering that the neighborhood window needs to be limited in the search window, the embodiment of the invention obtains the neighborhood window length of each pixel point according to the search window length, the noise interference degree and the obvious degree of the texture characteristics of the local image block where each pixel point is located.
Preferably, the method for acquiring the length of the neighborhood window includes:
Because when the noise interference is stronger and the texture characteristics are more obvious, the neighborhood window with more details is needed for the corresponding pixel point to perform contrast calculation, namely the length of the needed neighborhood window is larger; therefore, the noise interference degree and the obvious degree of texture features are in positive correlation with the length of the neighborhood window; the length of the neighborhood window needs to be smaller because the neighborhood window needs to be limited in the search window; taking into consideration that the value ranges corresponding to the noise interference degree and the obvious texture characteristic degree calculated by the embodiment of the invention are all 0 to 1; therefore, the embodiment of the invention takes the product of the noise interference degree and the local image block where the pixel point is positioned as the weight, and weights the length of the search window to obtain the length of the neighborhood window required by the embodiment of the invention; namely, the product of the length of the search window, the noise interference degree and the obvious degree of the texture feature of the local image block where each pixel point is positioned is up-odd integer to obtain the length of the neighborhood window of each pixel point.
In the embodiment of the invention, each pixel point in the pavement gray level image is sequentially used as the first pixel pointA pixel point is the firstThe method for obtaining the neighborhood window length of each pixel point is expressed as the following formula: /(I)
For/>Neighborhood window length of each pixel point,/>For/>The obvious degree of the texture characteristics of the local image block where each pixel point is positioned; /(I)Noise interference degree of the pavement gray level image; /(I)The length of the search window is the pavement gray level image; To take an odd integer function upwards. It should be noted that, the implementer may also obtain the length of the neighborhood window by other methods, for example, multiply the normalized value of the noise interference degree and the obvious degree and value of the texture feature by the length of the search window and then take the odd integer upwards to obtain the length of the neighborhood window.
Step S4: and in the search window where each pixel point is located, based on a non-local mean value filtering method, combining the noise interference degree according to the neighborhood window of each pixel point and the similarity degree of the gray distribution condition, the obvious degree of texture features and the road surface elevation value distribution condition between each neighborhood window obtained by sliding traversal, and obtaining the final filtering gray value of each pixel point.
After the search window length and the neighborhood window length are obtained, further carrying out self-adaptive calculation of final filtering gray values on each pixel point by combining non-local mean filtering; according to the principle of a non-local mean value filtering algorithm, similarity calculation is needed to be performed between a neighborhood window of each pixel point and each neighborhood window in a corresponding search window; the traditional similarity calculation method is to calculate the mean square error between the neighborhood windows in gray scale; however, for the damaged road image processed by the milling machine, different damaged areas usually show different image features, for example, a normal road area has less texture information compared with the damaged road area, but due to random distribution of noise, the noise points in the normal road are compared with the dense areas in a similarity manner, so that the weight of the damaged area is increased, and the filtering result is inaccurate. When the image is less interfered by noise, the traditional similarity measurement method based on the mean square error can truly obtain a relatively ideal calculation result; however, when the image is greatly interfered by noise, according to the calculation method of the length of the neighborhood window, the pixel points of different local image blocks can appear in the corresponding neighborhood window; and different image blocks are influenced by larger noise, so that the image characteristics of the neighborhood windows of the calculated pixel points are possibly more similar to those of other image blocks, and the similarity weight of the neighborhood windows which do not belong to the same local image block is indirectly improved, so that the accuracy of similarity measurement is influenced; because the elevation values in the same local image block are relatively close, the similarity weight of the neighborhood windows which do not belong to the same local image block is reduced; in the case of large noise interference, the similarity measure should be biased more toward regions with similar elevation values.
Therefore, in the search window where each pixel is located, the final filtering gray value of each pixel is obtained based on the non-local mean filtering method and according to the noise interference degree between the neighborhood window of each pixel and each neighborhood window obtained by sliding traversal, and the similarity degree of gray distribution, the obvious degree of texture features and the road surface elevation value distribution.
