Precast concrete surface crack detection method and system based on image recognition

Abstract

The invention relates to the technical field of image recognition, and discloses a concrete prefabricated part surface crack detection method and system based on image recognition, and the method comprises the steps: collecting a double-spectrum image sequence of a concrete prefabricated part in a near-infrared wave band and a visible light wave band; separating a surface texture mask and a subsurface structure gradient map of the precast concrete in the double-spectrum image sequence; according to the non-continuous mutation characteristics of the subsurface structure gradient map, identifying a density abnormal region of the concrete prefabricated member, and extracting a potential crack path of the density abnormal region in combination with the penetration depth difference of the near-infrared band; utilizing the surface texture mask to shield an environment interference area of the potential crack path; and collecting the surface vibration frequency spectrum of the shielded area under simulated vibration excitation. The method can solve the problems that the potential crack path cannot be positioned through the penetration depth difference and the detection coverage range and the precision cannot meet the engineering requirements in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a concrete prefabricated part surface crack detection method and system based on image recognition. BACKGROUND

[0002] The prior art relies on single visible light band imaging, which can only capture surface texture features and cannot separate surface texture and subsurface structure information, making it difficult to identify density anomaly areas formed by uneven aggregate distribution and internal voids, and further resulting in a large number of missed detections of hidden cracks in the area. At the same time, the penetration characteristics of the near-infrared band are not utilized, and the potential crack path cannot be located through the penetration depth difference, and the detection coverage and accuracy cannot meet the engineering requirements.

[0003] In addition, the prior art lacks an effective environmental interference exclusion mechanism, and surface-attached stains, light shadows, etc. are easily misjudged as cracks, and the crack is only determined by the visual features of the static image, and the real crack and surface scratch cannot be distinguished by the material vibration response difference, and there is a risk of false judgment. SUMMARY

[0004] The present application provides a concrete prefabricated part surface crack detection method and system based on image recognition, which is mainly aimed at solving the problems in the above background.

[0005] To achieve the above purpose, the concrete prefabricated part surface crack detection method based on image recognition provided by the present application comprises:

[0006] S1, collecting a dual-spectrum image sequence of a concrete prefabricated part in a near-infrared band and a visible light band;

[0007] S2, separating a surface texture mask and a subsurface structure gradient map of the concrete prefabricated part in the dual-spectrum image sequence;

[0008] S3, identifying a density anomaly area of the concrete prefabricated part according to the non-continuous mutation characteristics of the subsurface structure gradient map, and extracting a potential crack path of the density anomaly area in combination with the penetration depth difference of the near-infrared band;

[0009] S4, shielding an environmental interference area of the potential crack path by using the surface texture mask;

[0010] S5, collecting a surface vibration spectrum of the shielded area under simulated vibration excitation, and counting the frequency domain attenuation difference between the surface vibration spectrum and a preset reference spectrum;

[0011] S6, when the frequency domain attenuation difference exceeds an acoustic impedance mismatch threshold, marking a verified crack area, and generating a crack quantization result according to the geometric topological characteristics of the verified crack area.

[0012] Preferably, the acquisition of the concrete precast double-spectrum image sequence in the near-infrared band and the visible light band includes:

[0013] Control the near-infrared band and the visible light band to illuminate in turn to obtain a double-band illumination environment;

[0014] Based on the double-band illumination environment, image processing is performed on the concrete precast to obtain a double-spectrum image of the concrete precast;

[0015] The double-spectrum image is integrated in time sequence to obtain a double-spectrum image sequence of the concrete precast.

[0016] Preferably, the separation of the surface texture mask and the subsurface structure gradient map of the concrete precast in the double-spectrum image sequence includes:

[0017] The near-infrared image in the double-spectrum image sequence is subjected to an axial penetration separation operation to obtain a depth feature mapping set of the near-infrared image;

[0018] The irradiance gradient of the depth feature mapping set is analyzed, and a subsurface structure gradient map of the depth feature mapping set is constructed based on the irradiance gradient;

[0019] The visible light image in the double-spectrum image sequence is separated, and a surface texture mask of the visible light image is extracted.

[0020] Preferably, the identification of the density abnormal area of the concrete precast according to the non-continuous mutation feature of the subsurface structure gradient map includes:

[0021] The radiation attenuation gradient mutation of the subsurface structure gradient map is detected, and the area coordinates of the gradient discontinuity in the subsurface structure gradient map are marked through the radiation attenuation gradient mutation;

[0022] The penetration depth difference of the near-infrared band in the area coordinates is verified, and the density abnormal area in the area coordinates is divided based on the penetration depth difference.

[0023] Preferably, the extraction of the potential crack path of the density abnormal area combined with the penetration depth difference of the near-infrared band includes:

[0024] The radiation intensity difference rate of the density abnormal area is collected, and the distribution of the radiation intensity difference rate is counted;

[0025] The maximum gradient direction angle of the distribution is determined, and the potential crack path of the distribution is determined based on the maximum gradient direction angle.

[0026] Preferably, the environmental disturbance region shielded by the surface texture mask includes:

[0027] A spatial projection mapping relationship of the potential crack path under the surface texture mask is established;

[0028] The environmental disturbance feature of the surface texture mask is enhanced to obtain an environmental disturbance distribution map of the surface texture mask;

[0029] The path segment of the potential crack path that coincides with the area of the environmental disturbance distribution map is removed.

[0030] Preferably, the surface vibration frequency spectrum of the shielded region is collected under simulated vibration excitation, including:

[0031] An axial vibration excitation is applied to the side surface of the concrete precast member by a hydraulic pulse exciter, and a high-frequency response propagation field of the axial vibration excitation is synchronously collected;

[0032] A vibration displacement time-domain signal of the high-frequency response propagation field is collected, and the vibration displacement time-domain signal is converted into a surface vibration frequency spectrum by Fourier transform.

[0033] Preferably, the calculation formula of the frequency domain attenuation difference is:

[0034]

[0035] Wherein: is the frequency domain attenuation difference, is the lowest feature perception frequency, is the highest effective response frequency, is the vibration frequency spectrum after shielding, is a preset reference spectrum, is an anti-aliasing constant, is a crack attenuation feature amplifier, a spectrum feature selector, is a dynamic range compression scale.

[0036] Preferably, the crack quantization result is generated according to the geometric topological feature of the verified crack region, including:

[0037] A main path ridge line skeleton tracking is performed on the verified crack region to obtain a topological main ridge line skeleton of the verified crack region;

[0038] The opening and depth of the crack profile in the verified crack region are measured based on the topological main ridge line skeleton;

[0039] The opening and depth are integrated into the crack quantization result of the verified crack region.

[0040] The concrete prefabricated part surface crack detection system based on image recognition comprises:

[0041] An image acquisition module is used to collect a dual-spectrum image sequence of the concrete prefabricated part in a near-infrared wave band and a visible light wave band.

