Garden construction progress and quality intelligent inspection analysis method

Abstract

The application discloses a garden construction progress quality intelligent inspection analysis method and relates to the technical field of garden construction progress monitoring.The method comprises the following steps: acquiring an aligned image and a planned spreading degree of a planar process area; extracting a laid attribution degree and an un-laid attribution degree per spatial unit; taking the product of the two as attribution tension, determining a gradual transition zone of the attribution tension exceeding a threshold value, and dividing the rest into a laid sharp zone and an un-laid sharp zone according to a convergence direction; connecting a spreading front line in the transition zone, obtaining a front normal and a transition zone width; dividing an equivalent zone along the normal, obtaining a zone attribution tension and a zone transition gradient; determining a front section spreading degree by comprehensively considering the two; combining the front section spreading degrees and the sharp zones into a spreading degree map, comparing the spreading degree map with the planned spreading degree, and outputting a process progress and deviation result.The application can make the progress recognition result more consistent with the actual spreading situation on site, and support progress spatial distribution analysis and lagged landing point positioning.

Description

Technical Field

[0001] This invention relates to the field of landscape construction progress monitoring technology, specifically to an intelligent inspection and analysis method for landscape construction progress and quality. Background Technology

[0002] Landscape construction progress management requires continuous monitoring of the completion status of surface greening processes such as lawns, ground cover, and color blocks. With the popularization of drone aerial photography and on-site cameras, image-based automatic recognition of construction progress is gradually replacing manual on-site verification and becoming a common method for progress inspection.

[0003] Existing image progress recognition methods typically first segment the acquired image into elements, classifying each spatial region in the image into two categories: completed or incomplete. Then, they calculate the proportion of the area occupied by the completed regions, convert this proportion into the process completion rate, and compare it with the construction plan to determine whether there is a delay.

[0004] However, the process of surface greening is not completed in an instant, but rather spreads gradually along a front that is constantly moving forward. There is often no clear boundary between the paved and unpaved areas, but rather a blurred area that gradually transitions from complete coverage to complete exposure. In this transitional area, the ground surface is neither a completely finished product nor a completely blank space, but rather in an intermediate state of formation.

[0005] How to accurately quantify the actual progress reflected by the transition area, so that the progress identification results are more in line with the actual on-site paving situation, has become an urgent technical problem to be solved. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an intelligent inspection and analysis method for the progress and quality of garden construction.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] In a first aspect, this invention discloses an intelligent inspection and analysis method for the progress and quality of garden construction, comprising the following steps:

[0009] Obtain the alignment image and planned spread of the covered process area;

[0010] For each spatial cell of the aligned image, the tiling and non-tiling classifications are extracted. The tiling and non-tiling classifications represent the tendency of the spatial cell to belong to the completed and non-complete side categories, respectively.

[0011] The product of the paved and unpaved classification is used as the classification tension that each spatial unit belongs to both categories at the same time. The connected regions of spatial units with classification tension greater than the preset tension judgment parameter are defined as gradual transition zones. The remaining connected regions are divided into paved sharp regions and unpaved sharp regions according to the convergence direction of the paved and unpaved classification.

[0012] Within the gradual transition zone, the spatial units located between the paved and unpaved sharp areas are connected in series to form a paved frontal line. The frontal normal of each point on the paved frontal line from the unpaved side to the paved side is calculated, and the width of the transition zone along the frontal normal from the boundary of the unpaved sharp area to the boundary of the paved sharp area is calculated.

[0013] The gradual transition zone is divided into sequentially arranged normal isobars along the frontal normal. The zone assignment tension and zone transition gradient are calculated for each normal isobar. The zone transition gradient is the rate of change of the paved assignment degree along the frontal normal.

[0014] For each point of the spreading frontal line, the spreading degree of the continuous frontal segment is determined by the band-attached tension distributed along the frontal normal and the band transition gradient within the width of the transition zone.

[0015] The spread of each frontal segment, the spread sharp area and the unspread sharp area are combined according to their spatial location to form a spread map. The spread map is compared with the planned spread to output the progress and deviation results of the surface process.

[0016] Secondly, this invention discloses an intelligent inspection and analysis system for the progress and quality of garden construction, comprising:

[0017] The image acquisition module is used to acquire aligned images and planned spread of the covered process area;

[0018] The attribution extraction module is used to extract the attribution of the laid and unlaid parts of the aligned image on a per-spatial-cell basis. The attribution of the laid and unlaid parts are the tendency strengths of the spatial cell to belong to the completed and uncompleted side categories, respectively.

[0019] The region division module is used to use the product of the paved and unpaved classification as the classification tension that represents each spatial unit simultaneously belonging to both categories. The connected regions of spatial units with classification tension greater than the preset tension judgment parameter are determined as gradual transition zones. Based on the convergence direction of the paved and unpaved classification, the remaining connected regions are divided into paved sharp regions and unpaved sharp regions.

[0020] The frontal analysis module is used to connect spatial units located between paved and unpaved sharp areas into a paved frontal line within a gradual transition zone, calculate the frontal normal from the unpaved side to the paved side at each point on the paved frontal line, and calculate the width of the transition zone along the frontal normal from the boundary of the unpaved sharp area to the boundary of the paved sharp area.

[0021] The iso-band processing module is used to divide the gradual transition zone into sequentially arranged normal iso-bands along the frontal normal, and to calculate the band assignment tension and band transition gradient for each normal iso-band. The band transition gradient is the rate of change of the paved assignment degree along the frontal normal.

[0022] The spreading degree determination module is used to determine the spreading degree of continuous frontal segments at each point of the spreading frontal line within the width of the transition zone by the belt-attached tension distributed along the frontal normal and the belt transition gradient.

[0023] The splicing and comparison module is used to splice the spread of each front segment, the spread sharp area and the unspliced ​​sharp area according to their spatial positions to form a spread map, compare the spread map with the planned spread, and output the progress and deviation results of the surface process.

[0024] According to another aspect of this application, an electronic device is provided, including a memory and a processor, the memory being used to store computer-executable instructions, and the processor being used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method described above.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. By extracting the independent attribution degree of two paths on a per-spatial-unit basis, the attribution tension is constructed by multiplying the two attribution degrees. Connected regions with attribution tension greater than the threshold are identified as gradual transition zones. Furthermore, the frontal line is located and spread within the transition zone, the frontal normal and the width of the transition zone are obtained, and then isopleths are divided along the normal. The attribution tension and the transition gradient of the zone are combined to determine the spread of the frontal segment. Finally, the spread is combined into a per-unit spread map and compared with the planned spread. This scheme explicitly separates the previously ignored ongoing areas and assigns them continuous spread completion, so that the progress identification results can truly reflect the continuous advancement process of the planar process from unpaved to paved.

[0027] 2. By dividing the paving degree map into three non-overlapping zones—the completed zone, the transition zone, and the unpaved zone—and calculating the completion contribution of each zone separately, the paving degree of each process is weighted by area and synthesized. Simultaneously, the weighted root mean square of the paving degree relative to the process paving degree for each unit is calculated as the spatial dispersion of the schedule. The system outputs multi-dimensional results, including the process paving degree, the completion contribution of the three zones, and the spatial dispersion. This not only provides the overall schedule deviation but also presents the completion status of the completed zone, the transition zone, and the unpaved zone, as well as the spatial fluctuation of the schedule. This overcomes the limitation of traditional single averages, which cannot reflect the uneven spatial distribution within the schedule, and provides a more detailed decision-making basis for the construction party to strategically allocate work resources. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is an overall block diagram of the method in Embodiment 1 of the present invention;

[0030] Figure 2 This is an overall block diagram of the system according to Embodiment 2 of the present invention;

[0031] Figure 3 This is a timing diagram of the processing interactions of the present invention;

[0032] Figure 4 This is a block diagram of an electronic device according to the present invention. Detailed Implementation

[0033] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1:

[0035] like Figure 1 As shown, the intelligent inspection and analysis method for the progress and quality of garden construction includes the following steps:

[0036] To facilitate understanding, we will use a drone inspection of a lawn laying process in a park landscaping project as an example. This process covers an area of ​​approximately... The rectangular lawn area was laid from the pre-laid turf on one side to the bare ground on the other side. Near the advancing edge, there was a transition zone where the turf went from dense to sparse and from solid to bare. At a certain inspection moment, a drone took a vertical inspection image from above the area. At the same time, several ground reference points marked with spray paint were pre-placed at the four corners and boundaries of the area. This embodiment uses this inspection image as the original input to reconstruct the continuous progress of the lawn laying process at that moment.

[0037] Step S1: Obtain the alignment image and planned spread of the covered process area.

[0038] The aligned image is not a direct copy of the original inspection image taken by the drone, but rather a result of spatial alignment processing of the original inspection image. The original inspection image is a three-channel two-dimensional pixel array, where each pixel carries a grayscale value, and the pixels only have row and column positions without a true spatial scale. Due to factors such as the drone's shooting angle, lens distortion, and flight attitude, there is perspective distortion between the pixel positions in the original inspection image and the actual positions of the plots, making it unsuitable for direct use in progress statistics based on site area. Therefore, this step introduces a calibration control point set, which consists of several ground reference points with known actual site coordinates and their pixel coordinates in the original inspection image. Each control point corresponds to a coordinate pair of pixel coordinates and site coordinates.