Preferably, the method for obtaining the final filtered gray value includes:
Sequentially taking each pixel point as a target pixel point; taking the search window of the target pixel point as a target search window; and taking the neighborhood window taking the target pixel point as the center as a target neighborhood window. Namely, the centers of the target search window and the target neighborhood window are target pixel points.
Traversing the target search window by a sliding window with the same shape and size as the target neighborhood window to obtain at least two comparison neighborhood windows; taking the pixel point corresponding to the central position of each contrast neighborhood window as a contrast pixel point; arranging all pixel points in the target neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence of the target neighborhood window; and arranging all the pixel points in each comparison neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence corresponding to each comparison neighborhood window. The purpose of calculating the pixel point reference sequence of the target neighborhood window and the comparison neighborhood window is to facilitate the subsequent comparison of window similarity through similarity between pixel points with the same index value.
Further, taking the noise interference degree as a weight, and constructing a window similarity calculation model according to the pixel point reference sequence between each comparison neighborhood window and the target neighborhood window and the similarity on the gray level distribution condition, the texture feature obvious degree and the road surface elevation value distribution condition; obtaining the window similarity of each comparison neighborhood window according to the window similarity calculation model;
preferably, the window similarity calculation model includes:
Wherein, For target neighborhood window/>Corresponding/>Window similarity of each comparison neighborhood window; /(I)Noise interference degree of the pavement gray level image; /(I)For target neighborhood window/>And/>The number of pixels in the pixel reference sequence corresponding to the neighborhood windows is compared; /(I)For target neighborhood window/>Gray values and the/>, of all pixels in a pixel reference sequenceMean square error of gray values of all pixels in the pixel point reference sequence of each contrast neighborhood window; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>The obvious degree of the texture characteristics of the local image block where each pixel point is positioned; for/> The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowThe obvious degree of the texture characteristics of the local image block where each pixel point is positioned; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>Road elevation values of point cloud data points corresponding to the pixel points; /(I)For/>The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowRoad elevation values of point cloud data points corresponding to the pixel points; /(I)A function is selected for the maximum. It should be noted that, the method for calculating the mean square error is known in the prior art by those skilled in the art, and is not further limited and described herein.
Firstly, the value range of the noise interference degree in the embodiment of the invention is 0 to 1, and when the noise interference degree is larger, the noise interference on the pavement gray level image is larger; when the combined image is less interfered by noise, the characteristics of the traditional method based on the mean square error between neighborhood windows in gray scale can obtain a relatively ideal calculation result can be known; when the noise interference degree of the pavement gray level image is lower, the reliability of the window similarity obtained based on the mean square error of gray level between the neighborhood windows is higher; therefore, in the window similarity calculation model, the negative correlation mapping value of the noise interference degree is used as the target neighborhood windowAnd/>The weights of the mean square errors in gray scale between the neighborhood windows are compared, and the similarity of the corresponding windows is larger when the mean square errors are smaller, so that the mean square errors are multiplied by the negative correlation mapping value of the noise interference degree after the negative correlation mapping is carried out.
Under the condition of large noise interference, the measurement index of the similarity should be more biased to the comparison neighborhood window in the same local image block; therefore, under the condition of larger noise interference degree, the pixel points in the comparison neighborhood window and the pixel points in the target neighborhood window belong to a local image block, and the similarity weight of the corresponding comparison neighborhood window is larger; the pavement elevation values in the same local image block are relatively close and the corresponding texture features are identical in significance; therefore, the greater the noise interference degree is, the greater the similarity of the corresponding windows is when the pavement elevation value of the pixel points between the target neighborhood window and the contrast neighborhood window is close to the corresponding salient degree of the texture features; therefore, the invention calculates the difference between the pixel points at the corresponding positions of the target window and the contrast window on the basis of the principle of mean square error calculation, namelyAnd the robustness is ensured by normalizing the maximum value as a denominator; calculating the difference/>, between the saliency degrees of the texture features of the local image blocks where the pixel points at the positions corresponding to the target window and the contrast window are locatedAnd the robustness is ensured by normalizing the maximum value as a denominator; further, after the difference on the pavement elevation value corresponding to each pixel point and the difference on the salient degree of the texture feature are combined through the product, the difference/>, on the pavement elevation value and the salient degree of the texture feature, of the whole target neighborhood window and the contrast neighborhood window is calculated through the mean value modeAccording to the principle that the smaller the difference is, the higher the similarity is, the negative correlation mapping is carried out on the difference; it should be noted that, the implementer also performs the negative correlation mapping by other methods, such as inverse, etc., and further description is omitted herein. When the noise interference degree is larger, the similarity of the corresponding windows is larger when the pavement elevation value of the pixel point between the target neighborhood window and the contrast neighborhood window is closer to the corresponding texture feature significance degree; and therefore, the negative correlation mapping value is weighted according to the noise interference degree, and the corresponding window similarity is obtained.