[0042] A layering module is used to separate a surface texture mask and a subsurface structure gradient map of the concrete prefabricated part in the dual-spectrum image sequence.

[0043] A path generation module is used to identify a density abnormal area of the concrete prefabricated part according to a non-continuous mutation feature of the subsurface structure gradient map, and extract a potential crack path of the density abnormal area in combination with a penetration depth difference in the near-infrared wave band.

[0044] A shielding module is used to shield an environmental interference area of the potential crack path by using the surface texture mask.

[0045] An active disturbance module is used to collect a surface vibration frequency spectrum of the shielded area under simulated vibration excitation, and count a frequency domain attenuation difference between the surface vibration frequency spectrum and a preset reference frequency spectrum.

[0046] A result generation module is used to mark a verified crack area when the frequency domain attenuation difference exceeds an acoustic impedance mismatch threshold, and generate a crack quantization result according to a geometric topological feature of the verified crack area.

[0047] Compared with the prior art, the present application has the following beneficial effects:

[0048] By separating the surface texture mask and the subsurface structure gradient map of the concrete prefabricated part, and combining the penetration depth difference in the near-infrared wave band and the non-continuous mutation feature of the subsurface structure gradient map, the present application can accurately identify the density abnormal area and extract the potential crack path, effectively avoiding the missed detection problem of the subsurface implicit crack caused by information loss. At the same time, the surface texture mask is used to remove the local environmental interference section in the potential crack path, which can filter the interference factors such as stains and light shadows to the greatest extent, greatly reduces the false detection risk caused by environmental interference, and ensures the accuracy of crack identification.

[0049] The present application solves the pain points of positive determination in traditional static image detection by introducing hydraulic pulse vibration excitation to collect surface vibration spectrum, calculating the frequency domain attenuation difference with the preset reference spectrum, and verifying the crack authenticity with the acoustic impedance mismatch threshold, ensuring that the identified cracks are all real defects affecting the safety of the structure. In addition, based on the geometric topological features of the verified crack area, the crack opening and depth are accurately measured through main path ridge line skeleton tracking, which provides more comprehensive crack parameters compared to the traditional detection method that can only count the crack length, providing accurate data support for concrete precast structure safety evaluation and quality control. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A flowchart of a concrete precast surface crack detection method based on image recognition provided by an embodiment of the present application is shown in the figure.

[0051] Figure 2 A function module diagram of a concrete precast surface crack detection system based on image recognition provided by an embodiment of the present application is shown in the figure.

[0052] The purpose of the present application, functional characteristics and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0054] The embodiments of the present application provide a concrete precast surface crack detection method based on image recognition. The execution subject of the concrete precast surface crack detection method based on image recognition includes but is not limited to at least one of the electronic devices that can be configured to execute the method provided by the embodiments of the present application, such as a server, a terminal, etc. In other words, the concrete precast surface crack detection method based on image recognition can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server, or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms, etc. Basic cloud computing services.

[0055] Referring to Figure 1 A flowchart of a concrete precast surface crack detection method based on image recognition provided by an embodiment of the present application is shown in the figure. In this embodiment, the concrete precast surface crack detection method based on image recognition includes:

[0056] S1, collecting a dual-spectrum image sequence of the concrete prefabricated part in a near-infrared wave band and a visible light wave band.

[0057] In the embodiment, the collecting a dual-spectrum image sequence of the concrete prefabricated part in a near-infrared wave band and a visible light wave band comprises:

[0058] controlling near-infrared wave band and visible light wave band to illuminate in turn to obtain a dual-wave band illumination environment;

[0059] based on the dual-wave band illumination environment, performing imaging processing on the concrete prefabricated part to obtain a dual-spectrum image of the concrete prefabricated part;

[0060] integrating the dual-spectrum images in time sequence to obtain a dual-spectrum image sequence of the concrete prefabricated part.

[0061] Specifically, the near-infrared wave band refers to an electromagnetic wave band with a wavelength range of 760nm-2500nm. The light in this band has the characteristic of penetrating the surface layer of concrete and can irradiate to the subsurface layer of the concrete prefabricated part, such as a depth of several millimeters to tens of millimeters below the surface layer, and can reflect the information of the subsurface structure, such as internal aggregate distribution, void, and hidden cracks. It is a key wave band for subsequent acquisition of subsurface structure related images.

[0062] The visible light wave band refers to an electromagnetic wave band with a wavelength range of 380nm-760nm. The light in this band cannot penetrate the surface layer of concrete and can only be reflected by the surface layer of the concrete prefabricated part. It can clearly present the texture characteristics of the surface layer, such as surface stains, scratches, apparent cracks, and surface flatness. It is a key wave band for subsequent acquisition of surface texture related images and meets the needs of preliminary identification of surface defects of the concrete prefabricated part.

[0063] The dual-wave band illumination environment refers to an illumination scene that simultaneously has near-infrared wave band illumination and visible light wave band illumination capability, but the two wave bands of light are not turned on at the same time and work alternately in a specific order. This environment needs to cover the complete detection area of the concrete prefabricated part, such as the entire surface of the prefabricated slab, the side and top surface of the prefabricated beam, to ensure that the light uniformly irradiates the detection area without obvious shadows or uneven light intensity, thereby providing stable and effective light conditions for subsequent imaging processing.

[0064] In detail, through an illumination control device, such as a near-infrared light source controller and a visible light light source controller, the opening, closing and illumination intensity of the near-infrared wave band light source and the visible light wave band light source are precisely controlled according to a preset operation.

[0065] First check the initial state of the two light sources, and then set the illumination intensity parameters and detect the ambient light adjustment to avoid overexposure of the image caused by high intensity, and image blur caused by low intensity. The intensity of the near-infrared band light source is usually set to 500-1000 lux, and the intensity of the visible light band light source is usually set to 800-1500 lux.

[0066] Specifically, the concrete prefabricated part refers to a building component to be detected, which is pre-cast from concrete, such as a prefabricated concrete slab, a prefabricated concrete beam, a prefabricated concrete column, etc., and is the target object of imaging processing. The surface of the building component may have defects such as surface cracks, subsurface hidden cracks, surface stains, scratches, etc., and the imaging processing needs to capture the image information of the detection area of the building component completely.

[0067] The dual-spectrum image refers to a set of two types of images including a near-infrared band image of the concrete prefabricated part and a visible light band image of the concrete prefabricated part. The near-infrared band image reflects the subsurface structure information of the prefabricated part, and the visible light band image reflects the surface texture information of the prefabricated part. The two types of images need to correspond to the same detection area and the same detection time, i.e., the near-infrared image is taken when the near-infrared band light source is on, and the visible light image is taken when the visible light band light source of the corresponding batch is on.