[0039] The specific spatial alignment process includes: forming coordinate pairs by combining the image pixel coordinates of each control point in the calibration control point set with the actual site coordinates; and fitting a homography transformation matrix using least squares from four or more pairs of non-collinear control points. so that each control point passes The sum of squared residuals after mapping to its site coordinates is minimized; for each pixel of the inspection image, with... The pixel coordinates are mapped to the field coordinate system, and the grayscale is resampled using bilinear interpolation on the regular space grid of the field coordinate system to obtain the aligned image. The mapping relationship from pixel coordinates to field coordinates can be expressed as: ;

[0040] In the formula, Image pixel coordinates, The coordinates mapped to the site coordinate system. The homography transformation matrix is ​​a 3x3 matrix. This indicates equality in homogeneous coordinates, where the two sides differ by a non-zero scaling factor. Since this mapping holds in homogeneous coordinates, the homography transformation matrix... There is overall scale uncertainty, that is Equivalent to the mapping represented by any non-zero multiple, directly solving for the minimum sum of squared residuals will degenerate to the zero solution. Therefore, during fitting, ... Adding normalization constraints, such as setting the last element to 1 or the sum of squares of all elements to 1, yields a unique least-squares solution after eliminating zero solutions and scale redundancy. Four or more pairs of non-collinear control points are used because the homography transformation matrix has eight independent degrees of freedom in homogeneous coordinates. Each pair of control points provides two equation constraints, and four pairs of non-collinear control points provide exactly eight constraints to uniquely determine the matrix. When there are more than four pairs of control points, the equation system becomes overdetermined. Least-squares fitting minimizes the sum of squares of the mapping residuals of all control points, thus suppressing the impact of individual control point calibration errors on overall alignment. The aforementioned bilinear interpolation refers to taking the grayscale values ​​of the four original pixels surrounding the corresponding position in the original inspection image for each target position on the regular space grid of the site coordinate system, and then weighting and summing them according to the distance from the target position to these four pixels to obtain the resampled grayscale value, making the grayscale values ​​of the aligned image smooth and continuous on the integer grid.

[0041] Using the lawn example, ground reference points at the four corners and boundaries of the plot are used as control points. After fitting the homography transformation matrix, the original inspection image is mapped and corrected as a whole onto a site coordinate system with one corner of the plot as the origin and meters as the unit of length. The images are then plotted on the site coordinate system according to... The regular spatial grid is resampled to obtain the aligned image. At this point, the aligned image becomes a three-channel two-dimensional image with spatial coordinate indexing, where each spatial unit corresponds to a plot of land. That is, the area is The actual area, and all subsequent calculations involving area and distance will be performed within this site coordinate system. Planned paving depth. This is also obtained as raw input, representing the degree of paving completion that the process area should achieve at this inspection time according to the construction plan. The value is between 0 and 1; in this example, the planned paving completion is set to [value missing]. This means that half of the plan has been completed.

[0042] The above method corrects the original inspection images with perspective distortion into aligned images with the actual site spatial scale, so that the subsequent analysis results of each spatial unit can be statistically compared according to the actual site area, providing a spatial benchmark and comparison benchmark for the entire progress inversion.

[0043] Step S2: Extract the tiled and untilted classifications for each spatial cell of the aligned image. The tiled and untilted classifications represent the tendency of the spatial cell to belong to the completed and uncompleted side categories, respectively.

[0044] Traditional image progress recognition methods force each area in the image to be classified as either completed or incomplete, essentially a binary classification that ignores the intermediate state of the surface in the transition zone, which is neither a completely finished product nor completely blank. This step lays the foundation for subsequent quantification of the degree of semi-completion by simultaneously extracting two unconstrained attributions for the same spatial unit.

[0045] Specifically, the process of extracting the laid and unlaid membership values ​​of the aligned image per spatial unit is implemented using a semantic segmentation network with an encoder-decoder structure. The encoding path of this network consists of alternating convolutional and downsampling layers. The convolutional layers extract spatial features from the aligned image layer by layer, while the downsampling layers compress the spatial resolution to expand the receptive field. The decoding path consists of alternating upsampling and convolutional layers. The upsampling layers restore the spatial resolution layer by layer, and the total upsampling ratio of the decoding path matches the total downsampling ratio of the encoding path, ensuring that the network output restores the same spatial resolution as the aligned image and achieves per-spatial-unit output. The network ends with two segmentation feature channels corresponding to the completed and uncompleted categories, respectively. , The sigmoid function is applied independently to the activation responses of the two segmented feature channels to obtain the laid-up belonging degree. Belonging to Unpaved Areas :

[0046] ;

[0047] In the formula, The position of the spatial unit in the site coordinate system. , These represent the activation responses of the segmentation feature channels on the completed and incomplete sides, respectively. For the sigmoid function, For the already established degree of belonging, The degree of non-attachment is represented by two values, both of which are between 0 and 1.

[0048] The key to this step lies in obtaining the two classification scores independently using the sigmoid function, without imposing cross-class normalization that would sum the two scores to one, i.e., without using softmax. The sigmoid function monotonically compresses the activation response of each score to the 0-1 range, and the two scores do not affect each other. Therefore, it allows the two classification scores of the same spatial unit to be high or low simultaneously: when a unit has both finished turf features and residual bare ground features, its paved and unpaved classification scores can both take higher values, which corresponds to a semi-complete unit that is in the process of formation. However, if softmax is used to make the sum of the two scores always equal to one, then an increase in one score will inevitably force the other score to decrease, and the semi-complete state will be artificially merged into one side, thus losing the intermediate semantics carried by the transition zone. It is this design of independent dual-score taking and unconstrained sum that preserves the semi-complete semantics, allowing the subsequent use of the product of the two classification scores to characterize the degree of simultaneous classification into two categories. This is also the basis for this scheme to distinguish itself from traditional binary classification progress recognition and to characterize the continuous progress of the transition zone.

[0049] To enable the semantic segmentation network to distinguish between laid and unlaid semantic segments while preserving partially completed semantics, it needs to be pre-trained. The training samples consist of aligned images from historical inspections. For each sample image, two independent binary labels are labeled per spatial unit: the completed side label and the unlaid side label. with incomplete side label For clearly defined, completed spatial units, the completed side label is set to 1, and the incomplete side label is set to 0; for clearly defined, exposed spatial units, the opposite applies; for semi-complete spatial units in the transition zone, both the completed and incomplete side labels are allowed to be set to 1 simultaneously, to indicate that they possess features of both classes. This labeling method, which allows both labels to be set to 1 simultaneously, corresponds to the structure of two independent sigmoid functions in the inference phase, without cross-class normalization. Since the two membership degrees are independent and without softmax normalization, the training loss uses the sum of the binary cross-entropy of each of the two functions, rather than the single multi-class cross-entropy in the multi-class case.

[0050] ;

[0051] ;

[0052] In the formula, For total training losses, For binary cross-entropy, , These represent the network outputs of the laid and unlaid network affiliations, respectively. , These are the corresponding binary labels for the completed and incomplete sides. For the set of spatial units participating in the calculation, The reason for using the sum of two binary cross-entropies instead of a single multi-class cross-entropy is that the multi-class cross-entropy contains the normalization assumption that the sum of the probabilities of each class is one, which forces the laid and unlaid parts to be mutually exclusive, thus erasing the half-complete semantics during the training phase; while the two independent binary cross-entropies supervise the two parts separately and are not coupled with each other, which allows the network to learn the value patterns of the transition unit with high membership of both parts at the same time.

[0053] The training process includes: inputting the aligned image into the network to calculate the two-way attribution, calculating the total training loss according to the above formula, calculating the gradient through backpropagation, updating the network parameters, and iterating in this way until the total training loss converges or reaches the preset number of training rounds, thus obtaining a semantic segmentation network that can be used to extract laid and unlaid attribution per spatial unit.

[0054] Using the lawn example, after inputting the aligned image at that moment into the trained semantic segmentation network, a pair of attribution scores is output for each spatial unit. Spatial units located deep within the laid lawn have an attribution score of approximately 0.95 for laid and approximately 0.05 for unlaid. Conversely, spatial units located deep within bare ground have an attribution score of approximately 0.05 for laid and approximately 0.95 for unlaid. Spatial units located in the transition zone near the spreading front, such as a unit where the turf is halfway laid with bare soil in the gaps, have an attribution score of approximately 0.60 for laid and approximately 0.40 for unlaid. Both scores are at a moderate level, reflecting that the area is in the process of being laid.

[0055] The above scheme yields a classification map for each spatial unit, carrying both the classification degree of completed and uncompleted units. Compared to the traditional hard classification of single completed or incomplete units, this classification map additionally retains the semi-complete information of transition zone units that simultaneously belong to both categories, serving as input for subsequent regional division and semi-complete degree quantification.

[0056] Step S3: The product of the paved and unpaved attribution degrees is used as the attribution tension that represents each spatial unit simultaneously belonging to both categories. The connected regions of spatial units with attribution tension greater than the preset tension judgment parameter are determined as gradual transition zones. Based on the convergence direction of the paved and unpaved attribution degrees, the remaining connected regions are divided into paved sharp regions and unpaved sharp regions.