Further based on the principle of non-local mean filtering, the ratio between the window similarity of each comparison neighborhood window and the accumulated value of the window similarity of all the comparison neighborhood windows is used as the filtering weight of each comparison neighborhood window; taking the product of the filtering weight and the gray value of the contrast pixel point of each contrast neighborhood window as the reference weighted gray value of each contrast neighborhood window; and taking the accumulated value of the reference weighted gray values of all the comparison neighborhood windows as the final filtered gray value of the target pixel point. It should be noted that, the non-local mean filtering process after obtaining the similarity between the neighborhood windows is well known in the art, and will not be further described herein.
Step S5: obtaining a pavement denoising image according to the final filtering gray value of each pixel point in the pavement gray image; and optimizing and adjusting the travelling speed of the milling machine according to the pavement denoising image.
After the final filtering gray values of all the pixel points in the pavement gray image are obtained, namely, after the smoothed gray value corresponding to each pixel point, the original gray value is further replaced by the final filtering gray value, and a denoised image is obtained; according to the embodiment of the invention, the pavement denoising image is obtained according to the final filtering gray value of each pixel point in the pavement gray image. And the travelling speed of the milling machine is optimally adjusted according to the pavement denoising image.
Preferably, the method for optimally adjusting the traveling speed of the milling machine according to the pavement denoising image comprises the following steps:
Inputting the pavement denoising image into a trained convolutional neural network, and outputting pavement quality indexes of each advancing pavement in the advancing process of the milling machine; and adjusting the travelling speed of the milling machine according to the road surface quality index. In the embodiment of the invention, when the convolutional neural network is trained, the pavement denoising images with different pavement quality are manually scored, and the pavement quality is divided into 10 pavement quality index types of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1 in sequence from low pavement quality to high pavement quality, and the larger the pavement quality index is, the better the corresponding pavement quality is. The product of the maximum speed of the milling machine running on the current road and the road surface quality index is further taken as the required running speed of the embodiment of the invention. It should be noted that, in addition to changing the traveling speed of the milling machine, the operator may also change the working power of the milling machine, that is, the better the road quality index, the smaller the working power of the milling machine, for example, 1 minus the road quality index is multiplied by the maximum working power of the milling machine to obtain the adjusted working power of the milling machine; and will not be further described herein.
In summary, the method combines the position distribution of the point cloud data points corresponding to the pavement gray level image to divide each local image block; combining the gray gradient distribution in the local image block and the road elevation value of the corresponding point cloud data point to obtain the noise interference degree; determining the length of a search window according to the area and texture characteristics of the local image block; combining the noise interference degree and the obvious texture characteristic degree to obtain the length of a neighborhood window; on the basis of the search window and the neighborhood window, the road surface denoising image with better denoising effect is obtained according to the noise interference degree, the obvious texture characteristic degree and the gray value distribution by a non-local mean value filtering method, so that the travelling speed of the milling machine is more accurately optimized and adjusted based on the road surface denoising image.