[0068] In detail, first, an imaging system is built, the imaging device is fixed on a support, and the relative position of the imaging device and the detection area of the concrete prefabricated part is adjusted, such as aligning the center of the imaging device lens with the center of the detection area, controlling the distance to be 1-3 meters, adjusting according to the size of the detection area, ensuring that the detection area is completely presented in the image without edge cropping, and setting the imaging parameters, exposure time, and sensitivity.

[0069] When the near-infrared band light source is turned on in the dual-band illumination environment and reaches the preset illumination duration, the imaging device is started to take pictures of the detection area of the concrete prefabricated part, and one near-infrared band image is obtained;

[0070] Subsequently, the near-infrared band light source is turned off, the visible light band light source is turned on, and reaches the preset illumination duration, and then the imaging device is started again to take pictures of the same detection area, and one visible light band image is obtained. After the shooting is completed, the two images are preliminarily screened to remove blurred images, overexposed images, or underexposed images caused by light fluctuations or device shaking. The screened near-infrared band image and visible light band image are taken as a group to form a dual-spectrum image of the concrete prefabricated part. The relative positions of the imaging device, the concrete prefabricated part, and the light source need to remain unchanged during the entire imaging process to avoid mismatching of the detection areas of the two types of images due to position changes.

[0071] Specifically, the time sequence refers to the time sequence of the near-infrared band image and the visible light band image in the dual-spectrum image, and the time sequence of multiple groups of dual-spectrum images. Usually, the time sequence is recorded according to the time stamp, which is accurate to milliseconds, to ensure the logical sequence of image acquisition.

[0072] The dual-spectrum image sequence refers to a continuous image set formed by arranging multiple groups of dual-spectrum images according to the acquisition time sequence. The acquisition time interval of each group of dual-spectrum images in the sequence needs to be fixed, and it can continuously reflect the surface and subsurface image information of the concrete precast at different time points, providing continuous data support for subsequent image analysis, such as separating the surface texture mask and extracting the subsurface structure gradient.

[0073] In detail, first, add the acquisition time stamp to each group of dual-spectrum images. The time stamp needs to be accurate to the millisecond level of image shooting, to ensure that the time stamps of the near-infrared band image and the visible light band image of each group of images can correspond to the same lighting period.

[0074] Then set the integration time interval, and arrange multiple groups of dual-spectrum images in order from early to late according to the time stamp: first arrange the dual-spectrum image group corresponding to the earliest time stamp, then arrange the dual-spectrum image group corresponding to the second earliest time stamp, and so on, until the integration of the preset number of groups is completed.

[0075] Check whether the detection area of each group of dual-spectrum images is completely consistent. By comparing the edge features and surface fixed marker points of the concrete precast in the images, ensure that there is no area deviation caused by equipment displacement. Check whether the imaging parameters of each group of images are consistent with the initial settings. If an image group with a detection area deviation or parameter variation is found, it needs to be removed and re-acquired for supplementation; finally, save the multiple groups of dual-spectrum images arranged in sequence and passing the verification as a dual-spectrum image sequence.

[0076] S2, separate the surface texture mask and the subsurface structure gradient map of the concrete precast in the dual-spectrum image sequence.

[0077] In this embodiment, the separation of the surface texture mask and the subsurface structure gradient map of the concrete precast in the dual-spectrum image sequence includes:

[0078] Perform an axial penetration separation operation on the near-infrared image in the dual-spectrum image sequence to obtain a depth feature mapping set of the near-infrared image;

[0079] Analyze the irradiance gradient of the depth feature mapping set, and construct a subsurface structure gradient map of the depth feature mapping set based on the irradiance gradient;

[0080] Separate the visible light image in the dual-spectrum image sequence and extract the surface texture mask of the visible light image.

[0081] Specifically, the near-infrared image is an image in the dual-spectrum image sequence taken by near-infrared band illumination, and the near-infrared light can penetrate the concrete surface layer and reflect the information of the subsurface layer of the prefabricated part, such as subsurface hidden cracks and internal aggregate distribution.

[0082] The axial penetration separation operation is an operation of separating the information of the subsurface layer structure in the near-infrared image from the image in the direction of extending from the surface layer to the interior along the depth direction of the subsurface layer structure of the concrete prefabricated part.

[0083] The depth feature mapping set is a mapping result set of the depth features of the subsurface layer structure of the prefabricated part, such as subsurface hidden cracks and voids, which is obtained after the near-infrared image is subjected to the axial penetration separation.

[0084] In detail, all the near-infrared images are selected from the dual-spectrum image sequence, and it is necessary to ensure that the corresponding detection area of the same prefabricated part and the imaging parameters are stable. With the help of an industrial computer equipped with a near-infrared image special processing module, the pixel position of the surface layer of the prefabricated part in the image is first marked through edge detection to determine the axial direction.

[0085] Then, starting from the starting point of the surface layer, the pixel value, texture and other features of the near-infrared image are analyzed pixel by pixel or in a region-by-region manner at a step corresponding to an actual depth of about 1 mm, which can be set according to the penetration ability of the near-infrared light and the material of the prefabricated part. The features reflecting the subsurface layer structure are separated from the original image, and finally these separated features are integrated to form the depth feature mapping set.

[0086] Specifically, the irradiance gradient is the rate of change of the light intensity distribution of the near-infrared light in the subsurface layer region of the prefabricated part, which reflects the subsurface layer structure, such as the difference in the influence of hidden cracks on the near-infrared light. The change in the penetration and reflection of light at the cracks will cause a change in the irradiance.

[0087] The subsurface layer structure gradient image is an image in which the edges and changing regions of the subsurface layer structure, such as hidden cracks and subsurface layer voids, are visually presented in the form of gradient changes by analyzing the irradiance gradient of the depth feature mapping set.

[0088] In detail, the depth feature mapping set is imported into a workstation with image gradient analysis function, and the irradiance change amount of adjacent pixels or regions is calculated region by region for each depth feature mapping to determine the irradiance change rate in the transverse, longitudinal and other directions of the subsurface layer.

[0089] Subsequently, the image is reconstructed according to the gradient size and direction: the regions with large irradiance gradient are marked with specific gray scale or color, and the regions with small gradient are represented with uniform gray scale or color, and finally the subsurface layer structure gradient image capable of intuitively displaying the changes of the subsurface layer structure is generated.

[0090] Specifically, the visible light image is an image in the dual-spectrum image sequence taken by visible light band illumination, which can only reflect the information of the surface layer of the prefabricated part, such as surface layer cracks, stains, and normal texture, and the visible light cannot penetrate the surface layer.

[0091] The surface layer texture mask is used to mark the surface layer texture of the prefabricated part in the visible light image, and the mask contains the distribution area of the surface layer crack, the normal surface layer texture, and the surface layer stain. The surface layer texture information can be separated from the visible light image.

[0092] In detail, all visible light images are extracted from the dual-spectrum image sequence by image marking or time stamping. The visible light image is imported into a computer equipped with texture analysis software. The surface layer range of the prefabricated part in the image is first distinguished by color or gray scale, and then the texture features in the surface layer are analyzed pixel by pixel or in blocks. Finally, the area where the surface layer texture exists is marked as an effective area, and the background or non-surface layer area without surface layer texture is marked as an invalid area, to generate a surface layer texture mask.

