A terahertz-based conveyor belt core detection method and device

Abstract

The application provides a kind of based on terahertz conveyor belt core detection method and device, belong to conveyor belt detection technical field.The method comprises: by terahertz scanning equipment to the core of conveyor belt is scanned, obtains thermal map;Channel extraction and standardization processing are carried out to thermal map, and standardization two-dimensional coordinate matrix is obtained;Standardization two-dimensional coordinate matrix is carried out longitudinal projection and positioning processing, and the target position of steel core in the core is obtained;Carrying out horizontal projection and defect extraction processing, and the defect position of steel core is obtained;Based on target position and defect position, reconstruct image is generated, and based on the center coordinates of defect position and annotation parameter, it is annotated, and reconstruct image and annotation result are output.The method provided in the application can identify the position of steel core in the core and the defect position, and form corresponding reconstruct image and annotation result, realize the effective extraction and intuitive presentation of internal structure characteristics and defect position of the core.

Description

Technical Field

[0001] This application relates to the field of conveyor belt detection technology, and in particular to a method and apparatus for detecting conveyor belt cores based on terahertz. Background Technology

[0002] As conveyor belts are widely used in heavy-duty material handling and continuous production scenarios in industries such as mining, ports, and chemicals, their operational stability directly affects the efficiency and safety of the production system. The belt core, as the core load-bearing and force-transmitting component of the conveyor belt, typically uses a steel wire rope core structure to bear the tension transmission and load support during operation. Under long-term heavy-duty operation, high-frequency start-stop, and complex working conditions, the belt core is prone to defects such as fatigue fracture and corrosion fracture. If these defects are not detected in time, they can easily lead to malfunctions such as localized tearing or overall breakage of the conveyor belt, resulting in equipment damage, material loss, production stoppage, and even safety risks.

[0003] In related technologies, terahertz detection technology is commonly used for non-destructive testing of conveyor belt cores. This method features non-contact testing, strong penetration capability, no radiation damage, and high testing speed. It can penetrate the rubber coating of the conveyor belt to scan and image the internal structure of the core, completing the inspection without disassembling the conveyor belt, making it suitable for online inspection needs in continuous industrial production scenarios. However, existing terahertz core inspection technology still has shortcomings in practical applications: on the one hand, the clarity and resolution of the images obtained from scanning the internal structure of the core are limited, and the boundary distinction between the core area and the surrounding material in the image is not high, easily resulting in blurred outlines and unclear local features, affecting the accurate interpretation of the internal structure and defects of the core; on the other hand, the subsequent processing capability of the scanned data is still limited, making it difficult to effectively extract structural features and abnormal features of the core, and the accuracy and stability of identifying defects such as fractures and breaks still need to be improved, making it difficult to meet the requirements of detection accuracy and efficiency in industrial scenarios.

[0004] Therefore, in terahertz-based nondestructive testing of conveyor belt cores, the limited clarity of scanned images, the difficulty in identifying internal structural features, and the insufficient accuracy of defect identification have become urgent technical problems that need to be solved. Summary of the Invention

[0005] This application provides a terahertz-based method and apparatus for inspecting conveyor belt cores, aiming to solve the problems of limited scanning image clarity, difficulty in identifying internal structural features, and insufficient accuracy in defect identification in existing terahertz-based nondestructive testing of conveyor belt cores.

[0006] In a first aspect, this application provides a terahertz-based method for detecting the core of a conveyor belt, the method comprising: The core of the conveyor belt is scanned using a terahertz scanning device to obtain a thermal map of the core. Channel extraction and standardization are performed on the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core. The target position of the steel core in the strip is obtained by longitudinal projection and positioning processing of the standardized two-dimensional coordinate matrix; The defect locations of the steel core are obtained by performing lateral projection and defect extraction on the standardized two-dimensional coordinate matrix. Based on the target location and the defect location, a reconstructed image is generated; Based on the center coordinates and annotation parameters of the defect location, the steel core defect location in the reconstructed image is annotated to obtain the annotation result; Output the reconstructed image and the annotation results.

[0007] In one possible design, the heatmap is a heatmap in RGB format; The process of channel extraction and standardization of the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core includes: Extract the red channel values ​​from the heatmap; A two-dimensional coordinate matrix is ​​constructed based on the red channel values; The two-dimensional coordinate matrix is ​​standardized to obtain the standardized two-dimensional coordinate matrix.

[0008] In one possible design, the two-dimensional coordinate matrix is ​​a 1024×768 matrix, and the numerical value of the standardized two-dimensional coordinate matrix ranges from 0 to 255.

[0009] In one possible design, the step of performing longitudinal projection and positioning processing on the standardized two-dimensional coordinate matrix to obtain the target position of the steel core in the strip core includes: Calculate the column average of each column value in the standardized two-dimensional coordinate matrix to obtain a column average sequence; Maximum detection is performed on the average value sequence of the column to obtain the target maximum point; The target position of the steel core is determined based on the column position of the target maximum point.

[0010] In one possible design, the extreme value detection of the column average sequence to obtain the target maximum point includes: Extreme value detection is performed on the average value sequence of the column to obtain at least one first candidate maximum point; For each of the at least one first candidate maximum point, the adjacent column corresponding to the first candidate maximum point is determined according to the column position corresponding to the first candidate maximum point; The slope is calculated based on the column average of the column containing the first candidate maximum point and the column average of the adjacent columns. If the absolute value of the slope is greater than a preset slope threshold, the first candidate maximum point is determined as the target maximum point.

[0011] In one possible design, the process of performing lateral projection and defect extraction on the standardized two-dimensional coordinate matrix to obtain the defect location of the steel core includes: Calculate the row average of the values ​​in each row of the standardized two-dimensional coordinate matrix to obtain a sequence of row averages; Extreme value detection is performed on the row average sequence to obtain at least one candidate minimum point; Based on the at least one candidate minimum point, at least one initial defect interval is determined; Based on a preset length ratio threshold and interval inclusion relationship, the at least one initial defect interval is filtered to obtain the target defect interval; Based on the target defect range, the defect location of the steel core is determined.

[0012] In one possible design, the extreme value detection of the row average sequence to obtain at least one candidate minimum point includes: The row average sequence is negativeed to obtain a negative row average sequence. Maximum detection is performed on the negative row average sequence to obtain at least one second candidate maximum point; The corresponding point of the at least one second candidate maximum point in the row average sequence is determined as the at least one candidate minimum point.

[0013] In one possible design, generating the reconstructed image based on the target location and the defect location includes: Create a blank image with the same dimensions as the standardized two-dimensional coordinate matrix; In the blank image, the first interval corresponding to the defect location is assigned a black value, the area in the second interval corresponding to the target location other than the first interval is assigned a red value, and the area in the blank image other than the first interval and the second interval is assigned a white value, thus obtaining the reconstructed image.

[0014] In one possible design, the annotation of the steel core defect location in the reconstructed image based on the center coordinates and annotation parameters of the defect location, to obtain the annotation result, includes: The center coordinates of the defect location are determined based on the row and column ranges corresponding to the defect location in the reconstructed image. The annotation parameters are determined based on the row length and column length corresponding to the defect location; Based on the center coordinates and the annotation parameters, a rectangular box is drawn in the reconstructed image to obtain the annotation result; wherein, the annotation parameters include the width and height of the rectangular box.

[0015] Secondly, this application provides a terahertz-based conveyor belt core detection device, comprising: a module for performing the aforementioned method embodiment of the first aspect.

[0016] Thirdly, this application provides a terahertz-based conveyor belt core detection device, comprising: a memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method described in the first aspect or various possible designs of the first aspect.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method described in the first aspect or various possible designs of the first aspect.

[0018] Fifthly, this application provides a computer program product, which includes computer program code that, when run on a computer, causes the computer to implement the method described in the first aspect or various possible designs of the first aspect.

[0019] This application provides a terahertz-based method for detecting the core of a conveyor belt. The method involves scanning the conveyor belt core using a terahertz scanning device to obtain a corresponding heat map. The heat map is then processed by channel extraction and standardization to obtain a standardized two-dimensional coordinate matrix. This transforms the heat map, which is difficult to interpret directly, into matrix data that is easier to analyze uniformly. Furthermore, the standardized two-dimensional coordinate matrix is ​​subjected to vertical and horizontal projections. The vertical projection is used to extract the distribution characteristics of the steel core along the column direction from the two-dimensional matrix to determine the target position of the steel core. The horizontal projection is used to extract the change characteristics corresponding to abnormal areas from the two-dimensional matrix to determine the steel core's position. The method identifies the defect location, thus separating the identification of the internal structure of the strip core into specific location extraction and anomaly extraction processes. Subsequently, a reconstructed image is generated based on the target and defect locations to reorganize and visualize the normal, defective, and other regions of the steel core in the original scan results, providing a more intuitive presentation of the internal structural relationships and defect distribution relationships of the strip core. Finally, the defect locations of the steel core in the reconstructed image are labeled based on the center coordinates and annotation parameters of the defect location, and the reconstructed image and annotation results are output. This ensures that the detection results not only reflect the distribution of the steel core in the strip core but also directly indicate the defect location and its corresponding position. The detection method provided in this application can transform scan results with limited clarity into detection results with clear structural expression and defect indication capabilities, achieving effective extraction and intuitive presentation of the internal structural features and defect locations of the strip core. This solves the technical problems of limited scan image clarity, difficulty in identifying internal structural features, and insufficient defect identification accuracy in related technologies. Attached Figure Description

[0020] Figure 1 A schematic flowchart of a terahertz-based conveyor belt core detection method provided in this application embodiment; Figure 2a An original thermal image of a conveyor belt core obtained by a terahertz scanning device provided in this application embodiment; Figure 2b A reconstructed image provided in an embodiment of this application; Figure 3 This is a schematic diagram of extreme value extraction of a column average sequence provided in an embodiment of this application; Figure 4a A comparison diagram before and after optimization of defect detection in the beginning and end intervals provided for embodiments of this application; Figure 4b An optimized detection result diagram of the first and last interval defect detection provided in this application embodiment; Figure 5a This is a schematic diagram illustrating how to filter initial defect intervals based on interval inclusion relationships, as provided in an embodiment of this application. Figure 5b This is a diagram showing the detection results after filtering the initial defect intervals based on the interval inclusion relationship; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] 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 this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and drawings of this application are intended to cover non-exclusive inclusion.

[0023] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of the phrase "embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists, A and B can exist simultaneously, and B exists. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0025] Furthermore, the terms "first," "second," etc., in the specification and claims of this application or in the aforementioned drawings are used to distinguish different objects rather than to describe a specific order, and may explicitly or implicitly include one or more of the features.

[0026] In the description of this application, unless otherwise stated, "multiple" and "at least two" mean two or more (including two), and similarly, "multiple groups" and "at least two groups" mean two or more (including two groups).

[0027] In the description of this application, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, "connected" or "linked" can refer not only to a physical connection, but also to an electrical connection or a signal connection. For instance, it can be a direct connection, i.e., a physical connection, or an indirect connection through at least one intermediate component, as long as the circuit is connected. It can also refer to the internal connection between two components. A signal connection can refer not only to a signal connection through a circuit, but also to a signal connection through a medium, such as radio waves. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, different technical features in this application can be combined with each other.

[0029] Figure 1 This is a schematic flowchart illustrating a terahertz-based conveyor belt core detection method provided in an embodiment of this application. Figure 1 As shown, the conveyor belt core detection method provided in this application embodiment specifically includes S101 to S106, and S101 to S106 will be described in detail below.

[0030] It should be noted that the execution subject of the terahertz-based conveyor belt core detection method provided in this application embodiment can be a terahertz-based conveyor belt core detection device.

[0031] S101. The core of the conveyor belt is scanned using a terahertz scanning device to obtain a thermal map of the core.

[0032] In this embodiment, the object of detection is the steel wire rope core in the conveyor belt core. By controlling the terahertz scanning device to scan the core along the preset detection area of ​​the conveyor belt, the terahertz response data corresponding to the internal structure of the core is collected, and the terahertz response data is converted into a heat map.

[0033] Heat maps are used to characterize the distribution of response intensity at different locations inside the conveyor belt core, thus providing a raw data foundation for subsequent identification of the steel core location and extraction of defect locations.

[0034] In some implementations, the terahertz scanning device can use the 0.4THz frequency band for scanning, and the scanning resolution can be 8mm; the heat map generated after scanning can be an RGB format heat map, which is convenient for subsequent channel extraction processing.

[0035] S102. Perform channel extraction and standardization on the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core.

[0036] In this embodiment, by extracting data related to the internal structural response of the core from the heat map and performing uniform scaling on the extracted data, a standardized two-dimensional coordinate matrix suitable for subsequent location identification and defect extraction can be obtained.

[0037] S103. Perform longitudinal projection and positioning processing on the standardized two-dimensional coordinate matrix to obtain the target position of the steel core in the strip core.

[0038] The target location is used to characterize the distribution area of ​​the steel core in the strip core.

[0039] In this embodiment, by performing longitudinal projection processing on the standardized two-dimensional coordinate matrix, the two-dimensional matrix data characterizing the distribution of the steel core in the strip core can be converted into a first set of one-dimensional data that varies along the column direction, so that the position features corresponding to the steel core can be extracted from the first set of one-dimensional data to obtain the target position of the steel core in the strip core.

[0040] S104. Perform lateral projection and defect extraction on the standardized two-dimensional coordinate matrix to obtain the defect location of the steel core.

[0041] Among them, the defect location is used to characterize the area in the steel core where there is a fracture or damage.

[0042] In this embodiment, by performing lateral projection and defect extraction processing on the standardized two-dimensional coordinate matrix, the abnormal change information related to steel core defects can be converted into a second set of one-dimensional feature data distributed along the row direction, so as to extract the abnormal interval of the steel core from the second set of one-dimensional feature data and obtain the defect location of the steel core.

[0043] S105. Generate a reconstructed image based on the target location and the defect location.

[0044] The reconstructed image is used to visually present the position of the steel core in the strip, as well as the correspondence between the defective and non-defective positions of the steel core.

[0045] In this embodiment, by distinguishing different regions in the image based on the target location and the defect location, the normal region, defect region and background region of the steel core can be expressed in different display forms, thereby generating a reconstructed image that characterizes the distribution state of the steel core and the distribution state of the defects.

[0046] S106. Based on the center coordinates and annotation parameters of the defect location, the steel core defect location in the reconstructed image is annotated to obtain the annotation result.

[0047] The annotation results are used to further indicate the specific distribution of defect locations in the reconstructed image.

[0048] In this embodiment, by annotating the steel core defect location in the reconstructed image based on the center coordinates and annotation parameters of the defect location, a visual marker corresponding to the defect area can be formed in the reconstructed image, thereby obtaining the annotation result.

[0049] S107. Output the reconstructed image and annotation results.

[0050] In this embodiment, after determining the target location of the steel core, extracting the defect location, reconstructing the image, and annotating the defect, the reconstructed image and annotation results are output as the final detection result. The reconstructed image is used to display the distribution of the steel core in the conveyor belt and the distinction between defective and non-defective areas. The annotation results are used to display the visual marking information corresponding to the defect location, thereby facilitating the inspection personnel to observe, identify, and locate steel core defects in the conveyor belt core.

[0051] In some implementations, the output may include an image file containing only the reconstructed image, or it may include a detection result image overlaid with the labeled results.

[0052] In other implementations, the reconstructed image and annotation results can be displayed on a display terminal or stored as detection records for later viewing and analysis.

[0053] This application provides a terahertz-based method for detecting the core of a conveyor belt. The method involves scanning the conveyor belt core using a terahertz scanning device to obtain a corresponding heat map. The heat map is then processed by channel extraction and standardization to obtain a standardized two-dimensional coordinate matrix. This transforms the heat map, which is difficult to interpret directly, into matrix data that is easier to analyze uniformly. Furthermore, the standardized two-dimensional coordinate matrix is ​​subjected to vertical and horizontal projections. The vertical projection is used to extract the distribution characteristics of the steel core along the column direction from the two-dimensional matrix to determine the target position of the steel core. The horizontal projection is used to extract the variation characteristics corresponding to abnormal areas from the two-dimensional matrix to determine the defects in the steel core. The method identifies the core's internal structure and defects by defining their respective locations in the extraction and anomaly extraction processes. Subsequently, a reconstructed image is generated based on the target and defect locations to reorganize and visualize the normal, defective, and other regions of the core from the original scan, providing a more intuitive representation of the core's internal structural relationships and defect distribution. Finally, the defect locations in the reconstructed image are labeled using the center coordinates and annotation parameters, and the reconstructed image and annotation results are output. This ensures that the detection results not only reflect the distribution of the core but also directly indicate the defect location and its corresponding position. The detection method provided in this application transforms scan results with limited clarity into detection results with clear structural representation and defect indication capabilities, effectively extracting and intuitively presenting the core's internal structural features and defect locations. This addresses the technical problems of limited scan image clarity, difficulty in identifying internal structural features, and insufficient defect recognition accuracy in related technologies.

[0054] Figure 2a This application provides an embodiment of a terahertz scanning device for obtaining the original thermal image of a conveyor belt core. Figure 2b A reconstructed image is provided as an embodiment of this application. For example... Figure 2a and Figure 2b As shown, the core structure in the heatmap is somewhat blurred, the boundary between the steel core area and the background area is poorly distinguishable, and the internal structural features of the core are not clear enough. After reconstruction, the strip structure corresponding to the steel core in the core is more clearly presented in the reconstructed image, and a clear distinction is achieved between the steel core area and the background area. Figure 2a and Figure 2b This visually demonstrates the reconstruction effect of the method in this application on the conveyor belt core image.

[0055] In one possible embodiment, the method steps shown in S102 can be implemented by S1021 to S1023, which are described in detail below.

[0056] S1021, Extract the red channel values ​​from the heatmap.

[0057] In this embodiment, the heatmap contains data from multiple color channels. By converting the multi-channel image data in the heatmap into single-channel numerical data, the complexity of subsequent data processing can be reduced, and a unified data source can be provided for the construction of the subsequent two-dimensional coordinate matrix.

[0058] In some implementations, an image processing library can be used to separate the channels of an RGB format heatmap to obtain single-channel image data corresponding to the red channel.

[0059] For example, the red channel values ​​can be extracted using the cv2.split function in the OpenCV library to obtain a single-channel image.

[0060] S1022. Construct a two-dimensional coordinate matrix based on the red channel values.

[0061] In this embodiment, the red channel values ​​corresponding to each pixel position can be mapped to a two-dimensional coordinate plane according to the spatial distribution relationship of the red channel values ​​in the heat map, thereby constructing a two-dimensional coordinate matrix. This converts the heat map data, which was originally represented in image form, into matrix data with row and column indexing, so that subsequent vertical projection processing can be performed based on the column direction data in the matrix, or horizontal projection processing can be performed based on the row direction data in the matrix.

[0062] In some implementations, the two-dimensional coordinate matrix can be set to correspond to the resolution of the original heatmap; for example, the red channel values ​​can be converted into a 1024×768 two-dimensional coordinate matrix to characterize the distribution of response values ​​at each location within the core scanning area.

[0063] S1023. Standardize the two-dimensional coordinate matrix to obtain a standardized two-dimensional coordinate matrix.

[0064] In this embodiment, by standardizing the values ​​in the two-dimensional coordinate matrix, data with different value ranges in the original matrix can be mapped to a preset value interval, thereby obtaining a standardized two-dimensional coordinate matrix. The values ​​in the standardized two-dimensional coordinate matrix can be compared and calculated under a unified dimension, facilitating subsequent location identification and defect extraction based on column and row averages.

[0065] In some implementations, the min-max normalization method can be used to process the two-dimensional coordinate matrix, mapping the matrix values ​​of the two-dimensional coordinate matrix to the interval [0, 255] to eliminate scanning intensity deviation and obtain a normalized two-dimensional coordinate matrix.

[0066] In one possible embodiment, the method steps shown in S103 can be implemented by S1031 to S1033, which are described in detail below.

[0067] S1031. Calculate the column average of each column value in the standardized two-dimensional coordinate matrix to obtain the column average sequence.

[0068] In this embodiment, by calculating the average value of each column in the standardized two-dimensional coordinate matrix, the matrix data originally represented in two-dimensional form can be converted into one-dimensional sequence data that varies along the column direction, thereby reducing the data processing complexity of subsequent position identification and providing a serialized data foundation for the subsequent extraction of the steel core target position.

[0069] Since the steel core in the strip typically corresponds to the response change in the column direction in the standardized two-dimensional coordinate matrix, the positional features related to the steel core distribution can be preserved in the column average sequence by calculating the column average.

[0070] In some implementations, the average value of each column in the standardized two-dimensional coordinate matrix can be calculated separately, and the obtained average values ​​can be arranged in column index order to form a column average value sequence.

[0071] For example, when the standardized two-dimensional coordinate matrix is ​​a 1024×768 matrix, a sequence of column averages of length 1024 can be obtained by calculating the average value of each of the 1024 columns.

[0072] S1032. Perform maximum detection on the column average sequence to obtain the target maximum point.

[0073] In this embodiment, since the area where the steel core is located usually appears as a relatively prominent high value position in the column average sequence, by performing maximum value detection on the column average sequence, the peak feature corresponding to the position of the steel core can be extracted from the column average sequence, thereby obtaining the target maximum point used to characterize the position of the steel core.

[0074] In some implementations, an extreme value detection algorithm can be used to detect local peaks in the column average sequence, thereby determining the target maximum point.

[0075] For example, you can call the find_peaks function in the SciPy library to detect maximum values ​​in a column mean sequence, and set the minimum peak distance to 5 and the minimum peak height to 100 to obtain the corresponding maximum points.

[0076] S1033. Determine the target position of the steel core based on the column position of the target maximum point.

[0077] In this embodiment, by obtaining the column position corresponding to the target maximum point, the peak position in the column average sequence can be mapped inversely to the column direction position in the standardized two-dimensional coordinate matrix, thereby determining the target position of the steel core in the strip. In other words, the column position corresponding to the target maximum point can characterize the distribution area of ​​the steel core in the standardized two-dimensional coordinate matrix, so the target maximum point can be used as the basis for determining the target position of the steel core.

[0078] This application embodiment calculates the column average value of each column in a standardized two-dimensional coordinate matrix to obtain a column average value sequence. Then, it performs maximum value detection based on the column average value sequence and determines the target position of the steel core according to the column position of the target maximum value point. This realizes the transformation of two-dimensional matrix data into one-dimensional sequence data that is easy to analyze, and extracts the positional features of the steel core in the strip core from it, providing a positional basis for subsequent extraction of steel core defect positions and generation of reconstructed images.

[0079] Figure 3 This is a schematic diagram illustrating the extraction of extreme values ​​from a column average sequence, provided as an embodiment of this application. For example... Figure 3 As shown, Figure 3 The curve in the figure represents the column average sequence obtained by vertically projecting the standardized two-dimensional coordinate matrix, and the part circled by the dashed line represents the maxima detected from the column average sequence. By identifying the maxima in the column average sequence, the target maxima corresponding to the steel core position can be extracted, and the target position of the steel core can be determined based on the column position of the target maxima. Figure 3 It can be seen that this application can provide a basis for determining the target position of the steel core by projecting the standardized two-dimensional coordinate matrix vertically and extracting the maximum points in the column average value sequence.

[0080] In one possible embodiment, the method steps shown in S1032 can be implemented by Sa1 to Sa4, which are described in detail below.

[0081] Sa1. Perform extreme value detection on the column average sequence to obtain at least one first candidate maximum point.

[0082] It should be noted that since the column average sequence reflects the response change of the normalized two-dimensional coordinate matrix along the column direction, the local high points in the column average sequence usually correspond to the candidate positions of the steel core in the strip. By performing extreme value detection on the column average sequence, the peak positions corresponding to the steel core positions can be initially identified from the column average sequence, thereby obtaining at least one first candidate maximum point.

[0083] In some implementations, an extreme value detection algorithm can be used to detect local peaks in the column average sequence, thereby obtaining at least one first candidate maximum point.

[0084] For example, the extreme value detection function can be called to perform peak search on the column average sequence to obtain multiple candidate peak positions.

[0085] Sa2. For each first candidate maximum point among at least one first candidate maximum point, determine the adjacent column corresponding to the first candidate maximum point according to the column position corresponding to the first candidate maximum point.

[0086] It should be noted that since the first candidate maximum point itself only represents a local high value on a certain column, and the accurate determination of the steel core position also requires the combination of the numerical change relationship between the column and the adjacent columns, by determining the adjacent columns corresponding to each first candidate maximum point, we can obtain the column direction change information near the first candidate maximum point, thus providing a basis for subsequent slope calculation.

[0087] In some implementations, the left and right adjacent columns of the first candidate maximum point can be determined as the adjacent columns corresponding to the first candidate maximum point based on the column index of the first candidate maximum point.

[0088] For example, when the first candidate maximum point corresponds to the i-th column, the (i-1)-th column and the (i+1)-th column can be determined as the adjacent columns corresponding to the first candidate maximum point.

[0089] Sa3. Calculate the corresponding slope based on the column average of the column containing the first candidate maximum point and the column average of the adjacent columns.

[0090] It should be noted that since the peak points that truly correspond to the steel core locations usually exhibit significant upward or downward changes in the column average sequence, slope calculation can further distinguish between valid peak points and duplicate or interfering peak points. By calculating the corresponding slope based on the column average of the column containing the first candidate maximum point and the column averages of adjacent columns, the degree of change between the first candidate maximum point and the surrounding columns can be characterized, thereby determining whether the first candidate maximum point possesses sufficiently obvious peak characteristics.

[0091] In some implementations, the slope can be calculated based on the difference between the average values ​​of the columns containing the first candidate maximum point and the adjacent columns, combined with the positional differences between columns.

[0092] For example, the slope value can be obtained by dividing the difference between the column average of the first candidate maximum point and the column average of its adjacent columns by the corresponding column spacing.

[0093] Sa4. If the absolute value of the slope is greater than the preset slope threshold, determine the first candidate maximum point as the target maximum point.

[0094] It should be noted that by comparing the absolute value of the slope with a preset slope threshold, the first candidate maximum point can be further screened to determine the target maximum point. In other words, when the degree of change between the first candidate maximum point and its adjacent columns meets the preset requirements, it indicates that the first candidate maximum point can well characterize the peak feature corresponding to the steel core position, and therefore it can be determined as the target maximum point. Conversely, when the absolute value of the slope does not reach the preset slope threshold, it indicates that the degree of change between the first candidate maximum point and its adjacent columns does not meet the preset requirements, and the peak feature corresponding to the first candidate maximum point is not obvious, therefore it is not considered a target maximum point.

[0095] In some implementations, the preset slope threshold can be set by those skilled in the art based on actual detection needs and the changes in the average value sequence of the column; this embodiment does not impose specific limitations on this.

[0096] For example, the preset slope threshold is 0.5.

[0097] In this embodiment, by first performing extreme value detection on the column average sequence to obtain at least one first candidate maximum point, then determining the corresponding adjacent column for each first candidate maximum point, and calculating the corresponding slope based on the column average of the column where the first candidate maximum point is located and the column average of the adjacent columns, and finally determining the corresponding first candidate maximum point as the target maximum point when the absolute value of the slope is greater than a preset slope threshold, this can avoid misjudging candidate maximum points with indistinct peak features or those generated by local fluctuations as the target position corresponding to the steel core, thereby improving the accuracy of the target maximum point determination and providing a more accurate peak basis for the reliable identification of the subsequent steel core target position.

[0098] In one possible embodiment, the method steps shown in S104 can be implemented by S1041 to S1045, which are described in detail below.

[0099] S1041. Calculate the row average of each row value in the standardized two-dimensional coordinate matrix to obtain the row average sequence.

[0100] In this embodiment, since steel core defects typically correspond to abnormal changes in the row direction in a standardized two-dimensional coordinate matrix, the matrix data originally represented in two-dimensional form can be converted into one-dimensional sequence data that changes along the row direction by calculating the row average value of each row value in the standardized two-dimensional coordinate matrix. This reduces the data processing complexity of subsequent defect extraction and provides a serialized data basis for the subsequent extraction of steel core defect locations.

[0101] In some implementations, the average value of each row in the standardized two-dimensional coordinate matrix can be calculated separately, and the obtained average values ​​can be arranged in the order of row index to form a row average value sequence.

[0102] For example, when the standardized two-dimensional coordinate matrix is ​​a 1024×768 matrix, a sequence of row averages of length 768 can be obtained by calculating the average value of each of the 768 rows.

[0103] S1042. Perform extreme value detection on the row average sequence to obtain at least one candidate minimum point.

[0104] In this embodiment, since the steel core defect region usually shows a relatively low response position in the row average sequence, by performing extreme value detection on the row average sequence, abnormal low value points corresponding to the steel core defect position can be initially identified from the row average sequence, thereby obtaining at least one candidate minimum value point.

[0105] In some implementations, an extreme value detection algorithm can be used to detect local low value positions in the row average sequence, thereby obtaining at least one candidate minimum point.

[0106] For example, a sequence extremum search method can be used to detect local valley locations in the row average sequence to obtain at least one candidate minimum point.

[0107] S1043. Determine at least one initial defect interval based on at least one candidate minimum point.

[0108] In this embodiment, by determining the initial defect interval based on candidate minimum points, the single-point anomaly features in the row average sequence can be further expanded into interval features with range information. This allows the defect extraction results to not only reflect the center location of the defect but also characterize the distribution range of the defect in the row direction. That is, candidate minimum points are used to indicate the anomaly center corresponding to the defect region, while the initial defect interval is used to characterize the continuous anomaly range corresponding to the anomaly center.

[0109] In some implementations, the initial defect interval corresponding to each candidate minimum point can be determined by taking each candidate minimum point as the center and combining its adjacent changes in the row average sequence.

[0110] For example, the range corresponding to the candidate minimum point can be determined based on the continuity of the numerical changes on both sides of the candidate minimum point or the boundary changes, and this range can be used as the initial defect range.

[0111] S1044. Based on the preset length ratio threshold and interval inclusion relationship, at least one initial defect interval is screened to obtain the target defect interval.

[0112] In this embodiment, since the initial defect interval obtained from the preliminary determination of candidate minimum points may contain invalid intervals formed by local fluctuations, or may contain overlapping or mutually contained duplicate intervals, by filtering the initial defect interval based on a preset length ratio threshold and interval inclusion relationship, interference intervals that are too short or duplicate intervals can be removed, thereby improving the accuracy of the determination of the target defect interval.

[0113] In some implementations, before filtering the initial defect intervals, the average of row 1 and row 768 can be calculated as a reference for judging the first and last intervals. Since the first and last regions are at the boundary of the row average sequence, the reference information is relatively limited. Therefore, by introducing the average of row 1 and row 768 as a supplementary reference, a basis can be provided for judging defects in the first and last regions.

[0114] Figure 4a This is a comparison diagram before and after optimization of defect detection in the beginning and end intervals, provided as an embodiment of this application. Figure 4a It can be seen that in the process of determining candidate minimum points by performing extreme value detection based on the row average sequence, and further determining the initial defect interval based on the candidate minimum points, the abnormal features located at the beginning and end of the row average sequence are relatively limited in reference information due to their boundary positions, which can easily affect the judgment of the initial defect interval corresponding to the beginning and end intervals. Therefore, in some implementations, after determining the initial defect interval based on the candidate minimum points, the average of the first row and the 768th row can be calculated as a reference for judging the beginning and end intervals, thereby supplementing the identification of abnormal features corresponding to the beginning and end intervals, and combining the preset length ratio threshold and interval inclusion relationship to filter the initial defect intervals to obtain the target defect interval. Figure 4b This application provides an optimized detection result diagram for defect detection in the first and last intervals, as shown in the embodiments of this application. Figure 4b It can be seen that the detection method provided in this application can effectively identify the location of defects in the first and last sections of the steel core.

[0115] In some implementations, initial defect intervals with a length less than a preset length ratio threshold can be removed, and multiple initial defect intervals with an inclusion relationship can be merged or deleted to obtain the target defect interval.

[0116] For example, a preset length ratio threshold can be set to 0.05, and initial defect intervals with a length less than 5% of the total number of lines can be deleted; for multiple initial defect intervals that have an inclusion relationship, the included duplicate intervals can be deleted to obtain the target defect interval.

[0117] Figure 5aThis is a schematic diagram illustrating how to filter initial defect intervals based on interval inclusion relationships, as provided in an embodiment of this application. Figure 5b This is a screenshot showing the detection results after filtering the initial defect intervals based on interval inclusion relationships. Combined with... Figure 5a It can be seen that after obtaining candidate minimum points through extreme value detection based on the row average sequence and determining the initial defect interval based on these candidate minimum points, some initial defect intervals may have inclusion relationships, thus forming duplicate or invalid intervals. Therefore, this application filters the initial defect intervals by combining a preset length ratio threshold and interval inclusion relationships, eliminating duplicate and invalid intervals to obtain the target defect interval. Figure 5b It can be seen that the method of this application can effectively identify and mark multiple defect locations when there are multiple defect locations in the steel core in the same strip core.

[0118] S1045. Based on the target defect range, determine the defect location of the steel core.

[0119] In some implementations, the row range corresponding to each target defect interval can be directly determined as the defect location of the steel core; or, the target defect interval can be mapped to the area where the steel core is located in combination with the target location of the steel core to determine the defect location of the steel core.

[0120] For example, the abnormal area corresponding to the target defect interval can be output as the defect location of the steel core for subsequent defect display and annotation processing.

[0121] In this embodiment, the row average value of each row in the standardized two-dimensional coordinate matrix is ​​calculated to obtain a row average value sequence. Then, extreme value detection is performed on the row average value sequence to obtain at least one candidate minimum point. At least one initial defect interval is determined based on the at least one candidate minimum point. Then, the at least one initial defect interval is filtered based on a preset length ratio threshold and interval inclusion relationship to obtain a target defect interval. Finally, the defect location of the steel core is determined based on the target defect interval. This realizes the transformation of defect anomaly features in the two-dimensional matrix into one-dimensional interval features along the row direction. By effectively filtering the initial defect interval, the interference of invalid intervals and duplicate intervals on defect identification is reduced, thereby improving the accuracy of steel core defect location determination and providing a reliable defect region basis for subsequent image reconstruction and defect annotation.

[0122] In one possible embodiment, the method steps shown in S1042 can be implemented by Sc1 to Sc3, which are described in detail below.

[0123] Sc1. Negate the row average sequence to obtain the negative row average sequence.

[0124] In this embodiment, by negativeening the row average sequence, the minimum value feature originally used to characterize the defect location can be converted into a maximum value feature, which facilitates the subsequent identification of the defect candidate location using the maximum value detection method used in S1032 of the above embodiment.

[0125] In some implementations, each value in the row average sequence can be multiplied by a negative one to obtain a negative row average sequence. The negative row average sequence has the same length as the original row average sequence.

[0126] Sc2. Perform maximum detection on the negative row average sequence to obtain at least one second candidate maximum point.

[0127] In this embodiment, since the defective region is a local low value position in the original row average sequence, the corresponding position will be a local high value position in the negative row average sequence. By performing maximum detection on the negative row average sequence, the peak point corresponding to the minimum value position in the original row average sequence can be extracted from the negative row average sequence, thereby obtaining at least one second candidate maximum point.

[0128] Sc3. Determine the corresponding point in the row average sequence of at least one second candidate maximum point as at least one candidate minimum point.

[0129] In this embodiment, by mapping the second candidate maximum point detected in the negative row average sequence back to the corresponding position in the original row average sequence, the candidate minimum point in the original row average sequence can be obtained.

[0130] In some implementations, the point corresponding to the same index position in the original row average sequence can be found based on the sequence index position corresponding to the second candidate maximum point, and this point can be determined as the candidate minimum point.

[0131] For example, when the second candidate maximum point is located at the j-th position in the negative row average sequence, the point corresponding to the j-th position in the original row average sequence can be determined as the candidate minimum point.

[0132] In this embodiment, by negativeing ​​the row average sequence, the minimum feature originally used to characterize the defect location is converted into the maximum feature in the negative row average sequence. Then, the negative row average sequence is subjected to maximum detection, and the detected second candidate maximum point is mapped back to the corresponding point in the original row average sequence and determined as the candidate minimum point. This realizes the transformational extraction of the candidate location of steel core defects, so that the minimum location that is not easy to detect directly can be identified by the maximum detection method. This makes it easier to extract the candidate minimum point corresponding to the defect location from the row average sequence, and provides a basis for the subsequent determination of the initial defect interval.

[0133] In one possible embodiment, the method steps shown in S105 can be implemented by S1051 and S1052, which are described in detail below.

[0134] S1051. Create a blank image with dimensions consistent with the standardized two-dimensional coordinate matrix.

[0135] In this embodiment, since the data at each position in the standardized two-dimensional coordinate matrix corresponds to the spatial position in the core scanning area, by creating a blank image with the same size as the standardized two-dimensional coordinate matrix, the subsequent assignment results of different regions can be kept consistent with the positional relationship in the standardized two-dimensional coordinate matrix, thereby facilitating the generation of reconstructed images based on the target position and the defect position.

[0136] In some implementations, an initial blank image that corresponds to a standardized two-dimensional coordinate matrix in length and width can be created as the base image for reconstructing the image.

[0137] For example, when the normalized two-dimensional coordinate matrix is ​​a 1024×768 matrix, a blank image of size 1024×768 can be created for subsequent image reconstruction processing.

[0138] S1052. In the blank image, the first interval corresponding to the defect location is assigned the value of black, the area in the second interval corresponding to the target location other than the first interval is assigned the value of red, and the area in the blank image other than the first interval and the second interval is assigned the value of white, thus obtaining the reconstructed image.

[0139] In this embodiment, by classifying and assigning values ​​to different regions in the blank image, the steel core defect region, the steel core non-defect region, and the background region can be distinguished and displayed, so that the different regions in the subsequently obtained reconstructed image have a clear distinction relationship, thereby facilitating the subsequent observation and annotation of the steel core defect location.

[0140] Specifically, the first interval corresponding to the defect location is assigned a black value to represent the abnormal area in the steel core; the area in the second interval corresponding to the target location other than the first interval is assigned a red value to represent the non-defect area in the steel core; and the area other than the first and second intervals is assigned a white value to represent the background area.

[0141] In some implementations, pixel values ​​can be assigned to different regions in a blank image based on the corresponding positions of the target and defect locations in the blank image, thereby obtaining a reconstructed image.

[0142] For example, the pixels corresponding to the first interval can be assigned the value (0,0,0), the pixels corresponding to the area outside the first interval in the second interval can be assigned the value (255,0,0), and the pixels corresponding to the area outside the first and second intervals can be assigned the value (255,255,255), thereby completing the image reconstruction.

[0143] In this embodiment, a blank image with the same size as the standardized two-dimensional coordinate matrix is ​​created. In the blank image, the first interval corresponding to the defect location is assigned black, the area in the second interval corresponding to the target location other than the first interval is assigned red, and the area other than the first and second intervals is assigned white. This achieves the distinguishing display of the steel core defect area, the steel core non-defect area and the background area, so that the steel core distribution state and the defect distribution state can be presented intuitively in the form of a reconstructed image, providing a clear image basis for the subsequent annotation and observation of the steel core defect location.

[0144] In one possible embodiment, the method steps shown in S106 can be implemented by S1061 to S1063, which are described in detail below.

[0145] S1061. Determine the center coordinates of the defect location based on the row and column ranges corresponding to the defect location in the reconstructed image.

[0146] In this embodiment, since the defect location usually corresponds to a certain range of areas in the reconstructed image, the center coordinates of the defect location can be determined according to the row and column ranges corresponding to the defect location. This can convert the defect location from an interval range form to a center point form, thereby providing a positional basis for the determination of subsequent annotation parameters and the drawing of annotation boxes.

[0147] In some implementations, the center position of the row direction can be determined based on the starting and ending rows corresponding to the defect location, and the center position of the column direction can be determined based on the starting and ending columns corresponding to the defect location, thereby obtaining the center coordinates of the defect location.

[0148] For example, the midpoint of the column range corresponding to the defect location can be determined as the x-coordinate of the center coordinate, and the midpoint of the row range corresponding to the defect location can be determined as the y-coordinate of the center coordinate.

[0149] S1062. Determine the annotation parameters based on the row length and column length corresponding to the defect location.

[0150] In this embodiment, since the range of the intervals corresponding to different defect locations may differ, determining the annotation parameters based on the row length and column length corresponding to the defect location can enable the annotation results to adapt to defect areas of different sizes, thereby improving the accuracy of the annotation results in indicating the defect area.

[0151] In some implementations, the height of the annotation box can be determined based on the row length corresponding to the defect location, and the width of the annotation box can be determined based on the column length corresponding to the defect location, thereby obtaining the annotation parameters.

[0152] For example, the width of the rectangle can be determined by 1.2 times the column length corresponding to the defect location, and the height of the rectangle can be determined by 1.2 times the row length corresponding to the defect location, thus obtaining the annotation parameters.

[0153] S1063. Based on the center coordinates and annotation parameters, draw a rectangular box in the reconstructed image to obtain the annotation result; wherein, the annotation parameters include the width and height of the rectangular box.

[0154] In this embodiment, since the center coordinates can characterize the center position of the defect in the reconstructed image, and the annotation parameters can characterize the size range of the rectangular box, by combining the center coordinates and annotation parameters, a corresponding annotation box can be formed in the reconstructed image, so that the defect position can be displayed more intuitively in the reconstructed image, thereby facilitating the identification and location of the defect area.

[0155] In some implementations, the drawing position of the rectangle can be determined based on the center coordinates, and the width and height of the rectangle can be determined based on the annotation parameters, thereby drawing the rectangle in the reconstructed image and obtaining the annotation result.

[0156] For example, the cv2.rectangle function in the OpenCV library can be used to draw a rectangle in the reconstructed image based on the center coordinates and the width and height of the rectangle to obtain the annotation results.

[0157] In this embodiment, the center coordinates of the defect location are determined based on the row and column ranges corresponding to the defect location. Then, the annotation parameters are determined based on the row and column lengths corresponding to the defect location. A rectangular box is drawn in the reconstructed image based on the center coordinates and annotation parameters. This achieves the visualization annotation of the steel core defect location, so that the annotation results can correspond to the actual location and range of the defect area, thereby facilitating the identification and positioning of steel core defects.

[0158] This application provides a terahertz-based conveyor belt core detection device, which includes a heat map acquisition module, a standardized two-dimensional coordinate matrix generation module, a steel core target position determination module, a steel core defect position determination module, a reconstructed image generation module, a labeling result generation module, and a result output module.

[0159] The heat map acquisition module is used to scan the core of the conveyor belt using a terahertz scanning device to obtain the corresponding heat map of the core. The standardized two-dimensional coordinate matrix generation module is used to extract channels and standardize the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core. The steel core target position determination module is used to perform longitudinal projection and positioning processing on the standardized two-dimensional coordinate matrix to obtain the target position of the steel core in the strip core. The steel core defect location determination module is used to perform lateral projection and defect extraction processing on the standardized two-dimensional coordinate matrix to obtain the defect location of the steel core; The reconstructed image generation module is used to generate reconstructed images based on the target location and the defect location; The annotation result generation module is used to annotate the steel core defect locations in the reconstructed image based on the center coordinates and annotation parameters of the defect locations, and obtain the annotation results; The results output module is used to output the reconstructed image and annotation results.

[0160] It should be noted that the specific process of each module in the detection device executing the above method has been described in detail in the above embodiments, and this embodiment does not make specific limitations on it.

[0161] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 600 provided in this embodiment includes a memory 601 and a processor 602.

[0162] The memory 601 can be a separate physical unit, connected to the processor 602 via a bus 603. Alternatively, the memory 601 and processor 602 can be integrated and implemented in hardware. The memory 601 stores program instructions, which the processor 602 calls to execute the operations performed by the detection device in any of the above method embodiments.

[0163] Optionally, when some or all of the methods in the above embodiments are implemented by software, the electronic device 600 may also include only the processor 602. A memory 601 for storing programs is located outside the electronic device 600, and the processor 602 is connected to the memory via circuits / wires to read and execute the programs stored in the memory. The processor 602 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. The processor 602 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0164] The memory 601 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory may also include a combination of the above types of memory.

[0165] For example, this application provides a chip including: an interface circuit and a logic circuit. The interface circuit is used to receive signals from other chips outside the chip and transmit them to the logic circuit, or to send signals from the logic circuit to other chips outside the chip. The logic circuit is used to perform the operations performed by the detection device in the above method embodiments.

[0166] For example, this application provides a computer-readable storage medium storing computer program instructions thereon, which are executed by the processor of an electronic device to cause the electronic device to perform the operations performed by the detection device in the above method embodiments.

[0167] For example, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the operations performed by the detection device in the above method embodiments.

[0168] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A terahertz-based method for detecting the core of a conveyor belt, characterized in that, The method includes: The core of the conveyor belt is scanned using a terahertz scanning device to obtain a thermal map of the core. Channel extraction and standardization are performed on the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core. The target position of the steel core in the strip is obtained by longitudinal projection and positioning processing of the standardized two-dimensional coordinate matrix; The defect locations of the steel core are obtained by performing lateral projection and defect extraction on the standardized two-dimensional coordinate matrix. Based on the target location and the defect location, a reconstructed image is generated; Based on the center coordinates and annotation parameters of the defect location, the steel core defect location in the reconstructed image is annotated to obtain the annotation result; Output the reconstructed image and the annotation results.

2. The method according to claim 1, characterized in that, The heatmap is in RGB format; The process of channel extraction and standardization of the heat map to obtain the standardized two-dimensional coordinate matrix corresponding to the core includes: Extract the red channel values ​​from the heatmap; A two-dimensional coordinate matrix is ​​constructed based on the red channel values; The two-dimensional coordinate matrix is ​​standardized to obtain the standardized two-dimensional coordinate matrix.

3. The method according to claim 2, characterized in that, The two-dimensional coordinate matrix is ​​a 1024×768 matrix, and the numerical value range of the standardized two-dimensional coordinate matrix is ​​0 to 255.

4. The method according to claim 1, characterized in that, The process of longitudinally projecting and positioning the standardized two-dimensional coordinate matrix to obtain the target position of the steel core in the strip includes: Calculate the column average of each column value in the standardized two-dimensional coordinate matrix to obtain a column average sequence; Maximum detection is performed on the average value sequence of the column to obtain the target maximum point; The target position of the steel core is determined based on the column position of the target maximum point.

5. The method according to claim 4, characterized in that, The step of performing extreme value detection on the average value sequence of the column to obtain the target maximum point includes: Extreme value detection is performed on the average value sequence of the column to obtain at least one first candidate maximum point; For each of the at least one first candidate maximum point, the adjacent column corresponding to the first candidate maximum point is determined according to the column position corresponding to the first candidate maximum point; The slope is calculated based on the column average of the column containing the first candidate maximum point and the column average of the adjacent columns. If the absolute value of the slope is greater than a preset slope threshold, the first candidate maximum point is determined as the target maximum point.

6. The method according to claim 1, characterized in that, The process of performing lateral projection and defect extraction on the standardized two-dimensional coordinate matrix to obtain the defect location of the steel core includes: Calculate the row average of the values ​​in each row of the standardized two-dimensional coordinate matrix to obtain a sequence of row averages; Extreme value detection is performed on the row average sequence to obtain at least one candidate minimum point; Based on the at least one candidate minimum point, at least one initial defect interval is determined; Based on a preset length ratio threshold and interval inclusion relationship, the at least one initial defect interval is filtered to obtain the target defect interval; Based on the target defect range, the defect location of the steel core is determined.

7. The method according to claim 6, characterized in that, The step of performing extreme value detection on the row average sequence to obtain at least one candidate minimum point includes: The row average sequence is negativeed to obtain a negative row average sequence. Maximum detection is performed on the negative row average sequence to obtain at least one second candidate maximum point; The corresponding point of the at least one second candidate maximum point in the row average sequence is determined as the at least one candidate minimum point.

8. The method according to claim 1, characterized in that, The step of generating a reconstructed image based on the target location and the defect location includes: Create a blank image with the same dimensions as the standardized two-dimensional coordinate matrix; In the blank image, the first interval corresponding to the defect location is assigned a black value, the area in the second interval corresponding to the target location other than the first interval is assigned a red value, and the area in the blank image other than the first interval and the second interval is assigned a white value, thus obtaining the reconstructed image.

9. The method according to claim 1, characterized in that, The process of annotating the steel core defect location in the reconstructed image based on the center coordinates and annotation parameters of the defect location to obtain the annotation result includes: The center coordinates of the defect location are determined based on the row and column ranges corresponding to the defect location in the reconstructed image. The annotation parameters are determined based on the row length and column length corresponding to the defect location; Based on the center coordinates and the annotation parameters, a rectangular box is drawn in the reconstructed image to obtain the annotation result; wherein, the annotation parameters include the width and height of the rectangular box.

10. A terahertz-based conveyor belt core detection device, characterized in that, Includes memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any one of claims 1 to 9.

Images