The invention also provides an intelligent milling machine optimization control system, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes any one of the steps of the intelligent milling machine optimization control method when executing the computer program.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (5)
1. An intelligent milling machine optimization control method is characterized by comprising the following steps:
acquiring a pavement gray image and a corresponding point cloud data matrix of each advancing pavement in the advancing process of the milling machine; each point cloud data point in the point cloud data matrix corresponds to each pixel point in the pavement gray level image;
Dividing the pavement gray level image into at least two local image blocks according to the spatial position distribution condition of the point cloud data points in the point cloud data matrix; obtaining the obvious degree of the texture characteristics of each local image block according to the gray gradient distribution condition in each local image block; obtaining the noise interference degree of the pavement gray level image according to the pavement elevation value distribution concentration condition of the point cloud data points and the texture characteristic obvious degree;
obtaining the search window length of the pavement gray image according to the area of each local image block and the obvious degree of the texture characteristics; obtaining the neighborhood window length of each pixel point according to the search window length, the noise interference degree and the obvious degree of the texture characteristics of the local image block where each pixel point is located;
In a search window where each pixel point is located, combining the noise interference degree based on a non-local mean filtering method, and obtaining a final filtering gray value of each pixel point according to the gray distribution condition, the obvious degree of texture features and the similarity degree of road surface elevation value distribution condition between a neighborhood window of each pixel point and each neighborhood window obtained by sliding traversal;
obtaining a pavement denoising image according to the final filtering gray value of each pixel point in the pavement gray image; optimizing and adjusting the travelling speed of the milling machine according to the pavement denoising image;
The method for obtaining the obvious degree of the texture features comprises the following steps:
in each local image block, taking the distance between each pixel point and the centroid of the local image block as the reference distance of each pixel point;
Taking the product of the normalized value of the gray gradient value of each pixel point and the reference distance as a local texture characteristic value of each pixel point;
Taking the normalized value of the average value of the local texture characteristic values of all pixel points in each local image block as the texture characteristic obvious degree of each local image block;
the method for acquiring the noise interference degree comprises the following steps:
Taking the normalized value of the variance of the elevation values of the road surfaces of all the point cloud data points in each local image block as the road surface damage degree of each local image block; taking the product of the negative correlation mapping value of the pavement damage degree and the obvious degree of the texture feature as the texture feature parameter of each local image block; taking the average value of texture characteristic parameters of all local image blocks as the noise interference degree of the pavement gray level image;
the method for acquiring the length of the neighborhood window comprises the following steps:
Taking the product of the search window length, the noise interference degree and the obvious degree of the texture feature of the local image block where each pixel point is located to be an odd integer upwards to obtain the neighborhood window length of each pixel point;
The method for acquiring the final filtered gray value comprises the following steps:
sequentially taking each pixel point as a target pixel point; taking the search window of the target pixel point as a target search window; taking a neighborhood window taking a target pixel point as a center as a target neighborhood window;
Traversing the target search window by a sliding window with the same shape and size as the target neighborhood window to obtain at least two comparison neighborhood windows; taking the pixel point corresponding to the central position of each contrast neighborhood window as a contrast pixel point; arranging all pixel points in the target neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence of the target neighborhood window; arranging all the pixel points in each comparison neighborhood window in a sequence from left to right and from top to bottom to obtain a pixel point reference sequence corresponding to each comparison neighborhood window;
Taking the noise interference degree as a weight, and constructing a window similarity calculation model according to the pixel point reference sequence between each comparison neighborhood window and the target neighborhood window and the similarity on the gray level distribution condition, the texture feature obvious degree and the road surface elevation value distribution condition; obtaining the window similarity of each comparison neighborhood window according to the window similarity calculation model;
Taking the ratio between the window similarity of each comparison neighborhood window and the accumulated value of the window similarity of all the comparison neighborhood windows as the filtering weight of each comparison neighborhood window; taking the product of the filtering weight and the gray value of the contrast pixel point of each contrast neighborhood window as the reference weighted gray value of each contrast neighborhood window; taking the accumulated value of the reference weighted gray values of all the comparison neighborhood windows as the final filtering gray value of the target pixel point;
the window similarity calculation model includes:
Wherein, For target neighborhood window/>Corresponding/>Window similarity of each comparison neighborhood window; /(I)Noise interference degree of the pavement gray level image; /(I)For target neighborhood window/>And/>The number of pixels in the pixel reference sequence corresponding to the neighborhood windows is compared; /(I)For target neighborhood window/>Gray values and the/>, of all pixels in a pixel reference sequenceMean square error of gray values of all pixels in the pixel point reference sequence of each contrast neighborhood window; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>The obvious degree of the texture characteristics of the local image block where each pixel point is positioned; /(I)For/>The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowThe obvious degree of the texture characteristics of the local image block where each pixel point is positioned; /(I)For target neighborhood window/>The corresponding pixel point reference sequence is the/>Road elevation values of point cloud data points corresponding to the pixel points; /(I)For/>The/>, in the pixel point reference sequence corresponding to each comparison neighborhood windowRoad elevation values of point cloud data points corresponding to the pixel points; /(I)A function is selected for the maximum.