[0093] S3, according to the non-continuous mutation characteristics of the sub-surface structure gradient map, identifying the density abnormal area of the concrete prefabricated part, and combining the penetration depth difference of the near-infrared band to extract the potential crack path of the density abnormal area.

[0094] In this embodiment, according to the non-continuous mutation characteristics of the sub-surface structure gradient map, the density abnormal area of the concrete prefabricated part is identified, which includes:

[0095] Detecting the radiation attenuation gradient mutation of the sub-surface structure gradient map, and marking the region coordinates of the gradient discontinuity in the sub-surface structure gradient map through the radiation attenuation gradient mutation;

[0096] Verifying the penetration depth difference of the near-infrared band in the region coordinates, and dividing the density abnormal area in the region coordinates based on the penetration depth difference.

[0097] In this embodiment, the potential crack path of the density abnormal area is extracted in combination with the penetration depth difference of the near-infrared band, which includes:

[0098] Collecting the radiation intensity difference rate of the density abnormal area, and counting the distribution of the radiation intensity difference rate;

[0099] Determining the maximum gradient direction angle of the distribution, and determining the potential crack path of the distribution based on the maximum gradient direction angle.

[0100] Specifically, the radiation attenuation gradient mutation is a phenomenon that the gradient value of near-infrared light radiation attenuation suddenly changes in the sub-surface structure gradient map due to the density change in the concrete, such as the existence of cracks, which is a key feature of density abnormalities.

[0101] The region coordinate of gradient discontinuity is the position coordinate of the location where the gradient of radiation attenuation is suddenly changed in the subsurface structure gradient map, and the gradient value is no longer smooth and continuous in the image coordinate system, which is used to accurately locate the suspected abnormal region.

[0102] In detail, in the indoor environment of concrete precast member detection, the operator imports the subsurface structure gradient map into the industrial image analysis system equipped with a gradient analysis module. The gradient map is scanned pixel by pixel, and the gradient change rate of each position is calculated. When the gradient change rate of a certain position exceeds the preset threshold, for example, 3 times the gradient change rate of the normal region based on C30 concrete calibration, it is determined that the radiation attenuation gradient is suddenly changed. Subsequently, the image coordinates of these mutation positions are automatically recorded, and the region coordinate of gradient discontinuity is calibrated to form a preliminary abnormal region coordinate set.

[0103] Specifically, the region coordinate is the position coordinate of the region with gradient discontinuity in the subsurface structure gradient map in the image coordinate system calibrated in the above step.

[0104] The difference in penetration depth in the near-infrared band is that the penetration depth of near-infrared light in the normal region and the density abnormal region of concrete is different. The penetration depth in the normal region is deep and stable, and the penetration depth in the abnormal region deviates from the normal range, which is the core basis for distinguishing density abnormalities.

[0105] The density abnormal region is a region in the subsurface of concrete where the density is different from that in the normal region due to internal structures such as cracks and looseness, which causes abnormal penetration depth and radiation attenuation of near-infrared light.

[0106] In detail, for each coordinate corresponding to a subsurface position, the normal region of the precast member is compared, wherein: the normal region of the precast member is a normal region selected without defects, and the near-infrared penetration depth is measured, such as the penetration depth of about 5 mm for C30 concrete normal penetration depth;

[0107] Analyze whether there is a difference in penetration depth. If the near-infrared penetration depth at a certain region coordinate is 2 mm or more shallower than the normal region, it is determined that the internal density of the region is abnormal. Finally, the software divides the region coordinates corresponding to the regions that meet the penetration depth abnormality based on the judgment result of the penetration depth difference.

[0108] S4, shielding the environmental interference area of the potential crack path by using the surface texture mask.

[0109] In this embodiment, the shielding the environmental interference area of the potential crack path by using the surface texture mask comprises:

[0110] Establish a spatial projection mapping relationship of the potential crack path under the surface texture mask;

[0111] The environmental interference feature of the surface texture mask is strengthened to obtain an environmental interference distribution map of the surface texture mask.

[0112] The path segment on the potential crack path that coincides with the area of the environmental interference distribution map is removed.

[0113] Specifically, the potential crack path is a path in which a crack may exist in the subsurface of the concrete precast piece, and is an object that needs to be further processed to exclude interference, and is in a precast piece subsurface detection scene, such as an indoor detection workshop, with stable light.

[0114] The surface texture mask is a mask image that marks the surface texture of the concrete precast piece and contains environmental interference factors such as surface normal texture and stains, and is used to distinguish surface interference from subsurface potential cracks, and is extracted based on a visible light image and is adapted to the light conditions of surface detection.

[0115] The spatial projection mapping relationship is a positional correlation between the potential crack path and the plane on which the surface texture mask is located after the potential crack path is projected onto the plane, and is used for subsequent judgment of the overlap between the potential crack path and the surface interference.

[0116] In detail, in an indoor detection workshop of a concrete precast piece, three-dimensional spatial data of the potential crack path and two-dimensional plane data of the surface texture mask are imported into a professional image processing software.

[0117] The potential crack path is projected in a direction perpendicular to the surface of the precast piece, so that each point on the path corresponds to a plane coordinate of the surface texture mask. Through coordinate matching and spatial position correlation, the corresponding positions of the projection points of the potential crack path and the regions of the surface texture mask are recorded one by one, and finally a spatial projection mapping relationship of the potential crack path under the surface texture mask is established, ensuring the spatial correspondence between the subsurface path and the surface texture region.

[0118] Specifically, the environmental interference feature is a part of the surface texture mask that belongs to environmental interference, such as surface stains, non-crack scratches, irregular textures, etc., which will interfere with the judgment of the potential crack path.

[0119] The environmental interference distribution map is an image that clearly shows the distribution of the surface environmental interference region after the environmental interference feature is strengthened, and is used to clearly define the interference range that needs to be blocked.

[0120] In detail, in the image processing workstation, a surface layer texture mask is loaded. Using image enhancement tools, environmental interference features in the mask are enhanced: adjusting brightness and contrast, increasing the gray difference between surface stains and normal textures; using edge detection algorithms to highlight the edges of the environmental interference area, making the interference area profile clearer. After the enhancement operation, the originally insignificant environmental interference features become prominent, and finally an environmental interference distribution map is generated that can clearly show the location and range of all environmental interference areas on the surface layer, ensuring accurate identification of the interference areas that need to be blocked.

[0121] Specifically, the overlapping path segment is the part of the spatial projection of the potential crack path that overlaps with the interference area in the spatial projection of the environmental interference distribution map. Due to surface layer interference, non-true subsurface cracks need to be removed.

[0122] In detail, the spatial projection mapping relationship of the potential crack path is superimposed and compared with the environmental interference distribution map. In the image processing software, the projection of the potential crack path is checked segment by segment to see if it overlaps with the interference area of the environmental interference distribution map.

[0123] If the projection area of a certain segment of the potential crack path overlaps with the interference area, mark that segment as a part that needs to be removed. According to the marking, remove the path segments that overlap with the environmental interference area from the original potential crack path, and only keep the path segments that do not overlap with the environmental interference distribution map area after projection, to obtain the potential crack path excluding surface layer environmental interference. The entire process is carried out in a stable detection environment to ensure the accuracy of the image data and operation.

[0124] S5, collecting the surface vibration spectrum of the blocked area under simulated vibration excitation, and counting the frequency domain attenuation difference between the surface vibration spectrum and the preset reference spectrum.

[0125] In this embodiment, the collecting of the surface vibration spectrum of the blocked area under simulated vibration excitation comprises:

[0126] An axial vibration excitation is applied to the side surface of the concrete precast piece by a hydraulic pulse exciter, and a high-frequency response propagation field of the axial vibration excitation is synchronously collected.

[0127] The vibration displacement time domain signal of the high-frequency response propagation field is collected, and the vibration displacement time domain signal is converted into a surface vibration spectrum by Fourier transform.

[0128] Specifically, the hydraulic pulse exciter is a device for applying controllable vibration excitation to the concrete precast piece, which is used in the concrete precast piece detection laboratory and can stably output pulse-type hydraulic power to simulate vibration working conditions.

[0129] The side surface of the concrete precast piece is the side surface of the concrete precast piece, which is the contact position of the hydraulic pulse exciter for applying vibration, and the exciter needs to be in close contact with the side surface.

[0130] Axial vibration excitation is a vibration excitation transmitted along the axis of the concrete precast, causing the precast to vibrate along the axis.

[0131] High-frequency response propagation field is a vibration response distribution area formed by the high-frequency vibration component when the axial vibration excitation propagates inside and on the surface of the concrete precast, including the propagation path and intensity of the vibration.

[0132] In detail, in the concrete precast detection laboratory, first, the concrete precast is stably fixed on the test bench. Then, the excitation head of the hydraulic pulse exciter is aligned with the side of the precast, and coupling agent is applied at the contact to ensure that the excitation head is tightly attached to the side. Start the hydraulic pulse exciter, set the excitation parameters, and let the exciter apply axial vibration excitation to the side of the precast.

[0133] At the same time, the acquisition system equipped with multiple vibration sensors is turned on to synchronously acquire the high-frequency response propagation field formed by the axial vibration excitation propagating inside and on the surface of the precast. During the acquisition process, the exciter is kept in stable contact with the side, and the sensors are firmly attached to the surface of the precast to obtain accurate vibration response data.

[0134] Specifically, the vibration displacement time-domain signal is the signal of the vibration displacement of a certain (or multiple) position on the surface of the precast changing with time in the high-frequency response propagation field, reflecting the time-domain dynamic process of the vibration.

[0135] Fourier transform is a method for converting time-domain signals into frequency-domain signals, which can display the frequency component distribution of the signal.

[0136] The surface vibration frequency spectrum is a frequency spectrum graph with frequency as the horizontal axis and vibration amplitude as the vertical axis after the Fourier transform of the vibration displacement time-domain signal, reflecting the frequency characteristics of the vibration on the surface of the precast.

[0137] In detail, the vibration displacement time-domain signal in the high-frequency response propagation field is continuously acquired by the vibration sensor, which records the vibration displacement value at each time to form a continuous time-domain data sequence. The acquired vibration displacement time-domain signal is transmitted to the signal processing workstation through a stable data transmission line.

[0138] In the workstation, open the professional signal processing software, import the vibration displacement time-domain signal, and select the Fourier transform function. The software processes the time-domain signal, decomposes the vibration displacement signal changing with time into components of different frequencies, and finally generates the surface vibration frequency spectrum, providing a data basis for subsequent statistical frequency-domain attenuation difference.

[0139] In this embodiment, the calculation formula of the frequency-domain attenuation difference is:

[0140]

[0141] wherein: is the frequency domain attenuation difference, is the lowest feature perception frequency, is the highest effective response frequency, is the post-occlusion vibration spectrum, is the preset reference spectrum, is the anti-aliasing constant, is the crack attenuation feature amplifier, spectrum feature selector, is the dynamic range compression scale.

[0142] In detail, is the frequency domain attenuation difference, which is an index for measuring the difference in the degree of frequency domain attenuation between the post-occlusion vibration spectrum and the preset reference spectrum, and is used to determine the spectral anomaly caused by cracks in the detection of concrete prefabricated parts.

[0143] is the lowest feature perception frequency, which is the lowest frequency that can effectively perceive the vibration features of the prefabricated part in the detection, and is determined by the concrete material and the low-frequency response capability of the equipment. The laboratory needs to ensure that the exciter covers this frequency.

[0144] is the highest effective response frequency, which is the highest frequency at which the prefabricated part can produce effective vibration response, and is limited by the size, internal structure of the prefabricated part and the high-frequency acquisition capability of the equipment. The laboratory needs to ensure that the sensor can collect signals in this range.

[0145] is the post-occlusion vibration spectrum, which is the surface vibration spectrum of the target area under simulated vibration excitation after the surface texture mask is used to shield the environmental interference area, reflecting the vibration frequency characteristics after excluding the surface interference.

[0146] is the preset reference spectrum, which is the surface vibration spectrum under the same simulated vibration excitation when the prefabricated part has no cracks (normal state), and is used as the reference for abnormality. It needs to be pre-collected in the laboratory for normal prefabricated parts.

[0147] is the anti-aliasing constant, which avoids and ratio calculation instability caused by too small factor value, ensuring calculation accuracy under different detection conditions in the laboratory.

[0148] is the crack attenuation feature amplifier, which amplifies the ratio reflecting the crack attenuation feature, highlights the spectral difference caused by cracks, and facilitates capture of crack signals.

[0149] is a spectrum feature selector, selecting spectrum features related to cracks from preset reference spectrum combining dynamic range compression scale , filtering frequency components sensitive to cracks, excluding irrelevant interference.

[0150] More specifically, in the environment of laboratory simulation of vibration excitation, first determine and , covering the effective frequency range of the vibration characteristics of the preform; then obtain and , introduce to avoid aliasing, calculate to get the basis value of the difference between the two;

[0151] Then use to amplify the crack-related attenuation features in the ratio, while using to filter the spectrum part sensitive to cracks based on and ;

[0152] Finally, multiply with the result of , integrate in the interval to , and divide by to normalize, to get , quantifying the attenuation difference between the spectrum after shielding and the reference spectrum, to determine whether cracks exist.

[0153] Specifically, the calculation formula of the crack attenuation feature amplifier is:

[0154]

[0155] Where: is the crack attenuation feature amplifier, is the asymmetric enhancement coefficient.

[0156] In detail, is the input value, representing , reflecting the relative relationship between the spectrum after shielding plus the anti-aliasing constant and the reference spectrum plus the anti-aliasing constant.

[0157] is the asymmetric enhancement coefficient, which adjusts the enhancement degree of the amplifier to different directions, with the ratio being greater than 1 or less than 1, and the attenuation feature, to adapt to the asymmetric spectrum difference of different crack types in the laboratory.

[0158] is the inverse hyperbolic tangent function, which transforms the input , maps the difference to an easy-to-handle range, and highlights the change information in the ratio.

[0159] sinh: hyperbolic sine function, f(x) = sinh(x) = (ex - e-x) / 2 Operation, further amplifies the crack-related features, especially when the logarithmic value changes.

[0160] is the natural logarithm of , taking the natural logarithm, converting multiplication and division into addition and subtraction, facilitating the capture of relative changes, reflecting the proportion of spectral differences.

[0161] More specifically, in the laboratory signal processing environment, first calculate the input value ; then the inverse hyperbolic tangent transformation is performed on and the absolute value is taken, mapping the difference of to an easy-to-handle range; at the same time, the natural logarithm absolute value of is calculated, and after taking the 1 / 2 power, it is multiplied by , and the result is subjected to hyperbolic sine operation , and then multiplied by to adjust the degree of asymmetric enhancement;

[0162] Finally, the absolute value result of the inverse hyperbolic tangent transformation is added to the hyperbolic sine operation result to obtain , amplifying the crack-related attenuation features in the input value, providing more prominent crack information for .

[0163] Specifically, the calculation formula of the spectrum feature selector is:

[0164]

[0165] Where: the spectrum feature selector, is the preset reference spectrum, is the dynamic range compression scale, is the sub-resonance wave gain factor, is the acoustic coupling resonance offset, is the frequency band continuity coefficient.

[0166] Specifically, is the spectrum feature selector, which filters the spectrum features sensitive to cracks from the preset reference spectrum , excluding irrelevant components, and is used in the laboratory spectrum analysis link.

[0167] is the dynamic range compression scale, adjusting the compression degree of the exponential function, controlling the selection range of different amplitude parts, and adapting to the dynamic range differences of different preforms in the laboratory.

[0168] is a subharmonic wave gain factor, adjusting the gain of the sinusoidal function part, enhancing the influence of subharmonic waves, highlighting the subharmonic characteristics brought by cracks.

[0169] is a sound coupling resonance offset, adjusting the center position of the sinusoidal function, making the selector align with the sound coupling resonance frequency position related to the crack, adapting to the resonance offset of the vibration excitation coupled with the preform in the laboratory.

[0170] is a frequency band continuity coefficient, controlling the frequency band width of the sinusoidal function, adjusting the selection range of the continuous region of , ensuring that the selected spectral features meet the continuity of the crack vibration spectrum.

[0171] is an exponential function, operating on , forming a bell-shaped weight, giving high weight to the spectral components of low amplitude and concentrated range.

[0172] The function has a band-pass selection feature, which can accurately screen spectral components in a specific frequency range, is the function of the third power, further strengthening the selection ability of the target frequency range, so that the spectral components that meet the conditions are more prominently selected.

[0173] More specifically, through the spectral integral mechanism, the vibration response of the concrete surface is converted into a continuous energy flow function, avoiding the feature fragmentation caused by traditional discrete sampling, taking the reference spectrum as the core reference system, the function can automatically identify the inherent vibration characteristics of the material, and the dual-channel nonlinear transformation is adopted in the core processing unit to form a cascading amplification path of crack sensitivity;

[0174] The acoustic impedance boundary detector is established by the hyperbolic inverse tangent structure, the weak signal of micro-cracks in the critical domain is explosively developed, the logarithmic hyperbolic transformation constructs the dissipation energy model of material damage, and the linear vibration attenuation is mapped to a non-linear energy collapse event, the function output value has a super-linear relationship with the crack depth, and the self-calibration measurement of sub-millimeter cracks is realized.

[0175] The Gaussian component forms a spectral energy dome, automatically suppressing the pseudo-peak signal caused by environmental vibration, and the cubic window function configuration forms a quantum well effect in the main frequency offset domain, accurately positioning the abnormal frequency point of sound coupling resonance, and when the double modes work together, the cooperative effect of characteristic frequency capture of crack area and background noise annihilation is produced.

[0176] Further, by establishing a three-dimensional mapping of energy flow-acoustic impedance-fracture toughness in the frequency domain, the reverse tracking of the vibration mode is realized for the potential crack path on the surface, and the nondestructive diagnosis of micro-cracks is upgraded from qualitative judgment to quantitative reconstruction, and the final output has a clear material fault physical image interpretation.

[0177] S6, when the frequency domain attenuation difference exceeds the acoustic impedance mismatch threshold, marking as a verified crack area, and generating a crack quantization result according to the geometric topological features of the verified crack area.

[0178] In this embodiment, the generating of the crack quantization result according to the geometric topological features of the verified crack area comprises:

[0179] The main path ridge line skeleton tracking is performed on the verified crack area to obtain a topological main ridge line skeleton of the verified crack area.

[0180] The opening and depth of the crack profile in the verified crack area are measured based on the topological main ridge line skeleton.

[0181] The opening and depth are integrated into the crack quantization result of the verified crack area.

[0182] Specifically, the verified crack area is a concrete precast surface and subsurface area with real cracks, which is marked by comparing the surface vibration spectrum with the preset reference spectrum to calculate the frequency domain attenuation difference, and when the difference exceeds the acoustic impedance mismatch threshold.

[0183] This area has excluded environmental interference such as surface stains, light shadows, and surface scratches, and has also been verified by vibration excitation, and only contains real cracks that affect the safety of the structure, and is the only target area for subsequent main path ridge line skeleton tracking.

[0184] The topological main ridge line skeleton refers to a topological structure line reflecting the crack core extension trajectory extracted from the verified crack area after the main path ridge line skeleton tracking operation. The skeleton is composed of a series of continuous pixel points, accurately corresponding to the center axis of the crack, retaining only the main extension direction of the crack, and can intuitively reflect the trend, total length and key inflection point of the crack.

[0185] In detail, in the image processing workstation in the indoor detection workshop, first, the dual-spectrum image containing the verified crack area is imported into the professional image recognition software; the image display parameters are adjusted to ensure that the crack edges of the verified crack area and the surrounding concrete surface form a clear gray difference, and the crack profile is free of blur and noise interference; the boundary of the verified crack area is manually or automatically framed and selected to lock the tracking range and avoid the tracking range exceeding the real crack area.

[0186] Starting from the crack initiation point at one end, the neighborhood pixels of each pixel point are identified, and it is determined whether there are continuous crack pixels in the neighborhood. If so, the neighborhood pixels are marked as main path candidate points. Continue to track along the extension direction of the main path candidate points, and exclude the pixels of the non-main path in the crack branch. During the tracking process, if a crack inflection point is encountered, the inflection point coordinates are recorded, and the tracking continues in the new direction until all the main crack paths in the verified crack area are covered, forming the preliminary main path ridge line.

[0187] Delete isolated pixels caused by image noise, and supplement missing pixels caused by small gray difference. After optimization, the software generates a preview image of the topological main ridge skeleton. The operator compares the original image of the verified crack area to confirm whether the skeleton line completely fits the crack center axis without deviation or breakage. If there is deviation, manually adjust the local pixel position until the topological main ridge skeleton accurately reflects the main extension trajectory of the crack, and finally save the skeleton image.

[0188] Specifically, the crack profile refers to the edge boundary of the crack in the verified crack area, which is composed of edge pixels on both sides of the crack, reflecting the actual boundary range of the crack on the surface of the concrete prefabricated part.

[0189] Its profile characteristics include irregularity of edge, width variation, and gray difference with surrounding concrete surface. Its clarity depends on the quality of the pre-acquired dual-spectrum image. The measurement should be carried out in an environment without image stretching or deformation to avoid displacement of the profile edge pixel position and affect the measurement accuracy of the opening and depth.

[0190] The opening refers to the width parameter of the crack on the surface of the concrete prefabricated part in the verified crack area, i.e., the vertical distance between the crack profile edges on both sides of the crack, usually measured in millimeters.

[0191] Since the crack width may have distribution differences, the overall width characteristics need to be reflected through multiple measurement points. The measurement environment should be based on a clear visible light image to avoid the crack profile edge being blurred due to concrete surface reflection or stain residue, affecting the positioning accuracy of the two edges.

[0192] The depth refers to the vertical distance of the crack extending from the surface of the concrete prefabricated part to the subsurface in the verified crack area, usually measured in millimeters, reflecting the penetration degree of the crack.

[0193] Its measurement relies on the penetration characteristics of the near-infrared band image. The environment needs to ensure that the penetration depth of the near-infrared image is accurately calibrated without penetration deviation caused by uneven distribution of concrete aggregates, ensuring that the depth can truly reflect the subsurface extension of the crack.

[0194] In detail, in the image processing workstation, the fusion image of the topological main ridge line skeleton and the verified crack area is loaded; the measurement points are planned according to the principle of uniform distribution and covering the full length, taking the topological main ridge line skeleton as the reference: if the crack length is less than or equal to 100 mm, one measurement point is selected every 5 mm; if the crack length is greater than 100 mm, one measurement point is selected every 10 mm, and additional measurement points are added at positions where the crack width changes obviously; the position of each measurement point is marked on the topological main ridge line skeleton through the marking tool of the software, so that each measurement point is located on the central pixel point of the topological main ridge line skeleton.

[0195] For each measurement point, a measurement straight line perpendicular to the extension direction of the topological main ridge line skeleton is automatically generated; the edge pixel points of the crack profiles on both sides are identified by scanning pixel by pixel along the measurement straight line from the topological main ridge line skeleton to both sides; the coordinates of the edge pixel points on the measurement straight line are recorded, and the distance between the two points is calculated, which is the opening of the measurement point.

[0196] After the measurement is completed, the opening value of each measurement point is marked at the corresponding position of the image, and is recorded in the data table.

[0197] Further, for each measurement point, the near-infrared band image of the verified crack area is switched to; a depth measurement line perpendicular to the surface of the concrete prefabricated part is generated, taking the pixel position of the measurement point on the topological main ridge line skeleton as the starting point; the sub-surface end pixel point of the crack profile is identified by scanning pixel by pixel along the depth measurement line from the surface to the sub-surface.

[0198] The coordinates of the starting point and the end point are recorded, and the vertical distance between the two points is calculated, which is the depth of the measurement point; after the measurement is completed, the depth value of each measurement point is recorded in association with the corresponding opening value, so as to avoid data confusion.

[0199] Specifically, the crack quantification result refers to the final achievement of quantitatively describing the geometric characteristics of the cracks in the verified crack area, which is the comprehensive presentation of the opening and depth data, and includes three parts: original measurement data, statistical analysis data, and visualized diagram.

[0200] The generation environment needs to ensure the accuracy of data integration, and the report format conforms to the engineering detection standard, so as to facilitate the subsequent structural safety evaluation and quality control of the concrete prefabricated part.

[0201] In detail, in the data arrangement module of the image processing workstation, the opening and depth original data of all measurement points are imported; the data is sorted according to the position sequence of the measurement points on the topological main ridge line skeleton, so as to ensure that the data sequence is consistent with the crack extension direction.

[0202] Through the data statistical function of the software, the maximum opening, the minimum opening and the average opening are screened out; the depth statistical value is calculated in the same way; at the same time, the opening change trend graph and the depth change trend graph are generated through the line graph tool, and the change law of the opening and the depth of the crack in the extension process is intuitively displayed.

[0203] In the image editing module, the original image of the verified crack area and the topological main ridge line skeleton are loaded; the number of each measurement point is marked on the skeleton, and the corresponding opening value and depth value are marked beside the number; the maximum opening section and the maximum depth section are marked with different color line segments, and a legend is added in the corner of the image; at the same time, the basic information of the verified crack area is added below the image, so that the information of the drawing is complete and easy to understand.

[0204] The statistical analysis data visualized drawing of the original measurement data table is integrated into a comprehensive report, which is named as the crack quantification result report of the verified crack area;

[0205] The report format is selected as the commonly used PDF or Excel format in engineering detection, so that the table data can be edited and the drawing is clear and distortion-free; in the integration process, the correspondence between each item of data and the drawing needs to be checked to avoid mismatch between the data and the drawing;

[0206] Finally, the report is saved to the specified folder of the detection equipment and backed up to the cloud, and the generation of the crack quantification result is completed.

[0207] As shown in Figure 2 FIG. 1 is a functional module diagram of a concrete precast surface crack detection system based on image recognition provided by an embodiment of the present application.

[0208] The concrete precast surface crack detection system based on image recognition 100 can be installed in an electronic device. According to the functions implemented, the concrete precast surface crack detection system based on image recognition 100 can include an image acquisition module 101, a layering module 102, a path generation module 103, a shielding module 104, an active disturbance module 105 and a result generation module 106. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, which are stored in the memory of the electronic device.

[0209] In the present embodiment, the functions of each module / unit are as follows:

[0210] The image acquisition module 101 is used to acquire a dual-spectrum image sequence of the concrete precast in the near-infrared band and the visible light band;

[0211] The layering module 102 is used to separate the surface texture mask and the sub-surface structure gradient of the concrete precast in the dual-spectrum image sequence.

[0212] The path generation module 103 is configured to identify a density abnormal area of the concrete prefabricated part according to a discontinuous mutation feature of the subsurface structure gradient map, and extract a potential crack path of the density abnormal area in combination with a penetration depth difference of the near-infrared band;

[0213] The shielding module 104 is configured to shield an environmental interference area of the potential crack path by using the surface layer texture mask;

[0214] The active disturbance module 105 is configured to collect a surface vibration frequency spectrum of the shielded area under simulated vibration excitation, and count a frequency domain attenuation difference between the surface vibration frequency spectrum and a preset reference frequency spectrum;

[0215] The result generation module 106 is configured to mark a verified crack area when the frequency domain attenuation difference exceeds an acoustic impedance mismatch threshold, and generate a crack quantization result according to a geometric topological feature of the verified crack area.

[0216] In several embodiments provided in the present application, it should be understood that the disclosed method and system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner.

[0217] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0218] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software function modules.

[0219] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0220] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Artificial intelligence is a theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by a digital computer, perceiving an environment, acquiring knowledge and using the knowledge to obtain optimal results.

[0221] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for detecting surface cracks of a concrete precast product based on image recognition, characterized by, The method comprises: S1, collecting a dual-spectrum image sequence of the concrete prefabricated part in the near-infrared band and the visible light band; S2, separating the surface texture mask and the subsurface structure gradient map of the concrete prefabricated part in the dual-spectrum image sequence; S3, identifying the density abnormal area of the concrete prefabricated part according to the non-continuous mutation characteristics of the subsurface structure gradient map, and extracting the potential crack path of the density abnormal area in combination with the penetration depth difference of the near-infrared band; S4, shielding the environmental interference area of the potential crack path by using the surface texture mask; S5, collecting the surface vibration spectrum of the shielded area under simulated vibration excitation, and counting the frequency domain attenuation difference between the surface vibration spectrum and the preset reference spectrum; S6, when the frequency domain attenuation difference exceeds the acoustic impedance mismatch threshold, marking it as a verified crack area, and generating a crack quantization result according to the geometric topological characteristics of the verified crack area.

2. The image recognition-based concrete precast surface crack detection method of claim 1, wherein, The collection of the dual-spectrum image sequence of the concrete prefabricated part in the near-infrared band and the visible light band comprises: controlling the near-infrared band and the visible light band to illuminate in turn to obtain a dual-band illumination environment; based on the dual-band illumination environment, imaging processing is performed on the concrete prefabricated part to obtain a dual-spectrum image of the concrete prefabricated part; the dual-spectrum images are integrated in time sequence to obtain a dual-spectrum image sequence of the concrete prefabricated part.

3. The image recognition-based concrete precast surface crack detection method of claim 1, wherein, The separation of the surface texture mask and the subsurface structure gradient map of the concrete prefabricated part in the dual-spectrum image sequence comprises: performing an axial penetration separation operation on the near-infrared image in the dual-spectrum image sequence to obtain a depth feature mapping set of the near-infrared image; analyzing the irradiance gradient of the depth feature mapping set, and constructing a subsurface structure gradient map of the depth feature mapping set based on the irradiance gradient; separating the visible light image in the dual-spectrum image sequence and extracting the surface texture mask of the visible light image.

4. The image recognition-based concrete precast surface crack detection method of claim 3, wherein, The identification of the density abnormal area of the concrete prefabricated part according to the non-continuous mutation characteristics of the subsurface structure gradient map comprises: detecting the radiation attenuation gradient mutation of the subsurface structure gradient map, and marking the region coordinates of the gradient discontinuity in the subsurface structure gradient map through the radiation attenuation gradient mutation; verifying the penetration depth difference of the near-infrared band in the region coordinates, and dividing the density abnormal area in the region coordinates based on the penetration depth difference.

5. The image recognition-based concrete precast surface crack detection method of claim 1, wherein, The extraction of the potential crack path of the density abnormal area in combination with the penetration depth difference of the near-infrared band comprises: collecting the radiation intensity difference rate of the density abnormal area, and counting the distribution of the radiation intensity difference rate; determining the maximum gradient direction angle of the distribution, and determining the potential crack path of the distribution based on the maximum gradient direction angle.

6. The image recognition-based concrete precast surface crack detection method of claim 5, wherein, The shielding of the environmental interference area of the potential crack path by using the surface texture mask comprises: establishing a spatial projection mapping relationship of the potential crack path under the surface texture mask; enhancing the environmental interference characteristics of the surface texture mask to obtain an environmental interference distribution map of the surface texture mask; The potential crack path is pruned of a path segment coinciding with the environmental interference distribution map region.

7. The image recognition-based concrete precast surface crack detection method of claim 1, wherein, The surface vibration frequency spectrum of the shielded region is collected under simulated vibration excitation, including: An axial vibration excitation is applied to the side of the concrete precast piece by a hydraulic pulse exciter, and a high-frequency response propagation field of the axial vibration excitation is synchronously collected; A vibration displacement time-domain signal of the high-frequency response propagation field is collected, and the vibration displacement time-domain signal is converted into a surface vibration frequency spectrum by Fourier transform.

8. The image recognition-based concrete precast surface crack detection method of claim 7, wherein, The formula for calculating the frequency domain attenuation difference is: wherein: is a frequency domain attenuation difference, is a lowest characteristic perceptual frequency, is a highest effective response frequency, is a post-occlusion vibration spectrum, is a pre-set reference spectrum, is an anti-aliasing constant, is a crack attenuation characteristic amplifier, spectrum characteristic selector, is a dynamic range compression scale.

9. The image recognition-based concrete precast surface crack detection method of claim 1, wherein, The crack quantization result is generated according to the geometric topological features of the verified crack region, including: A main path ridge line skeleton is tracked for the verified crack region, to obtain a topological main ridge line skeleton of the verified crack region; The opening and depth of a crack profile in the verified crack region are measured based on the topological main ridge line skeleton; The opening and depth are integrated into a crack quantization result of the verified crack region.

10. A system for detecting surface cracks in a concrete precast based on image recognition, characterized by, The system includes: An image acquisition module for collecting a dual-spectrum image sequence of a concrete precast piece in a near-infrared wave band and a visible light wave band; A layering module for separating a surface layer texture mask and a subsurface structure gradient map of the concrete precast piece in the dual-spectrum image sequence; A path generation module for identifying a density abnormal region of the concrete precast piece according to a non-continuous mutation feature of the subsurface structure gradient map, and extracting a potential crack path of the density abnormal region in combination with a penetration depth difference in the near-infrared wave band; A shielding module for shielding an environmental interference region of the potential crack path by using the surface layer texture mask; An active disturbance module for collecting a surface vibration frequency spectrum of the shielded region under simulated vibration excitation, and counting a frequency domain attenuation difference between the surface vibration frequency spectrum and a preset reference spectrum; A result generation module for marking a verified crack region when the frequency domain attenuation difference exceeds an acoustic impedance mismatch threshold, and generating a crack quantization result according to geometric topological features of the verified crack region.

Images