A water ditch cable trough hoisting integrated machine construction method and equipment

Abstract

The present application relates to the technical field of image analysis, and particularly relates to a water ditch cable groove hoisting all-in-one machine construction method and equipment. Binocular images at each moment in the construction process are obtained, and registration is performed based on the gray level and position characteristics of the pixel points to generate a high-precision depth map; a monocular image is divided to obtain a to-be-processed region, and cross-moment matching is realized using region similarity to obtain a matching region set, which is used to accurately track concrete phase change dynamics and establish a quantitative benchmark for demolding decision-making; by analyzing the changes in the number of pixels, position and depth value between the to-be-processed regions in the matching region set, the template region and the pouring piece region are effectively distinguished in the monocular image at the current demolding moment, the concrete splashing interference is eliminated, and the region recognition accuracy is improved; based on the position difference characteristics of the pixel points between the template region and the pouring piece region, adhesion risk assessment is realized, and the demolding vibration parameters are dynamically regulated to ensure the demolding quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image analysis, in particular to a water ditch cable trench hoisting all-in-one machine construction method and equipment. BACKGROUND

[0002] The water ditch cable trench is an integrated multifunctional concrete ditch structure in tunnel engineering, which is usually longitudinally arranged along the bottom of both sides of the tunnel or the center, and is mainly used for centralized pipeline laying, drainage and flood control, and safe operation of equipment. When the water ditch cable trench in the tunnel is constructed, the water ditch cable trench hoisting all-in-one machine is usually used to realize the installation and construction of the water ditch cable trench, including trolley positioning, formwork lifting and installation, concrete pouring, demolding, and moving, etc. In the demolding process, adhesion is prone to occur. In order to ensure the construction quality, the accuracy of demolding needs to be improved.

[0003] In the prior art, when the demolding process is monitored during construction, the template region and the pouring piece region are usually segmented based on the gray scale gradient information of the pixel points, so as to judge the adhesion phenomenon in the demolding process. However, since concrete stains may splash during pouring, if the region segmentation is only based on the gray scale information, the pouring piece region may be inaccurately segmented, which affects the accuracy of adhesion phenomenon analysis, and finally affects the subsequent demolding process. SUMMARY

[0004] In order to solve the technical problem that concrete stains may splash during pouring, if the region segmentation is only based on the gray scale information, the pouring piece region may be inaccurately segmented, which affects the accuracy of adhesion phenomenon analysis, and finally affects the subsequent demolding process, the purpose of the present application is to provide a water ditch cable trench hoisting all-in-one machine construction method and equipment, and the technical solution adopted is as follows:

[0005] A water ditch cable trench hoisting all-in-one machine construction method, comprising:

[0006] Obtaining binocular images at each time during construction, wherein the time includes the current demolding time and a plurality of historical pouring times;

[0007] In the binocular images at each time, registration is performed based on the gray scale change and position distribution characteristics of the pixel points in the binocular images, and a depth map is obtained;

[0008] At each historical pouring time, the regions are divided based on the position distribution of the pixel points in the monocular images, and a to-be-processed region is obtained; the to-be-processed regions are matched based on the position distribution similarity of the pixel points between the to-be-processed regions in the monocular images at different historical pouring times, and a matching region set is obtained;

[0009] In all matching region sets, the number of pixel points in the to-be-processed region, the position change of the pixel points and the depth value change of the pixel points are analyzed to determine the template region and the cast part region in the monocular image at the current demolding time.

[0010] In the monocular image at the current demolding time, the demolding adhesion degree at the current demolding time is determined based on the position difference features of the pixel points between the template region and the cast part region; and the demolding adjustment is performed based on the demolding adhesion degree at the current demolding time.

[0011] Further, the depth map acquisition method comprises:

[0012] In the binocular image at each time, the edge lines and the gray gradient of each edge pixel point in each monocular image are obtained based on the Canny operator.

[0013] The edge pixel points in the left eye image are taken as reference points, and the edge pixel points in the right eye image are taken as contrast points, the Euclidean distance between each reference point and each contrast point is calculated as a distance factor, and the product of the absolute value of the difference between the gray gradients of each reference point and each contrast point and the distance factor is negatively correlated and normalized to obtain a first matching factor.

[0014] The curvature value of each edge pixel point on each edge line in each monocular image is obtained, and the ratio of the mean value of the curvature values between each reference point and each contrast point to the absolute value of the curvature value difference is taken as a second matching factor.

[0015] The product of the first matching factor and the second matching factor between each reference point and each contrast point is normalized to obtain a feature matching probability between each reference point and each contrast point.

[0016] In all feature matching probabilities corresponding to each reference point, the contrast point and the reference point corresponding to the feature matching probability exceeding the preset matching threshold and being the largest are taken as a matching point pair.

[0017] In each group of binocular images, based on the matching point pair, the left and right eye images are overall registered by using RANSAC, so as to obtain a depth map.

[0018] Further, the to-be-processed region acquisition method comprises:

[0019] In the left eye image at each historical casting time, edge detection and morphological operation are performed based on the Canny operator to obtain all closed regions as to-be-processed regions.

[0020] Further, the matching region set acquisition method comprises:

[0021] In the left-eye images at each of two adjacent historical pouring moments, the area to be processed in the left-eye image at the previous historical pouring moment is used as the reference area, and the area to be processed in the left-eye image at the next historical pouring moment is used as the comparison area;

[0022] In each reference area and each comparison area, the value after negative correlation mapping of the Euclidean distance between the centroids is used as the distance proximity parameter, and the positions of the pixels in the reference area are compared with the positions of the pixels in the comparison area, and the number of pixels with the same position is used as the overlap factor;

[0023] The normalized value of the product of the distance parameter between each reference area and each comparison area and the overlap factor is used as the matching index;

[0024] Among all the matching indices corresponding to each reference area, the comparison area corresponding to the maximum matching index is used as the matching area of ​​the reference area;

[0025] In the left-eye images at all historical pouring moments, the reference areas and the corresponding matching areas are grouped together to obtain all matching area sets.