2. The optimal control method of an intelligent milling machine according to claim 1, wherein the local image block acquisition method comprises:
In the pavement gray level image, carrying out cluster analysis according to pixel coordinates of all point cloud data points in a point cloud data matrix through a k-means clustering algorithm to obtain at least two point cloud data point cluster clusters; and clustering each point cloud data point into a corresponding image area in the pavement gray level image to serve as a local image block.
3. The optimal control method of an intelligent milling machine according to claim 1, wherein the search window length obtaining method comprises:
Taking the product of the arithmetic square root of the area of each local image block and the saliency of the texture feature as the reference window length of each local image block; and (3) taking the average value of the reference window lengths of all the local image blocks to be an odd integer upwards to obtain the search window length of the pavement gray level image.
4. The method for optimally controlling the intelligent milling machine according to claim 1, wherein the method for optimally adjusting the traveling speed of the milling machine according to the pavement denoising image comprises the following steps:
Inputting the pavement denoising image into a trained convolutional neural network, and outputting pavement quality indexes of each advancing pavement in the advancing process of the milling machine; and adjusting the travelling speed of the milling machine according to the road surface quality index.
5. An intelligent milling machine optimization control system comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the processor, when executing the computer program, implements the steps of the method according to any one of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410361430.XA CN117974492B (en) | 2024-03-28 | 2024-03-28 | Optimal control method and system for intelligent milling machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410361430.XA CN117974492B (en) | 2024-03-28 | 2024-03-28 | Optimal control method and system for intelligent milling machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117974492A CN117974492A (en) | 2024-05-03 |
CN117974492B true CN117974492B (en) | 2024-06-11 |
Family
ID=90858274
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410361430.XA Active CN117974492B (en) | 2024-03-28 | 2024-03-28 | Optimal control method and system for intelligent milling machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117974492B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346608A (en) * | 2013-07-26 | 2015-02-11 | 株式会社理光 | Sparse depth map densing method and device |
CN108921800A (en) * | 2018-06-26 | 2018-11-30 | 成都信息工程大学 | Non-local mean denoising method based on form adaptive search window |
CN109815961A (en) * | 2018-12-25 | 2019-05-28 | 山东省交通规划设计院 | A kind of pavement patching class Defect inspection method based on local grain binary pattern |
CN113838111A (en) * | 2021-08-09 | 2021-12-24 | 北京中科慧眼科技有限公司 | Road texture feature detection method and device and automatic driving system |
WO2022067647A1 (en) * | 2020-09-30 | 2022-04-07 | 华为技术有限公司 | Method and apparatus for determining pavement elements |
CN114463286A (en) * | 2022-01-13 | 2022-05-10 | 东南大学 | Road surface point cloud smoothing and crack information processing method based on moving window |
CN115825410A (en) * | 2022-11-22 | 2023-03-21 | 重庆大学 | Method for estimating tire road surface action noise level based on road surface structure |
CN115953678A (en) * | 2022-12-09 | 2023-04-11 | 浙大城市学院 | Pavement damage detection method based on local gray extreme point and feature fusion |
CN116092015A (en) * | 2023-04-06 | 2023-05-09 | 安徽乾劲企业管理有限公司 | Road construction state monitoring method |
CN116152239A (en) * | 2023-04-18 | 2023-05-23 | 中铁九局集团第一建设有限公司 | Expressway pavement disease detection method and system |
CN117036641A (en) * | 2023-06-08 | 2023-11-10 | 四川轻化工大学 | Road scene three-dimensional reconstruction and defect detection method based on binocular vision |
-
2024
- 2024-03-28 CN CN202410361430.XA patent/CN117974492B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346608A (en) * | 2013-07-26 | 2015-02-11 | 株式会社理光 | Sparse depth map densing method and device |