[0057] This step calculates the attribution tension on a spatial unit basis. The attribution tension is defined as the product of the attribution degree of the tiled area and the attribution degree of the untiled area in that unit: ;

[0058] In the formula, spatial unit Tension of belonging , These represent the paved and unpaved affiliations of the unit, respectively. Affiliation tension characterizes the degree to which a spatial unit simultaneously belongs to both the completed and uncompleted categories: only when both affiliations are high is the product large, indicating the unit is in a semi-complete transitional state, resembling both a finished product and bare land; if either affiliation is close to zero, the product is close to zero, indicating the unit clearly belongs to one side. Thus, affiliation tension precisely condenses the semi-complete semantics retained independently in step S2 into a scalar that can be used as a criterion. It should be noted that since both affiliations are between 0 and 1, the maximum value of their product is obtained when both are equal and both are 0.5, meaning the upper limit of affiliation tension is 0.25. This upper limit corresponds to the spatial unit within the transition zone where the two categories are most balanced and the semi-completeness is strongest. Therefore, a threshold of affiliation tension greater than a preset tension judgment parameter is used. Spatial units are extracted and merged according to the adjacency connectivity criterion, and the resulting connected regions form a gradual transition zone. Preset tension judgment parameters The value of is subject to the above upper limit and can only fall between 0 and 0.25: The smaller the value, the more spatial units are judged to be included in the gradient transition zone, and the wider the transition zone becomes. The larger the value, the narrower the transition zone. The specific value can be determined empirically based on the typical width of the turf transition zone.

[0059] For spatial units outside the gradual transition zone, they are further divided into paved sharp zones and unpaved sharp zones based on the convergence direction of the paved and unpaved membership degrees. The specific division process includes: for spatial units outside the gradual transition zone, calculating the absolute value of the difference between the paved and unpaved membership degrees as the membership convergence degree. ;

[0060] In the formula, spatial unit The degree of convergence, ranging from 0 to 1, characterizes the degree to which two degrees of belonging are separated and clearly converge to one side: the higher the degree of convergence, the greater the difference between the two degrees of belonging, and the more clearly the unit is biased towards a certain category. Based on this, if the degree of convergence of a spatial unit is greater than a preset convergence criterion parameter... If the degree of belonging to the paved area is greater than that to the unpaved area, then the connected area of ​​the spatial unit is determined as the paved sharp area. If the convergence of the spatial unit is greater than the preset convergence criterion parameter... If the degree of belonging to the unpaved area is greater than that of the paved area, then the connected area of ​​the spatial unit is defined as the unpaved sharp area. Preset convergence criteria parameters The value ranges from 0 to 1. The larger the value, the higher the convergence required to enter the sharp zone, and the more units remain in the transition state. Its specific value can be set empirically based on the confidence requirements for distinguishing between finished products and bare land. The connected domain extraction of the above three types of units all adopt the eight-adjacency criterion, that is, a spatial unit and its eight adjacent units (up, down, left, right, and four diagonals) are considered connected. The connected regions of the transition zone unit constitute the gradual transition zone, the connected regions of the paved sharp unit constitute the paved sharp zone, and the connected regions of the unpaved sharp unit constitute the unpaved sharp zone.

[0061] It should be noted that attribution tension and attribution convergence constitute a pair of complementary criteria for spatial units within and outside the transition zone: units with high attribution tension and low attribution convergence have balanced and equal attribution degrees, falling within the gradual transition zone; units with high attribution convergence have one high and one low attribution degree, falling within the sharp zone, and further converge to either the paved or unpaved sharp zone based on the larger of the paved and unpaved attribution degrees. However, in actual images, there may be instances where the attribution tension is greater than... And it does not satisfy the requirement that the degree of convergence of the membership is greater than 1. Spatial units, such as two edge units with low and similar affiliation levels, fall neither into the gradual transition zone nor any sharp zone. To ensure complete and comprehensive area division, a proximity-based affiliation rule is adopted for these spatial units: their paved affiliation level is compared with their unpaved affiliation level. If the paved affiliation level is not less than the unpaved affiliation level, it is assigned to the paved sharp zone; otherwise, it is assigned to the unpaved sharp zone. After the above processing, all spatial units within the process area are classified into one of the three categories: gradual transition zone, paved sharp zone, or unpaved sharp zone, achieving a complete three-part division of the spatial units within the process area.

[0062] Using the lawn example, preset tension judgment parameters are adopted. Preset convergence criteria parameters The unit located deep within the finished turf has an assigned occupancy rate of approximately 0.95 for laid turf and approximately 0.05 for unlaid turf, with an assigned occupancy tension of approximately 0.0475, which is less than... And the convergence degree of the association is approximately 0.90, which is greater than... The paved areas have a higher attribution degree than the unpaved areas, therefore they are classified as paved sharp zones. Units deep within bare land have an attribution degree of approximately 0.05 for paved areas and approximately 0.95 for unpaved areas, with an attribution tension of approximately 0.0475 and an attribution convergence of approximately 0.90, thus they are classified as unpaved sharp zones. Units near the aforementioned front with an attribution degree of approximately 0.60 for paved areas and approximately 0.40 for unpaved areas have an attribution tension of approximately 0.24, which is greater than 0 but does not exceed the upper limit of 0.25, therefore they are classified as gradual transition zones. If there are edge units within the plot with an attribution degree of approximately 0.40 for paved areas and approximately 0.20 for unpaved areas, their attribution tension is approximately 0.08, less than... The convergence degree of belonging is approximately 0.20, which is less than... If the paved area has a higher degree of paving, it will be classified as a paved sharp area according to the nearest paving rule. After processing all units according to the above criteria and extracting the eight adjacent connected domains, the plot is divided into a paved sharp area on one side, an unpaved sharp area on the other side, and a strip-shaped gradual transition zone extending along the front between the two.

[0063] The above scheme uses the scalar measure of attribution tension, which is half-complete, to explicitly separate the transition zone from the sharp zone, obtaining the attribution tension and bilateral attribution degree of each unit in the gradual transition zone, as well as the paved and unpaved sharp zones, providing a regional basis for locating the frontal line and quantifying the continuous paving degree in the transition zone.

[0064] Step S4: Within the gradient transition zone, connect the spatial units located between the paved and unpaved sharp areas to form a paved frontal line. Calculate the frontal normal from the unpaved side to the paved side at each point on the paved frontal line, and calculate the width of the transition zone along the frontal normal from the boundary of the unpaved sharp area to the boundary of the paved sharp area.

[0065] The process of connecting spatial units located between paved and unpaved sharp zones into a paving front line includes: connecting and aggregating spatial units within the transition zone whose absolute difference between paved and unpaved affiliations is no greater than a preset balance tolerance, based on their spatial adjacency, to obtain a paving front line marking the current spatial position of the paving advancement. The determination of this front candidate unit can be expressed as:

[0066] ;

[0067] In the formula, For the set of candidate frontal units, It is a gradual transition zone. , They are spatial units The degree of belonging to those who have been paved and those who have not been paved. To preset the balance tolerance, its value is between 0 and 1. The smaller the value, the more concentrated the frontal candidates are at the two points where the two attribution degrees are strictly equal; the thinner the frontal line, the more likely it is to be broken. The larger the value, the wider the frontal candidate zone. Its specific value can be determined empirically based on the width of the balance zone between the two degrees of belonging within the transition zone. The reason for selecting units where the absolute value of the difference between the two degrees of belonging is no greater than the balance tolerance as frontal candidates is that, within the gradual transition zone, the point where the paved and unpaved degrees of belonging are exactly equal or closest is precisely the critical position where paving transitions from completed to incomplete, i.e., the current spatial position of paving advancement. Candidate unit clusters are obtained by extracting the frontal candidate unit set using eight-adjacency connections as connected components. Since candidate unit clusters usually have a certain width, morphological refinement is performed on each candidate unit cluster, i.e., layer-by-layer stripping of the boundary units of the connected regions until it narrows to a centerline of unit width, thus reducing it to a spatial curve of unit width, i.e., a paving frontal line. Considering that the skeleton obtained after thinning in actual images may contain short spur branches, forks, or loops, resulting in some units having more than two eight-neighbor skeleton neighbors and thus not being able to directly form a single path from one end to the other, the skeleton is pruned after thinning. Short branches with a length less than a preset spur length threshold are removed, and the longest trunk path is retained as the spreading frontal line. Then, the trunk curve is traced along eight adjacent curve units from one end of the curve to the other end, assigning each point on the curve an increasing frontal arc length along the direction. .

[0068] In the actual process area, the spreading front may not be continuous everywhere. This step is handled according to the following cases: when the connection aggregation results in a single connected cluster, it is refined to obtain a spreading front line; when the result of the connection aggregation is multiple unconnected connected curves, each connected curve is determined as a spreading front line and the corresponding front segment spreading degree is calculated. Subsequent steps are performed on each front line separately; when a candidate unit cluster cannot be connected into a continuous unit width curve after refinement, that is, when the refinement result is an isolated point, it is switched to the bottom branch.

[0069] The catch-all branch works as follows: When the results of connected aggregation cannot be chained into a continuous curve, the spread frontal line is not calculated for the discrete spatial unit region. Instead, the average value of the affiliation tension of each spatial unit within the region is used as the local spread of the discrete spatial unit region, and this local spread is used as the spread of the discrete spatial unit region in the spread map for assembly. The reason for using the average value of the affiliation tension within the region as the local spread for discrete patches that cannot form a frontal line is that although such patches do not have a continuously advancing frontal structure, their affiliation tension still reflects the overall semi-completeness of the units within the region simultaneously affiliating to two categories. Taking the average value means using this overall semi-completeness to approximately characterize the spread completion of the patch, thus allowing discrete patches to be included in the subsequent spread map.

[0070] Using the lawn example, a preset balance tolerance is adopted. Within the gradual transition zone, units with an absolute difference between the paved and unpaved ...

[0071] Based on the obtained frontal line, this step further calculates the frontal normal. The calculation process for the frontal normal includes: obtaining the tangent at each point on the frontal line; taking the direction orthogonal to the tangent as the frontal normal for that point; comparing the paved area assignments of the neighboring areas on both sides of the point; determining the side with a higher paved area assignment than the other side as the pointing side of the frontal normal, so that the frontal normal points from the unpaved side to the paved side. Specifically, for each point on the frontal line, a tangent fitting window is taken with a fixed arc segment on each side of the frontal line centered at that point. A linear fitting is performed on the field coordinates of the curve unit within the window, and the direction of the resulting fitted line is the tangent at that point; two directions orthogonal to the tangent are selected as candidate normals; and the neighboring areas along each of the two candidate directions are compared at that point. Considering that the paving front line is located precisely where the two membership degrees are nearly equal, the difference in the paving membership degrees of individual neighboring units adjacent to the front line may be very small, making them susceptible to noise interference and leading to incorrect normal orientation. To ensure the stability and uniqueness of the orientation, multiple spatial units are selected from near to far along each of the two candidate directions, and the average value of their paving membership degrees is calculated and compared. The direction corresponding to the side with the larger average paving membership degree is determined as the front normal. The direction of the front is thus directed, causing the frontal normal to point from the unpaved side to the paved side, and... Normalized to a unit vector. The direction of the frontal normal is used to determine the positive and negative directions of the subsequent gradient along the frontal normal. A tangential linear fit and orthogonal normal selection are used as the defined geometric implementation.

[0072] Using the lawn example, for a point on the frontal line, take the tangential fitting window of several units on both sides to fit the tangential direction of the point along the frontal direction as the transverse direction of the plot, and the two directions orthogonal to it as the two longitudinal directions of the plot. Take multiple units from near to far along both sides of the longitudinal direction to calculate the average degree of paved area. The average degree of paved area on the side of the finished lawn is significantly greater than that on the bare ground side. Therefore, take the longitudinal direction pointing to the side of the finished lawn as the frontal normal, which points from the unpaved side to the paved side. After normalization, it is taken as the unit frontal normal of the point.

[0073] Based on the frontal normal, this step also calculates the transition zone width. For each frontal point, the transition zone width is calculated using the frontal arc length. Starting from this point, along the normal direction of the front In the forward direction, sampling points are taken at equal intervals in the normal direction, and the first point that falls into the already paved sharp area is selected. The signed normal distance between the sampling point and the starting point is used as ; Take sampling points at equal intervals along the negative normal direction of the front, and take the first point that falls into the unpaved sharp area. The signed normal distance between the sampling point and the starting point is used as The value is negative. Considering that in actual images, detection along the normal direction may not always fall into the corresponding sharp area—for example, the transition zone might be abnormally wide, there might be a gap or depression where the normal direction points precisely to the sharp area, or the detection might touch the image boundary—to avoid unbounded detection, a maximum search distance for normal detection is preset. This maximum distance can be set proportionally to the scale of the process area. If a point along a certain direction fails to fall into the corresponding sharp area within the maximum search distance, it is considered an invalid point and does not participate in the calculation of the transition zone width and subsequent frontal segment spread. The corresponding gradient transition zone spatial unit is incorporated into the nearest-neighbor assignment process in subsequent symbol consistency checks. The transition zone width is defined as:

[0074] ;

[0075] In the formula, For the arc length of the front The width of the transition zone at that point It is the marked normal distance from the front point along the positive direction of the frontal normal to the boundary of the paved sharp zone. This is the signed, negative normal distance from the front point along the negative direction of the frontal normal to the boundary of the unpaved sharp zone. Because... For positive The difference between the two values ​​is negative. Subtracting the two values ​​gives the total width of the transition zone at the front point, extending along the frontal normal from the boundary of the unpaved sharp area to the boundary of the paved sharp area. Its physical meaning is the lateral thickness of the transition zone at that point, reflecting the spatial span of the transition from incomplete to complete: the wider the transition zone, the smoother the transition from bare to finished, and the more diffuse the paving; the narrower the transition zone, the more concentrated the transition and the sharper the paving front.

[0076] Using the lawn example, for a point on the frontal line, from its position along the positive normal direction... Intermittent point-by-point detection, if the first sample falls into the already laid sharp area on one side of the finished lawn at the 8th sampling point in the forward direction, then... ; if the probe is conducted along the negative direction, and the first unpaved sharp area on the bare ground side is detected at the fourth negative sampling point, then... The width of the transition zone at this point This indicates that the lawn transitioned from bare to finished state over a period of approximately [time period missing]. The width.

[0077] Using the above method, the spreading front line marking the current position of the spreading advance is located within the gradual transition zone, and its front arc length, the front normal from the unpaved side to the paved side of each front point, the width of the transition zone, and the local spreading degree of discrete patches are obtained, which prepares for dividing the isobars along the front normal and characterizing the transition structure band by band.

[0078] Step S5: Divide the gradual transition zone into sequentially arranged normal isobars along the frontal normal. Calculate the zone assignment tension and zone transition gradient for each normal isobar. The zone transition gradient is the rate of change of the paved assignment degree along the frontal normal.

[0079] This step first involves dividing the transition zone into isotropic sections along the normal to the front. For each frontal point, the isotropic section is calculated using the frontal arc length. Simultaneously, mark the frontal point and, along the frontal normal, mark the isotropic step size. interval The area is uniformly divided into sequentially arranged normal contour bands, the first... The center of the normal contour band is Each transition zone unit is calculated by subtracting the coordinates of the corresponding frontal point from its site coordinates, and then calculating the frontal normal. The projected components along the direction fall into the corresponding normal contour band. This division can be expressed as:

[0080] ;

[0081] In the formula, It is a gradual transition zone. Marker front arc length First Normal isovalue band, The normal coordinates of the center of the isotropic zone are: Equivalent values ​​are numbered. This refers to the isovalue step size. The value of is taken as an integer multiple of the spatial unit size and much smaller than the width of the transition zone. This ensures that each isopleth contains a sufficient number of spatial units to stably calculate the mean within the zone, and also ensures that there are enough discrete intervals along the normal direction to guarantee the accuracy of subsequent Riemann sums and integrals. Through this division, the gradual transition zone is organized into a sequence of normal isopleths indexed by the frontal arc length and normal coordinates. Each isopleth corresponds to a thin layer along the frontal normal direction within the transition zone.

[0082] After obtaining the sequence of normal isotropic zones, the zone assignment tension is calculated for each normal isotropic zone. For each normal isotropic zone, the arithmetic mean of the assignment tensions of all spatial cells falling into that zone is taken as the zone assignment tension. :

[0083] ;

[0084] In the formula, For the arc length of the front First The tension assigned to the normal isotropic band. This is the set of spatial units contained in the isopleth band. This represents the number of spatial units within the isobaric band. spatial unit The attribution tension. The attribution tension characterizes the degree of partial completion of the normal thin layer as a whole: in the thin layer near the unpaved or paved side of the transition zone, the two attribution degrees are one high and one low, and the product is small, resulting in a lower attribution tension; in the thin layer in the middle of the transition zone where the two attribution degrees are nearly equal, the product is the largest, and the attribution tension is the highest.

[0085] After obtaining the sequence of normal contour zones, the transition gradient of each normal contour zone is calculated. The transition gradient represents the rate of change of the paved membership degree along the frontal normal. The calculation process of the transition gradient includes: calculating the mean of the paved membership degree within each normal contour zone, and then calculating the rate of change of the mean of the adjacent normal contour zones along the frontal normal to obtain the transition gradient, which characterizes the sharpness of the frontal advance at that normal position. Specifically, the arithmetic mean of the paved membership degree within each contour zone is first calculated. Then, by forward differencing the mean of the adjacent isobars along the normal coordinate system and dividing by the isobar step size, the transition gradient of the band is obtained. :

[0086] ;

[0087] In the formula, For the arc length of the front First Mean value of paved area within the normal isotropic strip. spatial unit The degree of ownership that has been established This is the band transition gradient of the isopleth band. This represents the mean of the paved affiliation within the next isotropic zone adjacent to the frontal normal. The isoband step size. The band transition gradient characterizes the rate at which the paved attribution increases along the frontal normal from the unpaved side to the paved side, i.e., the sharpness of the frontal advance: the larger its absolute value, the steeper the change in the paved attribution at that normal position, the sharper the front and the more concentrated the transition; the smaller its absolute value, the gentler the transition and the more diffuse the front. Since the band transition gradient is forward-differentiated along the frontal normal, and the frontal normal has been defined in step S4 as pointing from the unpaved side to the paved side, the positive and negative directions of the band transition gradient are thus determined, with a positive value corresponding to a positive increase in the paved attribution along the normal. Since the forward difference requires the use of the next adjacent isoband, the last isoband in the normal direction, since there is no next isoband, is supplemented by the backward difference between it and the previous isoband, thus ensuring that each isoband has a definite band transition gradient.

[0088] Using the lawn example, take the step size of the isopleth strip. That is, each isobar is two spatial units wide, so the aforementioned transition band width At the point of the front, its range Divided evenly into 7 thicknesses The normal isopleths are defined, with the normal coordinates of the center of each isopleth being -0.4, -0.2, 0, 0.2, 0.4, 0.6, and 0.8 (in meters). The point with normal coordinate 0 represents the thin layer containing the frontal point. Along the normal from the unpaved side to the paved side, the average paved attribution within each isopleth isopleth is approximately 0.16, 0.33, 0.52, 0.63, 0.73, 0.83, and 0.91, showing a monotonically increasing trend; the average unpaved attribution within each isopleth is approximately 0.80, 0.63, 0.46, 0.37, 0.27, 0.18, and 0.09, showing a monotonically decreasing trend. It can be seen that at the frontal point thin layer with normal coordinate 0, the average paved attribution is approximately 0.52, and the average unpaved attribution is approximately 0.46, the two being closest, consistent with the positioning of the paved frontal line at the point of equilibrium between the two attributions. Accordingly, the attribution tensions (the product of the average attribution degree of paved areas and the average attribution degree of unpaved areas) of each zone are approximately 0.128, 0.208, 0.239, 0.233, 0.197, 0.149, and 0.082, respectively. The peak value of approximately 0.239 is obtained near the thin layer at the front point at the normal coordinate 0, and does not exceed the upper limit of 0.25, which is consistent with the fact that the two attribution degrees are closest at the front point. By forward difference of the mean paved attribution of adjacent zones and dividing by 0.2m, the zonal transition gradients of the first six isozones are approximately 0.85, 0.95, 0.55, 0.5, 0.5, and 0.4, respectively. The last isozone is supplemented by backward difference with the previous isozone, resulting in a gradient of approximately 0.4. Thus, the zonal transition gradients of the seven isozones are approximately 0.85, 0.95, 0.55, 0.5, 0.5, 0.4, and 0.4, respectively. All zonal transition gradients are positive, reflecting a stable increase in paved attribution along the normal direction at this frontal point.

[0089] The above scheme organizes the gradient transition zone into a series of normal isobars along the frontal normal, and obtains the zone assignment tension characterizing the degree of thin layer semi-completion and the zone transition gradient characterizing the sharpness of thin layer advancement for each zone, providing a quantitative description along the normal for assembling the frontal gradient profile.

[0090] After obtaining the transition gradient and before determining the frontal segment spread, this method also performs a sign consistency check to identify valid frontal segments. For each point on the spread frontal line, the number of normal isopleths with positive transition gradients in the corresponding frontal segment is counted as the number of positive gradient bands, and the number of normal isopleths with negative transition gradients is counted as the number of negative gradient bands. When the number of positive gradient bands is greater than the number of negative gradient bands, the dominant sign of the frontal segment is determined to be positive; when the number of negative gradient bands is greater than the number of positive gradient bands, the dominant sign of the frontal segment is determined to be negative. The number of normal isopleths with the same sign as the dominant sign in the frontal segment is counted as the number of bands in the same direction. The number of normal isobands with transition gradients whose signs are opposite to the dominant sign is taken as the number of reverse bands. When the difference between the number of unidirectional and antidirectional frontal zones exceeds a preset monotonicity tolerance, the frontal segment corresponding to that point is determined as a valid frontal segment, and the frontal segment spread is determined only for the valid frontal segment. The above criterion can be expressed as:

[0091] ;

[0092] In the formula, For the arc length of the front The number of normal contour bands in the frontal segment with the same transition gradient sign as the dominant sign. The quantity of the normal isobars with signs opposite to the dominant sign. This is a preset monotonicity tolerance, a non-negative integer. Its value can be set according to the total number of normal isopleths within the transition zone; the more isopleths, the larger the value can be. The reason for performing sign consistency verification is that in a true spreading front, the degree of belonging to the spread front should monotonically increase from the unspread side to the spread side along the front normal, and the signs of the transition gradients at each normal position should be consistent. If the signs of the transition gradients in a certain front segment are disordered, and the number of bands in the same direction and the number of bands in opposite directions are close, it indicates that there is no truly monotonically advancing spreading front at that location, but rather it is more likely a pseudo-half-complete region caused by noise or missegmentation. Therefore, it is removed from the effective front segments to avoid interfering with the spread degree calculation, thereby improving the robustness of the spread degree inversion to image noise and missegmentation.

[0093] For the transition zone spatial units corresponding to points not identified as effective frontal segments, and for the transition zone spatial units corresponding to invalid frontal points detected along the normal direction but not falling into the sharp zone, a proximity-based assignment process is adopted: the Euclidean distance from the spatial unit to the boundary of the paved sharp zone and the Euclidean distance from the spatial unit to the boundary of the unpaved sharp zone are calculated. If the Euclidean distance from the spatial unit to the boundary of the paved sharp zone is less than the Euclidean distance from the spatial unit to the boundary of the unpaved sharp zone, the spatial unit is assigned to the paved sharp zone; otherwise, it is assigned to the unpaved sharp zone. Here, the Euclidean distance is the square root of the sum of the squares of the differences between the site coordinates of two points, and the Euclidean distance from a spatial unit to the boundary of a region is the minimum value of the Euclidean distances from the unit to all boundary units of that region. The reason for assigning pseudo-semi-complete zone units to the nearest point on the boundary between the two sharp zones is that such units neither constitute an effective frontal segment nor can they participate in the calculation of frontal segment spread. Therefore, determining their assignment based on whether they are closer to the paved or unpaved side in space is the most direct and reasonable approach.

[0094] Using the lawn example, we take the monotonicity tolerance. This is a small non-negative integer. The transition gradients of the seven normal isotropic bands at the aforementioned frontal point are all positive, with 7 positive gradient bands and 0 negative gradient bands. The dominant sign is positive, and the difference between the number of bands in the same direction (7) and the number of bands in the opposite direction (0) is 7, which is greater than... Therefore, it is determined to be an effective frontal segment, and the subsequent frontal segment spread determination will proceed. If a transitional zone unit within the plot experiences inconsistent paving intensity along the normal direction due to dappled sunlight and turf shadows, and its transition gradient exhibits alternating positive and negative values, with similar numbers of unidirectional and counter-directional zones, and the difference not exceeding [value missing], then [the remaining text appears to be incomplete and requires further context]. If the area is not fully completed, it is considered a pseudo-semi-complete area. For each unit in the transition zone, compare its Euclidean distance to the boundary of the sharp area on the finished lawn side and the boundary of the unfinished sharp area on the bare ground side. Units closer to the finished lawn side are included in the sharp area, and units closer to the bare ground side are included in the unfinished sharp area.

[0095] Step S6: For each point of the spread frontal line, the spread of the continuous frontal segment is determined by the band-attached tension distributed along the frontal normal and the band transition gradient within the width of the transition zone.

[0096] The calculation process of frontal segment spread includes: integrating the zone tension along the frontal normal at each point of the spread frontal line within the width of the transition zone as the main term; integrating the absolute value of the zone transition gradient along the frontal normal within the width of the transition zone to obtain the zone gradient product; weighting the zone gradient product with a preset gradient weight coefficient and summing it with the main term; and then normalizing it with the width of the transition zone to obtain the frontal segment spread.

[0097] Before formally calculating the frontal segment spread, the frontal gradient profile is assembled and the gradients with transitions are normalized. (Using normal coordinates...) Using the axis as the axis, the band assignment tension and band transition gradient of each normal isotropic band are bound to the normal position of that band, and assembled into a frontal gradient profile:

[0098] ;

[0099] In the formula, For the arc length of the front Location and normal coordinates Frontal gradient profile at the location. The tension assigned to the band at this normal position. This represents the transition gradient at that normal position. For each frontal arc length... Obtain a coordinate along the normal direction The binary sequence, traversing the entire frontal arc length, yields the semi-complete distribution structure of the entire front. The frontal gradient profile combines two types of information along the normal direction: the degree of semi-completeness of the transition zone as a whole (with attribution tension) and the sharpness of the frontal advance at that point (with transition gradient). It is the core data object for subsequent calculation of the frontal segment spread. Its physical significance lies in simultaneously characterizing the semi-complete intensity distribution and advance sharpness distribution of the transition zone at that frontal point using a profile along the normal direction.

[0100] Since band belonging tension and band transition gradient are two quantities with different meanings, the band transition gradient needs to be normalized before summing. Integrating the absolute value of the band transition gradient along the width of the transition band yields the band gradient product. The integral along the normal direction is implemented using the isobars of each normal direction as discrete intervals, and the Riemann sum is calculated according to the isobar step size. That is, the values ​​of each isobar are multiplied by the step size and then summed.

[0101] ;

[0102] In the formula, For the arc length of the front Gradient product at point, Normal coordinates The gradient at the point is transitional. , These are the lower and upper integration limits of the transition zone along the normal direction of the front, respectively. For the first The transition gradient of the normal contour band. The step size is the isotropic zone. Taking the absolute value of the zone transition gradient and then integrating it along the normal direction means collecting the cumulative increase of the paved attribution degree along the entire transition zone. Since the zone transition gradient is the rate of change of the paved attribution degree along the normal direction, when the frontal segment advances monotonically (i.e., the zone transition gradient does not change sign), the continuous integration of its absolute value along the normal direction equals the cumulative change of the paved attribution degree from the unpaved thin layer to the paved thin layer. This transforms the rate information of the advance sharpness into a cumulative amount that can be added to the zone attribution tension.

[0103] Based on this, the frontal segment spread is calculated. The integral of the band's assigned tension along the width of the transition zone is used as the principal term. The product of the band gradients is weighted by gradient weighting coefficients and summed with the principal term. This summation is then normalized by the transition zone width to obtain the frontal segment spread. :

[0104] ;

[0105] In the formula, For the arc length of the front The extent of the frontal segment at that point. This is the width of the transition zone at that location. Normal coordinates The tension of belonging at the location, , These are the lower and upper integration limits of the transition zone along the normal direction of the front, respectively. These are the preset gradient weight coefficients. This is the gradient product at that point. For the first The tension assigned to the normal isotropic band. The step size is the isobar. The integral of the main term is also implemented using Riemann sum discretization with each normal isobar as the discrete interval. Frontal segment spread is a dimensionless quantity, ranging from 0 to 1, representing the degree of spread completion in the transition zone where the frontal segment is located. The scientific rationale of this formula lies in: the main term incorporates the overall semi-completeness of the transition zone along the normal direction as the cumulative belt-attached tension along the normal direction, reflecting the total amount of spread in the transition zone; the supplementary term incorporates the cumulative sufficiency of frontal advancement as the belt gradient product, reflecting the sufficiency of the spread from incomplete to complete; the sum of the two is then normalized by the transition zone width, eliminating the influence of varying transition zone widths, allowing the frontal segment spread obtained from different frontal points and transition zones of varying widths to be compared on the same 0-1 scale. Gradient weight coefficient. This is used to adjust the proportion of propulsion sharpness information in the spread. Its value can be between 0 and 1, and is set based on the experience of emphasizing the total amount of semi-completed work and propulsion sharpness in the actual process evaluation of the transition zone completion.

[0106] Using the lawn example, we take the gradient weight coefficients. Regarding the aforementioned transition band width The frontal point, its The band tensions of the normal isotropic bands, 0.128, 0.208, 0.239, 0.233, 0.197, 0.149, and 0.082, are multiplied by the step size, respectively. After summing, the integral of the main term is approximately 0.247; the absolute values ​​of the transition gradients of each band (0.85, 0.95, 0.55, 0.5, 0.5, 0.4, 0.4) are multiplied by the step size. After summing, the product of the gradient is approximately 0.83. This product of the gradient is close to the cumulative increase of 0.75 in the degree of belonging at this frontal point along the normal direction from approximately 0.16 to approximately 0.91. This closeness confirms the physical meaning that the product of the gradient corresponds to the cumulative increase in the degree of belonging during monotonic progression. The slight difference of approximately 0.08 arises from the inclusion of an additional boundary term due to backward difference supplementation of the last isotropic band, which is an approximation effect of the discrete Riemann sum at the boundary. Summing the integral of the main term with 0.3 times the product of the gradient yields approximately... Divide by the width of the transition zone. The frontal segment spread at this point is approximately 0.41, which is between 0 and 1, indicating that the transition zone at this point has advanced to nearly half of its spread completion.

[0107] The above scheme normalizes and integrates the frontal gradient profile that bears the semi-complete strength and propulsion sharpness into a scalar between 0 and 1, thereby obtaining the continuous frontal segment spread at each point of the spread frontal line. This continuously characterizes the degree of spread completion of the transition zone, providing a value for the degree of completion of the transition zone for assembling a spread map that covers the entire area.

[0108] Step S7: Combine the spread of each front segment, the spread sharp area and the unspread sharp area according to their spatial positions to form a spread map. Compare the spread map with the planned spread and output the progress and deviation results of the surface process.

[0109] This step first involves combining the spreadability of each part according to the site coordinate system to create a spreadability map that covers the entire process area. The specific assembly process is as follows: The sharp area has already been paved. Each unit within the area is assigned a completed endpoint value, i.e., the upper end of the spread range is 1; unpaved sharp areas... Each unit within the range is assigned an incomplete endpoint value, i.e., the lower end of the value range, 0; mapping and filling are performed on the transition zone units where the effective frontal segment is located, i.e., for each transition zone unit, the arc length of the frontal segment to which it belongs is determined. (Take the front point corresponding to its normal projection) and normal coordinates According to the normal coordinates of this unit In the interval The position in the middle is assigned a transitional value between the incomplete and completed end values, and the frontal segment spread at that position is used as the basis for the frontal position. As the overall benchmark for assigning values ​​to the transition zone of this frontal segment, the spread of the transition unit under the same frontal arc length monotonically transitions from the incomplete end value to the completed end value along the normal coordinate, and the area-weighted average value of this segment is equal to the spread of the frontal segment. Because it requires both a monotonically transitioning along the normal from the incomplete endpoint to the completed endpoint, and a requirement that the area-weighted mean of this segment equals the given frontal segment spread. Both constraints must be satisfied simultaneously, and the mean obtained by simply linearly interpolating the two endpoints using normal coordinates is generally not equal to the mean. Therefore, an adjustable transition shape assignment method is adopted: the assignment of the transition region along the normal direction is taken as a piecewise linear curve that monotonically increases from the incomplete end value to the completed end value, and is... Find the inflection point of the piecewise linear curve such that the weighted average of the areas covered by the two segments before and after the inflection point is exactly equal to... This ensures that the mean value of the segment equals the frontal segment spread while maintaining a monotonic transition along the normal direction. When the frontal segment spread approaches the endpoint, making it impossible for a single turning point to satisfy the mean constraint while maintaining monotonicity, the number of segments is increased or the shape of the transition curve is adjusted to approximate the mean value. The pseudo-semi-complete area unit is assigned the corresponding sharp area endpoint value based on the nearest assignment result obtained from the sign consistency check; the discrete patch unit is assigned a local spread value based on the bottom branch obtained. This local spread value is the mean of the assignment tension within the area, also falling between 0 and 1, and is under the same spread completion semantic as the frontal segment spread value, allowing it to be combined with the spread values ​​of other parts. The above assignment and transition assignment are implemented with defined spatial assignments. After this combination, each spatial unit within the process area obtains a spread value between 0 and 1, collectively forming a unit-by-unit spread map.

[0110] After obtaining the spreadability map, it is compared with the planned spreadability and the results are output. Specifically, the spreadability map is divided into three zones: the laid summary zone, the transition summary zone, and the unlaid summary zone. The laid summary zone consists of spatial units of the laid sharp areas, the unlaid summary zone consists of spatial units of the unlaid sharp areas, and the transition summary zone consists of spatial units of the gradually transitioning zone where the spreadability of the frontal section is determined. These three zones do not overlap and collectively cover all the spatial units of the planar process area. Here, the transition summary zone... The transition zone, consisting of the effective frontal segment and the discrete patch zone, comprises all transitional spatial units outside of the paved and unpaved sharp zones. Pseudo-complete zone units and transition zone units corresponding to invalid frontal points have already been incorporated into the paved or unpaved sharp zones during sign consistency verification and are not included in the transition summary zone. Furthermore, since step S3 has already incorporated edge units that neither meet the transition zone criterion nor any sharp zone criterion into the sharp zone based on proximity, the paved summary zone, transition summary zone, and unpaved summary zone comprehensively cover all process spatial units without overlap or omission.

[0111] For each of the three zones, the spread of each spatial unit in the spread map is calculated, and the spread is then weighted by area and averaged to obtain the contribution of the spread area. The transition zone has completed its contribution. Contribution to unpaved areas :

[0112] ;

[0113] In the formula, , , These are the contributions made in the paved area, the transition area, and the unpaved area, respectively. , , These are respectively the already paved summary area, the transitional summary area, and the unpaved summary area. Spatial unit in the spread diagram The degree of spread, spatial unit The corresponding site area. It should be noted that in this embodiment, each spatial unit is resampled in step S1 into a uniform, regular spatial grid with equal area. If the value is a constant, then the area-weighted average degenerates into the arithmetic mean; the formula retains... This is to accommodate the general situation where spatial unit areas are not equal. The zoning completion contribution is used to separately present the completion status of the paved area, the transition area being paved, and the unpaved area.

[0114] After obtaining the completion contributions of the three zones, the completion contributions of the paved area, the transition area, and the unpaved area are weighted and combined according to the proportion of each zone's area to the area of ​​the planar process region to form the process paving degree. :

[0115] ;

[0116] In the formula, For the spreadability of the process, , , These represent the total area of ​​the already paved summary area, the transition summary area, and the unpaved summary area, respectively; that is, the sum of the unit areas within each zone. , , These represent the completion contribution of the corresponding partition. The process spread degree characterizes the overall spread completion degree of the entire planar process area. Since the three partitions cover all units without overlap or omission, the composite result is numerically equivalent to the weighted average of the entire area of ​​the spread degree map. This makes the partition completion contribution an independently output intermediate quantity, which can provide both the overall progress of the entire area and the completion status of each partition. This summary of partitions cannot be provided by a single average of the entire area.

[0117] After obtaining the process spread, the schedule spatial dispersion is further calculated to characterize the spatial fluctuations of the schedule. For each cell in the spread diagram, the deviation of its spread relative to the process spread is calculated. The mean square of this deviation is then calculated by weighting the spatial cell areas, and the square root is taken to obtain the schedule spatial dispersion. :

[0118] ;

[0119] In the formula, The schedule space dispersion, spatial unit The area of ​​the site, This represents the spread of this unit in the spread diagram. This refers to the spread of the work process. Schedule spatial dispersion characterizes the spatial fluctuation of the spread at various points relative to the overall schedule: low dispersion indicates similar spread at various points and uniform schedule space; high dispersion indicates large spatial fluctuations in spread and schedule concentration in localized areas. This quantity is obtained by using the spatial fluctuation of each unit value relative to the overall schedule on the spread diagram, which cannot be given by a single mean value for the entire area. When the schedule spatial dispersion... Greater than the discreteness criterion parameter At that time, it indicates that the spatial distribution of progress is uneven and there are local concentrations, which can be used by the construction party for targeted scheduling; dispersion judgment parameter The value can be set between 0 and 1, depending on the project's tolerance for uneven schedule space.

[0120] After obtaining the process spread, the difference between the process spread and the planned spread is used to obtain the schedule deviation. This serves as the result of the progress and deviation of the surface-level process, and the deviation segment is determined to locate the lag point. The progress deviation is: ;

[0121] In the formula, This is the schedule deviation. For the spreadability of the process, This refers to the spread of the schedule. Schedule deviation represents the degree to which the overall progress of a process is ahead of schedule or behind schedule; a positive value indicates aheadness, and a negative value indicates lag. For processes meeting the spread of schedule criteria... The spatial units are extracted and merged into several connected segments using eight adjacent connected components, which are then used as deviation segments. The deviation segment is a parameter for determining the deviation range. Its value is between 0 and 1 and can be set according to the requirements for defining significant lag. The deviation range is used to locate the spatial landing point that is significantly lagging behind the plan.

[0122] Using the lawn example, let the total area of ​​the paved area be... Approximately the area of ​​the plot Its contribution Take the completed end value of approximately 1.0, and the area of ​​the unpaved total area. Approximately Its contribution Taking the incomplete end value of approximately 0.0, the area of ​​the transition summary region. Approximately Its contribution The frontal segment spread degree of each frontal point in the transition zone is obtained by weighting the area, and is on the same order of magnitude as the frontal segment spread degree of 0.41 at a certain frontal point mentioned above, so it is approximated as 0.41. The completion contribution of the three zones is weighted and synthesized according to the area ratio to obtain the process spread degree. Compare the spreadability of this process with the planned spreadability. Calculate the difference to obtain the schedule deviation. A positive value indicates that the overall progress of the lawn laying process is slightly ahead of schedule. Simultaneously, the weighted root mean square of the relative laying degree of each unit on the spreading degree map is calculated to obtain the spatial dispersion of the progress. If this value is greater than the dispersion judgment parameter, it indicates that there is local concentration of laying progress within the plot, for example, one corner has been completely laid while another large area is still bare. Subtracting the planned laying degree from the spreading degree indicates a progress rate lower than the target. By performing eight-adjacent connectivity extraction on the unit, specific plots that are significantly lagging behind the plan can be located, allowing the construction team to allocate work resources to that section.

[0123] The above scheme summarizes the spread diagram of each spatial unit into process spread, three-zone completion contribution, progress spatial dispersion, progress deviation and deviation segment. The three-zone completion contribution respectively presents the completion status of the three parts: laid, transition, and unlaid. Progress spatial dispersion describes the spatial uniformity of progress. Progress deviation and deviation segment give the deviation of the overall progress relative to the plan and its spatial landing point. Thus, the complete inversion and output of the continuous progress of surface process and its spatial distribution deviation are realized.

[0124] After assembling the spread map, this method can also identify the advancing pattern of the spreading front based on the spatial distribution of the spread map. Specifically, within the transition zone, the spread spatial gradient of each spatial unit of the spread map is calculated. The direction of the spread spatial gradient in the neighborhood of each point on the spreading front line is taken as the advancing direction of the front from the unpaved side to the paved side at that point. The advancing uniformity is obtained by calculating the inner product of the unit direction of the spread spatial gradient and the front normal along each point on the spreading front line and taking the average value along the front arc length. The spread spatial gradient is calculated as follows: within the transition zone, the spatial gradient of the spread map is calculated unit by unit. The center difference of the spread degree between the unit and the adjacent unit is taken along the two orthogonal coordinate axes of the site coordinate system, that is, the value of the positive adjacent unit minus the value of the negative adjacent unit and then divided by twice the unit spacing. The difference between the two axes constitutes the spread spatial gradient. The two components are calculated only within the gradual transition zone because the spread is constant and its spatial gradient is zero in both the paved and unpaved sharp zones, while the spread varies spatially and its gradient is non-zero only within the transition zone. Considering that the central difference requires adjacent cells on both sides, for spatial cells at the boundary of the gradual transition zone, if one of their adjacent cells is not within the transition zone, then that direction uses the single-sided difference between that cell and its adjacent cell on the side within the zone, to avoid cross-zone value contamination of the gradient. The uniformity of advance is calculated by taking the unit direction of the spatial gradient of the spread in the neighborhood of each point along the spreading front line and the normal to the front. The inner product of the two fronts, taken as the arithmetic mean along the arc length of the front, can be expressed as:

[0125] ;

[0126] In the formula, To improve uniformity, For the arc length of the front The unit direction of the spatial gradient of the neighborhood spread. The normal direction of the front at that point. This represents the total number of discrete points along the spread frontal line. The inner product of two unit vectors is the cosine of the angle between them, increasing from zero to one as the directional consistency increases. Advance uniformity characterizes the degree of consistency between the advance direction indicated by the spatial distribution of spread and the geometric normal of the front: the frontal normal originates from the geometric orientation of the frontal line, and the spatial gradient of spread originates from the spatial distribution of spread. These two sources are independent, and their consistency is a non-trivial measure. When the advance uniformity exceeds a preset uniformity judgment parameter... At that time, it was determined that the spreading of the surface in the planar process was even and uniform; the uniformity judgment parameter was used. The value can be between 0 and 1, and can be set according to the requirements for the uniformity of the frontal advance.

[0127] Using the lawn example, within the transition zone, the spread map is subdivided unit by unit along the horizontal and vertical coordinate axes of the plot, and then subdivided by one side at the boundary of the transition zone to obtain the spatial gradient of spread for each unit. In the neighborhood of each point on the frontal line, the direction of the spatial gradient of spread is from the bare ground side with low spread to the finished side with high spread, i.e., the direction of frontal advancement. The inner product of the unit direction of the spatial gradient of spread and the frontal normal is taken at each point on the frontal line, and the average value is taken along the arc length of the front to obtain the uniformity of advancement. Since the front in this example extends horizontally and straight along the plot and the normal direction is consistent everywhere, the direction of spread gradient is highly consistent with the frontal normal, and the uniformity of advancement is close to 1, which is greater than the uniformity judgment parameter, indicating that the frontal advancement of the lawn is uniform. If a certain part of the front protrudes or sinks due to uneven operation, the direction of spread gradient at that point deviates from the frontal normal, and the uniformity of advancement decreases accordingly. Based on this, the frontal section with uneven advancement can be identified.

[0128] This embodiment takes a drone inspection of a lawn laying process as an example, demonstrating the entire process from aligning the image and the planned spread input, through membership extraction, region segmentation, frontal line positioning, normal isopleth segmentation, frontal gradient profile assembly and spread integration, spread map stitching, to outputting the process spread, zonal completion contribution, progress spatial dispersion, progress deviation, deviation segment, and spread advancement pattern. Compared with the traditional progress recognition method that divides the image into two parts and then calculates the area ratio, this solution retains the semi-complete semantics and explicitly quantifies the continuous spread of the transition zone, making the progress recognition results of the planar process closely match the actual frontal advancement on site. It also provides the spatial distribution uniformity of progress and the lag landing point, providing a more detailed basis for construction scheduling.

[0129] Example 2:

[0130] like Figures 2-3 As shown, the intelligent inspection and analysis system for the progress and quality of landscape construction includes:

[0131] The image acquisition module is used to acquire aligned images and planned spread of the covered process area;

[0132] The attribution extraction module is used to extract the attribution of the laid and unlaid parts of the aligned image on a per-spatial-cell basis. The attribution of the laid and unlaid parts are the tendency strengths of the spatial cell to belong to the completed and uncompleted side categories, respectively.

[0133] The region division module is used to use the product of the paved and unpaved classification as the classification tension that represents each spatial unit simultaneously belonging to both categories. The connected regions of spatial units with classification tension greater than the preset tension judgment parameter are determined as gradual transition zones. Based on the convergence direction of the paved and unpaved classification, the remaining connected regions are divided into paved sharp regions and unpaved sharp regions.

[0134] The frontal analysis module is used to connect spatial units located between paved and unpaved sharp areas into a paved frontal line within a gradual transition zone, calculate the frontal normal from the unpaved side to the paved side at each point on the paved frontal line, and calculate the width of the transition zone along the frontal normal from the boundary of the unpaved sharp area to the boundary of the paved sharp area.

[0135] The iso-band processing module is used to divide the gradual transition zone into sequentially arranged normal iso-bands along the frontal normal, and to calculate the band assignment tension and band transition gradient for each normal iso-band. The band transition gradient is the rate of change of the paved assignment degree along the frontal normal.

[0136] The spreading degree determination module is used to determine the spreading degree of continuous frontal segments at each point of the spreading frontal line within the width of the transition zone by the belt-attached tension distributed along the frontal normal and the belt transition gradient.

[0137] The splicing and comparison module is used to splice the spread of each front segment, the spread sharp area and the unspliced ​​sharp area according to their spatial positions to form a spread map, compare the spread map with the planned spread, and output the progress and deviation results of the surface process.

[0138] Figure 4 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0139] like Figure 4 As shown, the electronic device includes one or more processors and memory.

[0140] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0141] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0142] In one example, the electronic device may also include input devices and output devices, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0143] Of course, for the sake of simplicity, Figure 4 Only some of the components of the electronic device relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device may include any other suitable components depending on the specific application.

[0144] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0145] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent inspection and analysis of the progress and quality of landscape construction, characterized in that, Includes the following steps: Obtain the alignment image and planned spread of the covered process area; For each spatial cell of the aligned image, the laid-up and unlaid-up classifications are extracted, where the laid-up and unlaid-up classifications represent the tendency of the spatial cell to belong to the completed and uncompleted side categories, respectively. The product of the paved and unpaved attribution degrees is used as the attribution tension that characterizes each spatial unit to simultaneously belong to two categories. The connected regions of spatial units whose attribution tension is greater than the preset tension judgment parameter are determined as gradual transition zones. The remaining connected regions are divided into paved sharp regions and unpaved sharp regions according to the convergence direction of the paved and unpaved attribution degrees. Within the gradient transition zone, the spatial units located between the paved sharp area and the unpaved sharp area are connected in series to form a spreading frontal line. The frontal normal direction from the unpaved side to the paved side at each point of the spreading frontal line is obtained, and the width of the transition zone along the frontal normal direction from the boundary of the unpaved sharp area to the boundary of the paved sharp area is obtained. The gradient transition zone is divided into sequentially arranged normal isobars along the frontal normal. The belt assignment tension and belt transition gradient are calculated for each normal isobar. The belt transition gradient is the rate of change of the paved assignment degree along the frontal normal. For each point of the spread frontal line, the continuous frontal segment spread is determined by the band-attached tension distributed along the frontal normal and the band transition gradient within the width of the transition zone. The spread of each frontal segment, the spread sharp area and the unspread sharp area are combined according to their spatial positions to form a spread map. The spread map is compared with the planned spread to output the progress and deviation results of the surface process.

2. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, The process of dividing sharp areas into paved and unpaved areas includes: For spatial units outside the gradient transition zone, the absolute value of the difference between the paved and unpaved membership degrees is taken as the membership convergence degree. If the convergence degree of a spatial unit is greater than the preset convergence judgment parameter and the degree of belonging of the laid unit is greater than the degree of belonging of the unlaid unit, then the connected region of the spatial unit is determined as a laid sharp region. If the convergence degree of a spatial unit is greater than the preset convergence judgment parameter and the unpaved degree of belonging is greater than the paved degree of belonging, then the connected region of the spatial unit is determined as an unpaved sharp region.

3. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, Connecting the spatial units located between the paved sharp area and the unpaved sharp area to form a paved frontal line includes: Spatial units within the gradient transition zone whose absolute value of the difference between the laid and unlaid affiliations is not greater than a preset balance tolerance are connected and clustered according to their spatial adjacency relationship to obtain a laying front line that marks the current spatial position of the laying advancement. When the result of the connectivity aggregation is multiple unconnected connected curves, each connected curve is determined as a spreading frontal line and the corresponding frontal segment spreading degree is calculated. When the results of the connectivity aggregation cannot be chained into a continuous curve, the spreading front line is not obtained for the discrete spatial unit region. Instead, the average value of the belonging tension of each spatial unit in the region is used as the local spreading degree of the discrete spatial unit region, and the local spreading degree is used as the spreading degree of the discrete spatial unit region in the spreading degree diagram for assembly.

4. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 3, characterized in that, The calculation process for the frontal segment spread includes: For each point of the spreading frontal line, the integral of the belt-assigned tension along the frontal normal within the width of the transition zone is taken as the principal term. The absolute value of the belt transition gradient is integrated along the frontal normal within the width of the transition zone to obtain the belt gradient product. The belt gradient product is weighted with a preset gradient weight coefficient and then summed with the principal term. The result is then normalized with the width of the transition zone to obtain the frontal segment spreading degree. The calculation process of the transition gradient includes: obtaining the average value of the degree of belonging within each normal contour band, and then calculating the rate of change of the average value of the adjacent normal contour bands along the frontal normal to obtain the transition gradient of the band, which characterizes the sharpness of the frontal advance at the normal position.

5. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, After obtaining the transition gradient and before determining the frontal segment spread, the method further includes: For each point of the spread frontal line, the number of normal contour bands with positive transition gradients in the frontal segment corresponding to that point is counted as the number of positive gradient bands, and the number of normal contour bands with negative transition gradients is counted as the number of negative gradient bands. When the number of positive gradient bands is greater than the number of negative gradient bands, the dominant sign of the front segment is determined to be positive; when the number of negative gradient bands is greater than the number of positive gradient bands, the dominant sign of the front segment is determined to be negative. The number of normal isobars with the same sign as the dominant sign in the frontal segment is taken as the number of bands in the same direction, and the number of normal isobars with the opposite sign to the dominant sign is taken as the number of bands in the opposite direction. When the difference between the number of bands in the same direction and the number of bands in the opposite direction is greater than the preset monotonicity tolerance, the frontal segment corresponding to that point is determined as an effective frontal segment, and the frontal segment spread is determined only for the effective frontal segment. For the spatial unit of the gradient transition zone corresponding to the point that is not determined as an effective frontal segment, calculate the Euclidean distance from the spatial unit to the boundary of the paved sharp zone and the Euclidean distance from the spatial unit to the boundary of the unpaved sharp zone. If the Euclidean distance from the spatial unit to the boundary of the paved sharp zone is less than the Euclidean distance from the spatial unit to the boundary of the unpaved sharp zone, the spatial unit is assigned to the paved sharp zone; otherwise, the spatial unit is assigned to the unpaved sharp zone.

6. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, The calculation process for the frontal normal includes: The tangent of each point on the spread frontal line is obtained, and the direction orthogonal to the tangent is taken as the frontal normal of the point. The degree of paving belonging of the neighboring areas on both sides of the point is compared. The side with the paving belonging degree greater than the paving belonging degree of the neighboring area on the other side is determined as the pointing side of the frontal normal, so that the frontal normal points from the unpaved side to the paved side. The direction of the frontal normal is used to determine the positive and negative directions of the gradient along the frontal normal.

7. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, The spreadability diagram is compared with the planned spreadability, and the results of the surface process progress and deviation are output, including: The spreadability map is divided into three zones: the spread summary zone, the transition summary zone, and the unspread summary zone. The spread summary zone is composed of spatial units of the spread sharp zone, the unspread summary zone is composed of spatial units of the unspread sharp zone, and the transition summary zone is composed of spatial units of the gradual transition zone that determines the spreadability of the frontal segment. The three zones do not overlap and together cover all the spatial units of the planar process area. For the three partitions, the spread of the spatial unit in the spread diagram is calculated for each spatial unit within the partition, and the average is calculated by weighting the spatial unit area to obtain the completed contribution of the paved area, the completed contribution of the transition area, and the completed contribution of the unpaved area, respectively. The completed contribution of the paved area, the completed contribution of the transition area, and the completed contribution of the unpaved area are weighted and combined according to the proportion of the area of ​​each zone to the area of ​​the planar process area to obtain the process spread degree. The difference between the process spread degree and the planned spread degree is used to obtain the progress deviation. The progress deviation is used as the progress and deviation result of the planar process.

8. The intelligent inspection and analysis method for the progress and quality of garden construction according to claim 1, characterized in that, After combining the data into a spreadability map, the method further includes: Within the gradient transition zone, the spread degree spatial gradient of each spatial unit of the spread degree map is obtained. The direction of the spread degree spatial gradient in the neighborhood of each point on the spread front line is taken as the frontal advancement direction from the unpaved side to the paved side at that point. The advancement uniformity is obtained by calculating the inner product of the unit direction of the spread degree spatial gradient and the frontal normal along each point on the spread front line and taking the average value along the frontal arc length. When the uniformity of the advance is greater than the preset uniformity judgment parameter, it is determined that the surface process spreads the front edge evenly.

9. A smart inspection and analysis system for the progress and quality of landscape construction, characterized by: The intelligent inspection and analysis method for the progress and quality of garden construction as described in any one of claims 1-8 includes: The image acquisition module is used to acquire aligned images and planned spread of the covered process area; The attribution extraction module is used to extract the laid attribution and unlaid attribution for each spatial cell of the aligned image. The laid attribution and the unlaid attribution are the tendency strengths of the spatial cell to belong to the completed side category and the uncompleted side category, respectively. The region division module is used to take the product of the paved and unpaved attribution as the attribution tension that characterizes each spatial unit to simultaneously belong to two categories, to determine the connected regions of spatial units whose attribution tension is greater than a preset tension judgment parameter as a gradual transition zone, and to divide the remaining connected regions into paved sharp regions and unpaved sharp regions according to the convergence direction of the paved and unpaved attribution. The frontal analysis module is used to connect the spatial units located between the paved sharp area and the unpaved sharp area into a paved frontal line within the gradient transition zone, calculate the frontal normal from the unpaved side to the paved side at each point of the paved frontal line, and calculate the width of the transition zone from the boundary of the unpaved sharp area to the boundary of the paved sharp area along the frontal normal. The iso-band processing module is used to divide the gradual transition zone into sequentially arranged normal iso-bands along the frontal normal, and to calculate the band assignment tension and band transition gradient for each normal iso-band. The band transition gradient is the rate of change of the paved assignment degree along the frontal normal. The spreading degree determination module is used to determine the spreading degree of a continuous front segment at each point of the spreading front line within the width of the transition zone by the zone-assigned tension distributed along the normal direction of the front and the zone transition gradient. The splicing and comparison module is used to splice the spread of each front segment, the spread sharp area and the unspliced ​​sharp area according to their spatial positions to form a spread map, compare the spread map with the planned spread, and output the progress and deviation results of the surface process.

10. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method as described in any one of claims 1 to 8.

Images