[0026] Furthermore, determining the template area and the casting area in the monocular image at the current demoulding moment includes:

[0027] In all matching area sets, analyze the fluctuation differences in the number of pixels and the depth values ​​between the areas to be processed, and divide all matching area sets into template area sets and deformation area sets;

[0028] In the deformation area set, the number of pixel points and the depth value change trend of the area to be processed are analyzed, and the casting area set is screened out from all deformation area sets based on the position change of the pixel points;

[0029] The area to be processed in the left image at the last historical casting moment in the template area set and the area corresponding to it in the left image at the current demoulding moment are used as the template area, and the area to be processed in the left image at the last historical casting moment in the casting part area set and the area corresponding to it in the left image at the current demoulding moment are used as the casting part area.

[0030] Furthermore, dividing all matching area sets into template area sets and deformation area sets includes:

[0031] In each set of matching region sets, a difference between a maximum value and a minimum value of the pixel point number of all to-be-processed regions is taken as a first number range, and a product of the first number range and a variance of the pixel point number of all to-be-processed regions is subjected to a negative correlation mapping process to obtain a first stability factor;

[0032] In each set of matching region sets, edge pixel points in each to-be-processed region are obtained based on a Canny operator, a difference between a maximum value and a minimum value of the number of edge pixel points in all to-be-processed regions is taken as a second number range, and a value obtained by subjecting a product of the second number range and a variance of the number of edge pixel points in all to-be-processed regions to a negative correlation mapping is taken as a second stability factor;

[0033] A value obtained by subjecting a variance of the depth value of the pixel point in all to-be-processed regions to a negative correlation mapping is taken as a third stability factor;

[0034] A value obtained by normalizing a product of the first stability factor, the second stability factor and the third stability factor is taken as a template index;

[0035] In all matching region sets, a matching region set corresponding to a maximum template index is taken as a template region set, and the remaining matching region sets are taken as morphing region sets.

[0036] Further, the method for obtaining the pouring member region set comprises:

[0037] In each set of morphing region sets, all to-be-processed regions are arranged in a time sequence of historical pouring time to obtain a sorting sequence, a first-order difference sequence of the pixel point number of the to-be-processed regions in the sorting sequence is calculated as a number change sequence, and continuous positive values in the number change sequence are taken as number increase segments to obtain all number increase segments;

[0038] In the sorting sequence, a first-order difference sequence of the mean value of the depth value of the pixel point of the to-be-processed region is calculated as a depth change sequence, and continuous positive values in the depth change sequence are taken as depth increase segments to obtain all depth increase segments;

[0039] A value obtained by normalizing a product of the mean value of the length of all number increase segments and the mean value of the length of depth increase segments is taken as a first pouring member factor;

[0040] In the sorting sequence, a Euclidean distance between the centroid of each to-be-processed region and the centroid of the template region in the monocular image is taken as a distance coefficient, a first-order difference sequence of the distance coefficient of all to-be-processed regions is calculated as a position change sequence, and a proportion of negative values in the position change sequence is taken as a second pouring member factor;

[0041] normalizing a product of the first casting factor and the second casting factor as a casting probability of each set of deformation region groups;

[0042] deformation region groups with a casting probability greater than a preset casting threshold are taken as a casting region set.

[0043] Further, the acquisition method of the demolding adhesion degree comprises:

[0044] At the current demolding moment, for any pixel point in each casting region, the Euclidean distance between the pixel point and the pixel point in the nearest template region is calculated as a distance reference value, the mean value of the distance reference values of all pixel points in the casting region is taken as a reference value, and the absolute value of the difference between the distance reference value of each pixel point in the casting region and the reference value is normalized as the adhesion probability of the pixel point.

[0045] In each casting region, the pixel point with an adhesion probability greater than a preset adhesion threshold is taken as a demolding adhesion point.

[0046] In each casting region, the demolding adhesion points in the obtained clustering cluster are composed of a region as an adhesion region based on K-means clustering algorithm and a preset K value, wherein the distance measurement is the Euclidean distance between the demolding adhesion points.

[0047] In each adhesion region, the product of the mean value of the adhesion probabilities of all pixel points in the adhesion region and the number of pixel points is normalized as an adhesion index of each adhesion region.

[0048] At the current demolding moment, the mean value of the adhesion indices of all adhesion regions is taken as the demolding adhesion degree at the current demolding moment.

[0049] Further, the demolding adjustment based on the demolding adhesion degree at the current demolding moment comprises:

[0050] The sum of the demolding adhesion degree at the current demolding moment and a preset parameter is taken as an adjustment degree value.

[0051] The product of the adjustment degree value and a preset demolding vibration frequency is taken as a demolding vibration adjustment frequency for demolding.

[0052] A water trench cable slot hoisting all-in-one machine construction equipment, comprising a processor and a memory, the memory stores at least one instruction, at least one program, a code set or an instruction set, at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to realize the steps of a water trench cable slot hoisting all-in-one machine construction method.

[0053] The present application has the following beneficial effects:

[0054] By acquiring binocular images at each moment in the construction process, the pouring and demolding processes are monitored. The feature points are determined and registered based on the gray scale change and position distribution characteristics of the pixel points in the binocular images, and a high-precision depth map is generated, which breaks through the perception limitation of monocular vision on spatial information and helps to provide multi-dimensional information for subsequent demolding adhesion analysis. At the historical pouring time, the pixel position distribution of the monocular image is used to divide the to-be-processed area, and cross-time matching is performed based on the area similarity, so that the dynamic precision tracking of the concrete solidification is realized, and a quantitative basis for demolding judgment is provided. By analyzing the change characteristics of the number, position and depth value of the pixel points in the matching area set, the template area and the pouring piece area are accurately distinguished in the monocular image at the current demolding time, the area misjudgment problem caused by concrete stain splashing is solved, and the demolding adhesion detection accuracy is improved. The demolding adhesion degree is calculated based on the position difference characteristics of the template area and the pouring piece area, the demolding is adjusted in real time according to the demolding adhesion degree, a closed-loop control system of "monitoring-analysis-regulation" is formed, the pouring piece area is accurately determined, and the calculation accuracy of the demolding adhesion degree is improved, so that the demolding quality is finally ensured. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art and the advantages thereof, a brief introduction will be given to the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0056] Figure 1 A method flowchart of a water ditch cable slot hoisting all-in-one machine construction method provided by an embodiment of the present application is shown in the following figure.

[0057] Figure 2 A scene schematic diagram of water ditch cable slot hoisting all-in-one machine construction provided by an embodiment of the present application is shown in the following figure.

[0058] Figure 3 A method flowchart of a template area and pouring piece area determination method in an embodiment of the present application is shown in the following figure.

[0059] Figure 4 An equipment structure schematic diagram of a water ditch cable slot hoisting all-in-one machine construction equipment provided by an embodiment of the present application is shown in the following figure.

[0060] Reference signs: 1 - upper longitudinal beam, 2 - lower longitudinal beam, 3 - upright column, 4 - steering oil cylinder, 5 - horn steering mechanism, 6 - track, 7 - electric hoist, 8 - cycloidal gear reducer, 9 - rubber tire, 10 - formwork, 11 - demolding screw rod, 12 - positioning pressure lever. DETAILED DESCRIPTION

[0061] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined object of the application, the specific implementation, structure, features and effects of the water trench cable trench hoisting integrated machine construction method and equipment according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0063] The specific scheme of the water trench cable trench hoisting integrated machine construction method and equipment provided by the present application is described in detail below in combination with the drawings.

[0064] Please refer to Figure 1 which shows the method flowchart of the water trench cable trench hoisting integrated machine construction method provided by one embodiment of the present application, and the method comprises the following steps:

[0065] Step S1: acquiring binocular images at each moment in the construction process, wherein the moments include the current demolding moment and a plurality of historical pouring moments.

[0066] When the water trench cable trench is constructed, the water trench cable trench hoisting integrated machine is usually used to realize the installation and construction of the water trench cable trench, including trolley positioning, formwork lifting and installation, concrete pouring, demolding, and moving again, etc. Please refer to Figure 2 which shows the scene schematic diagram of the water trench cable trench hoisting integrated machine construction in one embodiment of the present application, including the upper longitudinal beam 1, the lower longitudinal beam 2, the upright column 3, the steering oil cylinder 4, the horn steering mechanism 5, the track 6, the electric hoist 7, the cycloidal gear reducer 8, the rubber tire 9, the formwork 10, the demolding screw rod 11, and the positioning pressure lever 12. Among them, the upper longitudinal beam 1, the lower longitudinal beam 2 and the upright column 3 are used to support the whole trolley main body, the electric hoist 7 can move along the track 6 for hoisting the formwork 10, the cycloidal gear reducer 8 drives the rubber tire 9 to move, the steering oil cylinder 4 drives the horn steering mechanism 5 to steer the trolley main body, and the demolding screw rod 11 and the positioning pressure lever 12 are used for positioning and demolding the formwork 10.

[0067] But in its construction process, when demolding, the adhesion between the formwork and the pouring piece is prone to occur, and in the process of identifying the formwork area and the pouring piece area, if only according to the gray scale information, the area division will be inaccurate, therefore, in the embodiment of the present application, the construction process of the water ditch cable trough hoisting integrated machine is analyzed, the disturbance of the concrete stain to the formwork structure is excluded by combining the image of the pouring process with the change of the time sequence, so as to obtain the accurate area of the concrete pouring piece and the formwork in the image at the current demolding moment, and according to the relative movement of the formwork and the concrete pouring piece in the demolding process, the possibility of local adhesion at the current demolding moment is analyzed, which is used for demolding adjustment.

[0068] Firstly, in the embodiment of the present application, binocular images at each moment in the construction process need to be obtained, the reason for obtaining the binocular images is that they can not only provide gray scale texture and other information in monocular images, but also obtain spatial information, which is helpful to provide multi-dimensional data for subsequent demolding adhesion analysis.

[0069] Specifically, a binocular camera can be installed on the top of the water ditch cable trough hoisting integrated machine (such as the upper longitudinal beam 1), and the image in the construction process is obtained by using the overhead view angle, so as to obtain the binocular images at each moment, wherein the time includes the current demolding moment and a plurality of continuous historical pouring moments, the time interval of the historical pouring moment is 2 seconds, and the specific time interval can be adjusted according to the implementation scene, which is not limited here.

[0070] It should be noted that the installation of the binocular camera needs to ensure that the protection shooting range can completely obtain the images of the formwork and the pouring piece, if a single binocular camera cannot achieve this, a plurality of binocular cameras need to be installed, so as to splice the images as the binocular images at each moment.

[0071] Step S2: In the binocular image at each moment, the registration is carried out based on the gray scale change and position distribution characteristics of the pixel points in the binocular image, and the depth map is obtained.

[0072] In the construction scene of the water ditch cable trough in the tunnel, artificial lighting can cause local strong light or shadow in the image obtained by the binocular camera, and if the image registration is directly carried out, the accurate structure details cannot be obtained, therefore, in the left and right eye images obtained at each moment, the feature points with high geometric feature contribution degree and high stability are selected for the registration of the left and right eye images, so as to obtain the depth map in the construction process, which is used to provide spatial information, wherein the geometric feature contribution degree can be represented by the position distribution information of the pixel points, and the stability is reflected by the gray scale change of the pixel points.

[0073] Preferably, in an embodiment of the present application, the method for obtaining the depth map comprises:

[0074] Firstly, in each binocular image at each time, the edge lines and the gray gradient of each edge pixel point in each monocular image are obtained based on the Canny operator, reflecting the significant structural edge features in the image.

[0075] If the position coordinates of the pixel points in the two monocular images are similar, the probability of being matched points will also increase, so the same coordinate system can be constructed in each monocular image, for example, taking the lower left corner of the image as the origin, and the horizontal right direction and the vertical upward direction as the coordinate axes. Then the edge pixel points in the left eye image are taken as reference points, and the edge pixel points in the right eye image are taken as contrast points. The Euclidean distance between each reference point and each contrast point is calculated as a distance factor. The smaller the distance factor, the greater the probability of being matched points.

[0076] The images obtained by the left and right cameras are images of the same object from different angles, so it is necessary to find pixel points with less angle interference as feature points. Less angle interference means that the gray scale does not change much, so the product of the absolute value of the difference between the gray gradients of each reference point and each contrast point and the distance factor is negatively correlated and normalized, and the value after the mapping is taken as the first matching factor. The greater the first matching factor, the higher the probability of being matched points between the reference point and the contrast point. The negative correlation mapping and normalization here can use the formula wherein, represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0077] The curvature value of each edge pixel point on each edge line in each monocular image is calculated. The curvature value can reflect the bending degree of the edge line. High curvature points often correspond to structural key points, which can improve the registration accuracy as feature points. Therefore, the ratio of the mean value of the curvature value between each reference point and each contrast point to the absolute value of the difference of the curvature value is taken as the second matching factor. The greater the second matching factor, the higher the probability of the reference point and the contrast point being structural key points, and the greater the similarity between them (the smaller the absolute value of the difference of the curvature value), the greater the probability of being matched points.

[0078] The product of the first matching factor and the second matching factor between each reference point and each contrast point is normalized, and the value after the normalization is taken as the feature matching probability between each reference point and each contrast point. Based on the foregoing analysis, the greater the feature matching probability, the higher the matching degree between them and the higher the probability of being feature points. Therefore, among all the feature matching probabilities corresponding to each reference point, the contrast point corresponding to the feature matching probability that exceeds the preset matching threshold and is the largest forms a matched point pair with the reference point. The normalization is a well-known technical means to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0079] Finally, in each set of binocular images, based on the matched point pairs, the left and right images are globally registered by using the RANSAC algorithm, so as to obtain the depth map.

[0080] It should be noted that the RANSAC algorithm is a known technology, and the specific process is not described here; the preset matching threshold is 0.7, and the specific value can be adjusted according to the implementation scene, which is not limited here; the Canny operator is a known technology, and the specific process of obtaining the edge line and the gray gradient is not described here.

[0081] At this point, the depth map corresponding to the binocular image at each time can be obtained, which is used to provide the depth information of each pixel point.

[0082] Step S3: At each historical pouring time, the regions are divided based on the position distribution of the pixel points in the monocular image to obtain the to-be-processed regions; and the to-be-processed regions are matched based on the similar position distribution of the pixel points between the to-be-processed regions in the monocular images at different historical pouring times to obtain a matching region set.

[0083] During the pouring process, the form of the template has spatiotemporal stability, so first, at each historical pouring time, the regions are divided based on the position distribution of the pixel points in the monocular image to obtain the to-be-processed regions, which can divide the whole image into a plurality of local regions, which is helpful for subsequent division of the template region and the pouring part region.

[0084] Preferably, in an embodiment of the present application, the method for obtaining the to-be-processed region comprises:

[0085] In the left image at each historical pouring time, edge detection is performed based on the Canny operator and morphological operation is performed to obtain all closed regions as the to-be-processed regions.

[0086] It should be noted that in this embodiment of the present application, the monocular image is uniformly selected as the left image; the edge detection and morphological operation by the Canny operator are both known technologies, and the specific process is not described here.

[0087] At this point, the to-be-processed region in the left image at each historical pouring time can be obtained.

[0088] In the construction process of the water channel cable trench, usually, the formwork is placed at the specified position by using the trolley, and then the concrete is poured and demoulded, so as to realize the construction of the water channel cable trench. In the pouring process, the concrete stains may be attached to the formwork, forming the areas similar to the formwork or the concrete components in color, texture, etc., which makes it difficult to accurately identify the specific contour of the formwork, and the concrete pouring component and the formwork area cannot be accurately separated, resulting in missed or misjudged local adhesion of the demoulding. In the single pouring construction process, the formwork has spatial and temporal stability, so the analysis is performed based on the time sequence difference of the images obtained in the pouring process, and the formwork area and the pouring component area are identified.

[0089] Before that, the to-be-processed areas in the left-eye images at different historical pouring moments can be matched across moments to obtain a matching area set, which is helpful for subsequent analysis of the time sequence change characteristics.

[0090] Preferably, in an embodiment of the present application, the method for obtaining the matching area set comprises:

[0091] In the pouring process, the to-be-processed areas representing the same part usually do not have obvious displacement at continuous moments, so in the left-eye images at each adjacent two historical pouring moments, the to-be-processed area in the left-eye image at the previous historical pouring moment is taken as a reference area, and the to-be-processed area in the left-eye image at the next historical pouring moment is taken as a comparison area.

[0092] Then, in each reference area and each comparison area, the value obtained by negatively correlating the Euclidean distance between the centroids is taken as a distance proximity parameter. The smaller the Euclidean distance, the greater the distance proximity parameter, and the more similar the centroid positions of the reference area and the comparison area, which can be regarded as the higher the matching degree. The negative correlation mapping here can adopt wherein x represents the independent variable, represents a preset constant, which prevents the denominator from being 0, and can be taken as 0.001; at the same time, the positions (specifically, the coordinates) of the pixel points in the reference area and the pixel points in the comparison area are compared, and the number of the pixel points with the same position is taken as an overlap number factor. The greater the overlap number factor, the greater the possibility that the reference area and the comparison area represent the same part. In this embodiment of the present application, the same coordinate system is constructed in all left-eye images, for example, the lower left corner of the image is taken as the origin, and the horizontal right direction and the vertical upward direction are taken as the coordinate axes, so as to obtain the position coordinates of each pixel point.

[0093] The product of the distance similarity parameter between each reference region and each contrast region and the normalized value of the overlap number factor is taken as a matching index. Based on the foregoing analysis, the greater the matching index, the greater the probability that the reference region and the contrast region represent the same part. The normalization is a technique well known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0094] In all matching indexes corresponding to each reference region, the contrast region corresponding to the maximum matching index is taken as the matching region of the reference region.

[0095] Finally, in the left-eye image at each historical pouring time, the reference region and the corresponding matching region form a set, thereby obtaining all matching region sets. For example, there are three historical pouring times, the to-be-processed region b in the second historical pouring time is the matching region of the to-be-processed region a in the first historical pouring time, and the to-be-processed region c in the third historical pouring time is the matching region of the to-be-processed region b in the second historical pouring time, so a matching region set is obtained, which includes the to-be-processed region a, the to-be-processed region b, and the to-be-processed region c.

[0096] Step S4: In all matching region sets, the number change of the pixel points, the position change of the pixel points, and the depth value change of the pixel points in the to-be-processed region are analyzed, for determining the template region and the pouring piece region in the monocular image at the current stripping time.

[0097] With the progress of the pouring process, the area and the depth of the concrete pouring piece region change to a certain extent, and the change degree is relatively stable. In the pouring process, the concrete flow of the pouring piece may generate vortex-like textures, etc., so that the internal texture fluctuation degree is large. The area and the depth of the template region have no large change (the depth may have a small fluctuation due to vibration in operation), and the internal texture also has no large fluctuation. In the pouring process, there is also concrete stain interference (which may be caused by splashing in the concrete pouring process, and some may slide down with the progress of the construction process). The area and the depth of the stain region change unstably, so the number change of the pixel points, the position change of the pixel points, and the depth value change of the pixel points in the to-be-processed region can be analyzed in the matching region set, the interference of the concrete stain is excluded in all matching region sets, the accurate template region set and the pouring piece region set are screened out, and the accurate template region and the pouring piece region are determined in the left-eye image at the current stripping time based on the positional relationship, for subsequent stripping adhesion analysis.

[0098] Preferably, in one embodiment of the present application, the template region and the pouring piece region are determined in the monocular image at the current stripping time, comprising:

[0099] Referring to Figure 3 which shows a method flow chart of the method for determining the template region and the pouring member region in one embodiment of the present application, the method comprises the following steps:

[0100] Step S401: In all the matching region sets, analyze the number fluctuation difference of the pixel points between the to-be-processed regions and the depth value fluctuation difference, and divide all the matching region sets into a template region set and a deformation region set.

[0101] With the progress of the pouring process, the area and depth of the concrete pouring member region change, while the template region is relatively more stable, and the area size can be measured by the number of pixel points. Therefore, in each matching region set, the difference between the maximum value and the minimum value of the number of pixel points of all the to-be-processed regions is taken as a first number range, and the greater the first number range, the greater the range of area change. At the same time, the variance of the number of pixel points of all the to-be-processed regions is calculated, which can more intuitively reflect the fluctuation characteristics of the area, and the greater the variance, the greater the fluctuation. The value obtained by negatively correlating the product of the variance and the first number range is taken as a first stability factor. Based on the foregoing analysis, the greater the first stability factor, the more stable the change of the area, and the higher the possibility of being a template region set. The negative correlation mapping here can use the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0102] Then, the change of the texture characteristics of the to-be-processed regions is analyzed, which can be represented by the number change of the edge pixel points. Therefore, in each matching region set, the edge pixel points in each to-be-processed region are obtained based on the Canny operator. Similarly, the difference between the maximum value and the minimum value of the number of edge pixel points in all the to-be-processed regions is calculated as a second number range, and the greater the second number range, the greater the range of texture characteristic change, and the poorer the stability. The variance of the number of edge pixel points in all the to-be-processed regions is calculated, and the greater the variance, the greater the fluctuation degree of the texture characteristics, and the poorer the stability. Therefore, the value obtained by negatively correlating the product of the variance and the second number range is taken as a second stability factor. Based on the foregoing analysis, the greater the second stability factor, the more stable the change of the texture characteristics, and the higher the possibility of being a template region set. The negative correlation mapping here can use the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0103] The depth value of the template region will not have a large fluctuation change as the pouring process proceeds, so in each set of matching region collection, the variance of the depth value of the pixel point in all the to-be-processed regions is calculated, the greater the variance, the greater the fluctuation degree of the depth value, and the smaller the possibility of the template region, so the variance is negatively correlated and mapped, the logical relationship is corrected, and the third stability factor is obtained, and the greater the third stability factor, the higher the possibility of the template region collection. The negative correlation mapping can be realized by the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0104] Based on the foregoing logic, the first stability factor, the second stability factor, and the third stability factor are all positively correlated with the possibility of the matching region collection being the template region collection, so the product of the first stability factor, the second stability factor, and the third stability factor is normalized and taken as the template index, and the greater the template index, the greater the probability of the template region collection. The normalization is a technique well known to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0105] Finally, in all the matching region collections, the matching region collection corresponding to the maximum template index is taken as the template region collection, and the remaining matching region collections are taken as the deformation region collection.

[0106] Step S402: In the deformation region collection, the number variation trend and the depth value variation trend of the pixel points of the to-be-processed region are analyzed, and the pouring member region collection is screened out from all the deformation region collections in combination with the position change of the pixel points.

[0107] The deformation region collection screened out in step S401 includes the pouring member region collection and the concrete stain region collection, the concrete stain is caused by splashing during the concrete pouring process, the area and the depth change are unstable, and the deformation process usually occupies a relatively short time of the entire pouring process, and as the pouring process proceeds, the area and the depth of the pouring member both increase to a certain extent, and the continuity change is higher.

[0108] Therefore, in each set of deformation region collection, all the to-be-processed regions are arranged in time sequence according to the historical pouring time, to obtain a sorting sequence, and a first-order difference sequence can reflect the variation trend of the data value, so in the sorting sequence, the first-order difference sequence of the number of pixel points of the to-be-processed region is calculated as a number variation sequence, and the number of continuous positive values in the number variation sequence is taken as a number increase segment, thereby obtaining all the number increase segments, and the greater the length of the number increase segment, the more obvious the continuous growth of the number of pixel points of the to-be-processed region, and the higher the possibility of the pouring member region collection.

[0109] Similarly, in the sorting sequence, the first-order difference sequence of the mean value of the depth values of the pixels of the to-be-processed region is calculated as the depth change sequence, and consecutive positive values in the depth change sequence form a depth increase segment, and all depth increase segments are obtained. Similarly, the greater the length of the depth increase segment, the more obvious the continuous increase of the depth values of the pixels of the to-be-processed region, and the higher the possibility of the to-be-processed region being a cast member region.

[0110] Then, the product of the length mean value of all the number increase segments and the length mean value of the depth increase segments is normalized as the first cast member factor. The greater the first cast member factor corresponding to the deformation region set, the greater the probability that the deformation region set is a cast member region set. The normalization is a technique known to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0111] As the concrete fills the formwork, the center position of the cast member region gradually approaches the center position of the formwork region. Therefore, in the sorting sequence, the Euclidean distance between the centroid of each to-be-processed region and the centroid of the formwork region in the monocular image is taken as the distance coefficient. The smaller the distance coefficient, the closer the position. The first-order difference sequence of the distance coefficients of all to-be-processed regions is calculated as the position change sequence. The proportion of the number of negative values in the position change sequence is taken as the second cast member factor. The greater the second cast member factor, the closer the centroid of the to-be-processed region in the deformation region set to the centroid of the formwork region over time, and the greater the probability that the deformation region set is a cast member region set.

[0112] Based on the foregoing analysis, the first cast member factor and the second cast member factor of the deformation region set are positively correlated with the possibility of being a cast member region set. Therefore, the product of the first cast member factor and the second cast member factor is normalized as the cast probability of each deformation region set. The greater the cast probability, the greater the possibility of the deformation region set being a cast member region set. The normalization is a technique known to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0113] Finally, the deformation region set with a cast probability greater than a preset cast member threshold is taken as a cast member region set, and the remaining deformation region sets are taken as concrete stain sets.

[0114] It should be noted that in this embodiment of the present application, the preset cast member threshold is 0.6, and the specific value can be adjusted according to the implementation scenario, and is not limited herein.

[0115] Step S403: Determine the corresponding template region and the cast region in the left-view image at the current stripping time based on the template region set and the cast region set.

[0116] The concrete undergoes a phase transition process from liquid to plastic to solid from casting to stripping. The cast region at the last historical casting time can be regarded as the final form of the cast. Therefore, the region corresponding to the region to be processed in the left-view image at the last historical casting time in the template region set at the current stripping time is regarded as the template region, and the region corresponding to the region to be processed in the left-view image at the last historical casting time in the cast region set at the current stripping time is regarded as the cast region. The correspondence mentioned here refers to the correspondence of position coordinates, which can be located based on the position coordinates of the pixel points.

[0117] At this point, the template region and the cast region in the current stripping image can be obtained.

[0118] Step S5: Determine the stripping adhesion degree at the current stripping time based on the position difference features of the pixel points between the template region and the cast region in the monocular image at the current stripping time, and perform stripping adjustment based on the stripping adhesion degree at the current stripping time.

[0119] In the case of good stripping effect, the template is stably separated from the cast, so that the displacement of each point of the cast region relative to the template is highly similar. When there is local adhesion, the adhesion region of the cast may be deformed due to stress, forming local protrusions or depressions, that is, the displacement of the adhesion region and the template during the process of being pulled by the template will deviate from the displacement between the cast as a whole and the template. Therefore, in the left-view image at the current stripping time, the stripping adhesion degree at the current stripping time is determined based on the displacement difference features of the pixel points between the template region and the cast region in the embodiment of the present application.

[0120] Preferably, in an embodiment of the present application, the method for obtaining the stripping adhesion degree comprises:

[0121] At the current demolding moment, for any pixel point in each pouring piece region, the Euclidean distance between the pixel point and the pixel point in the nearest template region is calculated as a distance reference value, the average of the distance reference values of all pixel points in the pouring piece region is taken as a reference value, which can be regarded as the displacement feature of the whole pouring piece region relative to the template region, then the absolute value of the difference between the distance reference value of each pixel point in the pouring piece region and the reference value is calculated, the greater the absolute value of the difference, the greater the deviation of the displacement of the pixel point relative to the displacement of the whole pouring piece region relative to the template, and the greater the possibility of adhesion of the pixel point, so the absolute value of the difference is normalized as the adhesion probability of the pixel point, the greater the adhesion probability, the higher the possibility of the pixel point being an adhesion region pixel point; and in each pouring piece region, the pixel point with an adhesion probability greater than a preset adhesion threshold is taken as a demolding adhesion point. The normalization is a known technical means to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0122] In each pouring piece region, the K-means clustering algorithm and a preset K value are used for clustering analysis of all demolding adhesion points, and the region composed of the demolding adhesion points in the obtained clustering cluster is taken as an adhesion region, wherein the distance measurement is the Euclidean distance between the demolding adhesion points.

[0123] The greater the number of pixel points in the adhesion region and the greater the adhesion probability, the worse the demolding effect at the current demolding moment, so in each adhesion region, the product of the average of the adhesion probabilities of all pixel points in the adhesion region and the number of pixel points is normalized as the adhesion index of each adhesion region, the greater the adhesion index, the worse the demolding effect. The normalization is a known technical means to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0124] Finally, at the current demolding moment, the average of the adhesion indices of all adhesion regions is taken as the demolding adhesion degree at the current demolding moment, the greater the demolding adhesion degree, the worse the demolding effect, and the more need to increase the vibration frequency of demolding to realize demolding and ensure the demolding quality.

[0125] It should be noted that the K-means clustering algorithm is a known technology, and the specific process is not described herein; the preset K value is 3, and the specific value can be adjusted according to the implementation scene, which is not limited herein; the preset adhesion threshold is 0.75, and the specific value can be adjusted according to the implementation scene, which is not limited herein.

[0126] After obtaining the demolding adhesion degree at the current demolding moment, the demolding process can be adjusted based on the index.

[0127] Preferably, in one embodiment of the present application, the demolding adjustment is made based on the demolding adhesion degree at the current demolding time, including:

[0128] The sum of the demolding adhesion degree at the current demolding time and the preset parameter is taken as an adjustment degree value, and the product of the adjustment degree value and the preset demolding vibration frequency is taken as a demolding vibration adjustment frequency for demolding.

[0129] It should be noted that the preset parameter in the embodiment of the present application is set to 1; the preset demolding vibration frequency is 5HZ, and the specific value can be adjusted according to the implementation scene, which is not limited here.

[0130] In order to facilitate operation, all index data involved in operation in the embodiment of the present application are preprocessed, and the dimensional influence is cancelled. The means for removing the dimensional influence is a technical means familiar to those skilled in the art, which is not limited here.

[0131] In summary, by acquiring binocular images at each time in the construction process, the monitoring of the pouring and demolding process is realized. The feature points are determined based on the gray scale change and position distribution characteristics of the pixel points in the binocular images, and the registration is performed to generate a high-precision depth map, which breaks through the perception limitation of monocular vision on spatial information and helps to provide multi-dimensional information for subsequent demolding adhesion analysis. At the historical pouring time, the pixel position distribution of the monocular image is used to divide the to-be-processed region, and the cross-time matching is performed based on the region similarity, so as to realize the accurate tracking of the concrete solidification dynamics and provide a quantitative basis for demolding judgment. By analyzing the change characteristics of the pixel number, position and depth value in the matching region set, the template region and the pouring piece region are accurately distinguished in the monocular image at the current demolding time, the region misjudgment problem caused by concrete stain splashing is solved, and the demolding adhesion detection accuracy is improved. The demolding adhesion degree is calculated based on the position difference characteristics of the template region and the pouring piece region, and the demolding is adjusted in real time according to the demolding adhesion degree, forming a closed-loop control system of "monitoring-analysis-regulation", accurately determining the pouring piece region, improving the calculation accuracy of the demolding adhesion degree, and finally ensuring the demolding quality.

[0132] The embodiment of the present application also provides a water ditch cable slot hoisting integrated machine construction equipment, please refer to Figure 4A device structure schematic diagram of a water ditch cable trench hoisting integrated machine construction equipment provided by one embodiment of the present application is shown, and the device structure schematic diagram comprises a processor 500, a memory 501, a bus 502 and a communication interface 503, the processor 500, the communication interface 503 and the memory 501 are connected through the bus 502; wherein the memory 501 can contain a high-speed random access memory, the bus 502 can be an ISA bus, a PCI bus or an EISA bus, etc., the processor 500 can be an integrated circuit chip, and has a signal processing capability; the memory 501 stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to realize steps in a water ditch cable trench hoisting integrated machine construction method.

[0133] It should be noted that the above-mentioned embodiment sequence of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0134] Each of the embodiments in the specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. A construction method for a ditch cable trough hoisting machine, characterized in that: The method comprises: Acquire binocular images at various moments during the construction process, including the current demoulding moment and multiple historical pouring moments; In the binocular image at each moment, the depth map is obtained by registering the pixels based on their grayscale changes and position distribution characteristics. At each historical pouring moment, the regions are divided based on the position distribution of pixels in the monocular image to obtain the areas to be processed. Based on the similarity of the position distribution of pixels between the areas to be processed in the monocular images at different historical pouring moments, the areas to be processed are matched to obtain a set of matching areas. In all matching area sets, the changes in the number of pixels, the positions of the pixels, and the depth values ​​of the pixels in the area to be processed are analyzed to determine the formwork area and the casting area in the monocular image at the current demoulding moment; In the monocular image at the current demolding moment, the degree of demolding adhesion at the current demolding moment is determined based on the positional difference characteristics of the pixels between the template area and the casting area; and demolding adjustments are made based on the degree of demolding adhesion at the current demolding moment. The method for obtaining the demoulding adhesion degree includes: At the current demoulding moment, for any pixel point in each casting area, the Euclidean distance between the pixel point and the nearest pixel point in the template area is calculated as the distance reference value. The average of the distance reference values ​​of all pixels in the casting area is used as the benchmark value. The absolute value of the difference between the distance reference value of each pixel point in the casting area and the benchmark value is normalized as the sticking probability of the pixel point. In each casting area, the pixel points with a sticking probability greater than a preset sticking threshold are regarded as demoulding sticking points; In each casting area, all demoulding adhesion points are clustered based on the K-means clustering algorithm and the preset K value. The area consisting of the demoulding adhesion points in the obtained cluster is regarded as the adhesion area. The distance metric is the Euclidean distance between the demoulding adhesion points. In each adhesion area, the product of the mean adhesion probability of all pixels in the adhesion area and the number of pixels is normalized to obtain the adhesion index of each adhesion area; At the current demoulding moment, the average of the adhesion indices of all adhesion areas is taken as the demoulding adhesion degree at the current demoulding moment.

2. The method for constructing a water ditch cable trough hoisting machine according to claim 1, characterized in that: The method for obtaining the depth map includes: In the binocular image at each moment, the edge line and the grayscale gradient of each edge pixel in each monocular image are obtained based on the Canny operator; The edge pixels in the left image are used as reference points, and the edge pixels in the right image are used as comparison points. The Euclidean distance between each reference point and each comparison point is calculated as the distance factor. The absolute value of the grayscale gradient difference between each reference point and each comparison point is negatively correlated with the product of the distance factor and the normalized value is used as the first matching factor. Obtain the curvature value of each edge pixel on each edge line in each monocular image, and use the ratio of the mean of the curvature values ​​between each reference point and each comparison point to the absolute value of the curvature difference as the second matching factor; Normalizing the product of the first matching factor and the second matching factor between each reference point and each comparison point as the feature matching probability between each reference point and each comparison point; Among all the feature matching probabilities corresponding to each reference point, the comparison point and the reference point corresponding to the feature matching probability exceeding the preset matching threshold are taken as the matching point pair; In each set of binocular images, RANSAC is used to perform overall registration of the left and right images based on matching point pairs to obtain a depth map.

3. The method for constructing a water ditch cable trough hoisting machine according to claim 1, characterized in that: The method for obtaining the area to be processed includes: In the left-eye image at each historical pouring moment, edge detection and morphological operations are performed based on the Canny operator to obtain all closed areas as the areas to be processed.

4. The method for constructing a water ditch cable trough hoisting machine according to claim 1, characterized in that: The method for obtaining the matching area set includes: In the left-eye images at each of two adjacent historical pouring moments, the area to be processed in the left-eye image at the previous historical pouring moment is used as the reference area, and the area to be processed in the left-eye image at the next historical pouring moment is used as the comparison area; In each reference area and each comparison area, the value after negative correlation mapping of the Euclidean distance between the centroids is used as the distance proximity parameter, and the positions of the pixels in the reference area are compared with the positions of the pixels in the comparison area, and the number of pixels with the same position is used as the overlap factor; The normalized value of the product of the distance parameter between each reference area and each comparison area and the overlap factor is used as the matching index; Among all the matching indices corresponding to each reference area, the comparison area corresponding to the maximum matching index is used as the matching area of ​​the reference area; In the left-eye images at all historical pouring moments, the reference areas and the corresponding matching areas are grouped together to obtain all matching area sets.

5. The method for constructing a water ditch cable trough hoisting machine according to claim 1, characterized in that: Determining the template area and the casting area in the monocular image at the current demoulding moment includes: In all matching area sets, analyze the fluctuation differences in the number of pixels and the depth values ​​between the areas to be processed, and divide all matching area sets into template area sets and deformation area sets; In the deformation area set, the number of pixel points and the depth value change trend of the area to be processed are analyzed, and the casting area set is screened out from all deformation area sets based on the position change of the pixel points; The area to be processed in the left image at the last historical casting moment in the template area set and the area corresponding to it in the left image at the current demoulding moment are used as the template area, and the area to be processed in the left image at the last historical casting moment in the casting part area set and the area corresponding to it in the left image at the current demoulding moment are used as the casting part area.

6. The method for constructing a water ditch cable trough hoisting machine according to claim 5, characterized in that: The method of dividing all matching area sets into a template area set and a deformation area set includes: In each set of matching regions, the difference between the maximum and minimum values ​​of the number of pixels in all the regions to be processed is used as a first quantity range, and a product of the first quantity range and the variance of the number of pixels in all the regions to be processed is subjected to negative correlation mapping processing to obtain a first stability factor; In each set of matching regions, edge pixels in each to-be-processed region are obtained based on the Canny operator, the difference between the maximum and minimum values ​​of the number of edge pixels in all to-be-processed regions is used as the second quantity range, and the product of the second quantity range and the variance of the number of edge pixels in all to-be-processed regions is negatively correlated and mapped as the second stabilization factor. The value after negative correlation mapping of the variance of the depth values ​​of all pixels in the area to be processed is used as the third stabilization factor; Normalizing the product of the first stability factor, the second stability factor, and the third stability factor as a template index; Among all matching region sets, the matching region set corresponding to the maximum template index is used as the template region set, and the remaining matching region sets are used as the deformation region set.

7. The method for constructing a water ditch cable trough hoisting machine according to claim 5, characterized in that: The method for obtaining the casting area set includes: In each set of deformed regions, all the regions to be processed are arranged according to the chronological order of their historical pouring moments to obtain a sorted sequence. In the sorted sequence, a first-order difference sequence of the number of pixels in the regions to be processed is calculated as a quantity change sequence. Continuous positive values ​​in the quantity change sequence are grouped into quantity increasing segments, and all quantity increasing segments are obtained. In the sorted sequence, a first-order difference sequence of the mean depth values ​​of the pixels in the area to be processed is calculated as a depth change sequence, and continuous positive values ​​in the depth change sequence are grouped into depth increase segments to obtain all depth increase segments; The normalized value of the product of the mean length of all the segments with increased quantity and the mean length of the segments with increased depth is used as the first casting factor; In the sorting sequence, the Euclidean distance between the centroid of each area to be processed and the centroid of the template area in the corresponding monocular image is used as a distance coefficient, and a first-order difference sequence of the distance coefficients of all areas to be processed is calculated as a position change sequence, and the proportion of negative values ​​in the position change sequence is used as a second casting factor; The product of the first casting factor and the second casting factor is normalized to obtain a value as the casting probability of each set of deformation regions; The set of deformation regions whose casting probability is greater than a preset casting threshold is taken as the casting region set.

8. The method for constructing a water ditch cable trough hoisting machine according to claim 1, characterized in that: The demoulding adjustment based on the demoulding adhesion at the current demoulding moment includes: The sum of the demoulding adhesion degree at the current demoulding moment and the preset parameter is used as the adjustment degree value; The product of the adjustment degree value and the preset demoulding vibration frequency is used as the demoulding vibration adjustment frequency for demoulding.

9. A construction equipment for a ditch cable trough hoisting machine, characterized in that: It includes a processor and a memory, and the memory stores at least one instruction, at least one program, code set or instruction set. When the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, the steps of the construction method of a ditch cable trough hoisting integrated machine as described in any one of claims 1-8 are implemented.

Images