CN108921800A (en) * | 2018-06-26 | 2018-11-30 | 成都信息工程大学 | Non-local mean denoising method based on form adaptive search window |
CN109815961A (en) * | 2018-12-25 | 2019-05-28 | 山东省交通规划设计院 | A kind of pavement patching class Defect inspection method based on local grain binary pattern |
WO2022067647A1 (en) * | 2020-09-30 | 2022-04-07 | 华为技术有限公司 | Method and apparatus for determining pavement elements |
CN113838111A (en) * | 2021-08-09 | 2021-12-24 | 北京中科慧眼科技有限公司 | Road texture feature detection method and device and automatic driving system |
CN114463286A (en) * | 2022-01-13 | 2022-05-10 | 东南大学 | Road surface point cloud smoothing and crack information processing method based on moving window |
CN115825410A (en) * | 2022-11-22 | 2023-03-21 | 重庆大学 | Method for estimating tire road surface action noise level based on road surface structure |
CN115953678A (en) * | 2022-12-09 | 2023-04-11 | 浙大城市学院 | Pavement damage detection method based on local gray extreme point and feature fusion |
CN116092015A (en) * | 2023-04-06 | 2023-05-09 | 安徽乾劲企业管理有限公司 | Road construction state monitoring method |
CN116152239A (en) * | 2023-04-18 | 2023-05-23 | 中铁九局集团第一建设有限公司 | Expressway pavement disease detection method and system |
CN117036641A (en) * | 2023-06-08 | 2023-11-10 | 四川轻化工大学 | Road scene three-dimensional reconstruction and defect detection method based on binocular vision |
Non-Patent Citations (3)
Title |
---|
卢紫微 ; 张燕 ; 常东超 ; 吴成东 ; .基于多特征分析的路面裂缝检测算法.控制工程.2018,(04),全文. * |
基于图像局部均值标准差算法的沥青路面裂缝检测;罗玉柱;;测绘与空间地理信息;20171225(第12期);全文 * |
基于多特征分析的路面裂缝检测算法;卢紫微;张燕;常东超;吴成东;;控制工程;20180420(04);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN117974492A (en) | 2024-05-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110443836A (en) | A kind of point cloud data autoegistration method and device based on plane characteristic | |
CN109903319B (en) | Multi-resolution-based fast iteration closest point registration algorithm | |
CN107038717A (en) | A kind of method that 3D point cloud registration error is automatically analyzed based on three-dimensional grid | |
CN116309757B (en) | Binocular stereo matching method based on machine vision | |
CN108596975A (en) | A kind of Stereo Matching Algorithm for weak texture region | |
CN113628263A (en) | Point cloud registration method based on local curvature and neighbor characteristics thereof | |
CN117496359B (en) | Plant planting layout monitoring method and system based on three-dimensional point cloud | |
CN113093216A (en) | Irregular object measurement method based on laser radar and camera fusion | |
CN112085675A (en) | Depth image denoising method, foreground segmentation method and human motion monitoring method | |
CN113887624A (en) | Improved feature stereo matching method based on binocular vision | |
CN111105452A (en) | High-low resolution fusion stereo matching method based on binocular vision | |
CN109657525A (en) | A kind of rapid correction method for traditional LiDAR filtering error | |
CN114663373A (en) | Point cloud registration method and device for detecting surface quality of part | |
CN109741358A (en) | Superpixel segmentation method based on the study of adaptive hypergraph | |
CN106991660B (en) | The three dimensional ultrasonic image data methods of sampling decomposed based on modified Octree | |
CN105488798B (en) | SAR image method for measuring similarity based on point set contrast | |
CN109359604A (en) | Meter recognition method under shadow interference towards crusing robot | |
CN109272508B (en) | Petri network image segmentation method based on rough set and rough entropy | |
CN114612412A (en) | Processing method of three-dimensional point cloud data, application of processing method, electronic device and storage medium | |
CN114494704A (en) | Method and system for extracting framework from binary image in anti-noise manner | |
CN114120012A (en) | Stereo matching method based on multi-feature fusion and tree structure cost aggregation | |
CN106683125A (en) | RGB-D image registration method based on 2D/3D mode switching | |
CN117974492B (en) | Optimal control method and system for intelligent milling machine | |
CN114120098A (en) | SAR image lakeshore line detection method and system based on MRSF model | |
CN115953604B (en) | Real estate geographic information mapping data acquisition method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |