Multi-probe based ultrasonic inspection system and method

Abstract

The present invention provides a multi-probe-based ultrasound inspection system and method capable of providing uniform, high-resolution integrated images by resolving the problem of reduced inspection accuracy caused by image overlap and non-uniformity of image quality among multiple ultrasound probes. An ultrasonic inspection system according to one aspect of the present invention is an ultrasonic inspection system for inspecting a sample using multiple ultrasonic probes, comprising a memory configured to store commands and a processor configured to perform the following operations by executing said commands. The processor transmits an ultrasonic signal to the surface of a sample through a plurality of probes, receives an ultrasonic echo signal reflected from the sample to acquire ultrasonic data for each probe, generates an ultrasonic image for each probe based on the ultrasonic data acquired from the plurality of probes, analyzes image data of overlapping regions existing between the ultrasonic images for each probe and the differences in image quality characteristics of the images for each probe, adjusts at least one image quality control factor among brightness, contrast, resolution, and gain of each ultrasonic image for each probe based on the analyzed result to uniformly correct the image quality of the ultrasonic images between probes, fuses the corrected ultrasonic images for each probe and image data of the overlapping regions to generate a single integrated inspection image, and is configured to output the generated inspection image to a display device.

Description

Technology Field

[0001] The present invention relates to a multi-probe-based ultrasonic inspection system and method for inspecting a sample at high speed using multiple ultrasonic probes and correcting image overlap and non-uniformity of image quality between probes. Background Technology

[0002] Recently, ultrasonic inspection technology is being widely utilized in manufacturing processes for semiconductor wafers, electronic devices, and precision components to non-destructively inspect for micro-defects within products. Ultrasonic inspection offers the advantage of identifying the presence of defects, cracks, delamination, and voids within samples at high resolution, establishing itself as an essential non-destructive testing (NDT) technology for quality control and process management. In particular, for microstructures such as semiconductor chips-on-wafers, precise imaging using high-frequency ultrasonic probes is required to detect internal defects that cannot be observed with the naked eye from the outside.

[0003] However, since conventional ultrasonic inspection technology primarily operates based on a single probe, there are limitations in inspection speed for samples with large inspection areas or high resolution. Because the single-probe method scans only one point at a time, multiple scan lines must be traversed sequentially to inspect the entire sample, which prolongs the overall inspection time and reduces production efficiency. To address this problem, a multichannel or multi-probe method that operates multiple probes simultaneously has been proposed; however, applying multiple probes introduces new technical challenges.

[0004] First, image overlap issues occur between probes. Multiple probes are arranged at a constant pitch interval in the X and Y directions, and each probe scans adjacent areas simultaneously. During this process, the irradiation areas of each probe partially overlap, resulting in the duplicate detection of ultrasonic reflection signals from the same sample area. This duplicated data causes excessive emphasis or distortion of brightness or contrast during the imaging process, thereby reducing the coherence of the overall image. In particular, in boundary zones, differences in signal intensity cause seam lines or banding, which lowers the accuracy of defect detection.

[0005] Second, there is non-uniformity in image quality due to differences in characteristics among probes. Even if ultrasonic probes are designed for the same frequency band, their sensitivity, gain, and frequency response characteristics differ slightly depending on actual manufacturing tolerances or operating environments (temperature, medium conditions, contact pressure, etc.). These differences in physical characteristics between probes affect brightness, contrast, and resolution during image generation, leading to visual non-uniformity between images from different probes. Consequently, a single inspection image appears as if multiple partial images have been simply stitched together, making it difficult to identify the boundaries of actual defects.

[0006] Third, interference and signal distortion between probes are problematic. When multiple probes transmit ultrasound simultaneously, the transmitted signal from an adjacent probe interferes with the received signal of another probe, causing distortion of the echo signal. This results in some areas of the image becoming blurred or noise increasing, and may reduce the accuracy of detecting fine defects in the high-frequency band. Additionally, if independent signal amplifiers and data processing circuits are used for each probe, timing delays or phase differences accumulate, making image matching difficult.

[0007] Fourth, there is the complexity of the post-processing of image data. Ultrasound data acquired from multiple probes generates individual images corresponding to the number of probes, and complex alignment, normalization, correction, and fusion processes are required to combine them into a single integrated image. In particular, if the signal weighting factor in overlapping regions is not properly adjusted, the balance of the entire image is compromised. Existing simple averaging or linear correction methods make it difficult to completely eliminate differences in brightness and contrast, and struggle to resolve the problem of distinct boundaries appearing in certain areas of the image.

[0008] Consequently, although multi-probe ultrasonic inspection technology holds the potential to improve inspection speed, it faces limitations in widespread industrial application due to issues such as handling overlapping areas between probes, ensuring image quality uniformity, eliminating signal interference, and ensuring the alignment of image synthesis. In particular, in fields requiring high resolution and precise boundary detection, such as semiconductor wafers or precision metal materials, these image non-uniformity and redundancy problems lead to reduced inspection reliability and decreased accuracy in defect detection.

[0009] Therefore, a technology is required that can generate a single, integrated, high-quality ultrasound image by using multiple probes, correcting differences between probe images, and uniformly fusing overlapping boundary regions. This technology must be able to achieve a clean image of the same quality as one acquired with a single probe by adjusting image parameters such as brightness, contrast, resolution, and gain of each probe image, and adaptively correcting overlapping regions. The problem to be solved

[0010] The present invention has been devised to solve the aforementioned problems, and aims to provide a multi-probe-based ultrasonic inspection system and method capable of providing a uniform, high-resolution integrated image by resolving the problem of reduced inspection accuracy caused by image overlap and non-uniformity of image quality among multiple ultrasonic probes.

[0011] However, the technical problems that the present invention aims to solve are not limited to those described above, and other unmentioned problems will be clearly understood by those skilled in the art from the description of the invention below. means of solving the problem

[0012] An ultrasonic inspection system according to one aspect of the present invention is an ultrasonic inspection system for inspecting a sample using multiple ultrasonic probes, comprising a memory configured to store commands and a processor configured to perform the following operations by executing said commands. The processor transmits an ultrasonic signal to the surface of a sample through a plurality of probes, receives an ultrasonic echo signal reflected from the sample to acquire ultrasonic data for each probe, generates an ultrasonic image for each probe based on the ultrasonic data acquired from the plurality of probes, analyzes image data of overlapping regions existing between the ultrasonic images for each probe and the differences in image quality characteristics of the images for each probe, adjusts at least one image quality control factor among brightness, contrast, resolution, and gain of each ultrasonic image for each probe based on the analyzed result to uniformly correct the image quality of the ultrasonic images between probes, fuses the corrected ultrasonic images for each probe and image data of the overlapping regions to generate a single integrated inspection image, and is configured to output the generated inspection image to a display device.

[0013] Preferably, the processor may be configured to correct image quality deviations between probes by controlling at least one image quality control factor among brightness, contrast, resolution, and gain of the ultrasound image for each probe according to a parameter level calculated from the ultrasound image for each probe.

[0014] Preferably, the processor is, After identifying overlapping image data in the boundary regions between probes and calculating the reliability weights of ultrasound images for each probe based on the difference in image quality adjustment factors within the overlapping regions, The corrected image data to which the above weights have been applied can be configured to be fused using interpolation or seamless blending.

[0015] Preferably, the processor is, It can be configured to set a reference probe and detect differences in brightness, contrast, and resolution of the non-reference probe image to perform adaptive calibration to minimize the error with the reference probe image.

[0016] Preferably, the processor is, Automatically updates the reference values ​​of image quality adjustment factors based on overlapping areas or statistical characteristic values ​​of the entire image, and adjusts image correction parameters in response to environmental changes per probe. It can be configured to update adaptively.

[0017] A multi-probe based ultrasound examination method according to one aspect of the present invention is, Transmitting an ultrasonic signal to the surface of a sample through a plurality of probes and receiving an ultrasonic echo signal reflected from the sample Thus, the step of acquiring ultrasound data for each probe, and Ultrasonic data obtained from the above plurality of probes A step of generating ultrasound images per probe based on, and Existence between the ultrasound images for each of the above probes doing Differences in image data in overlapping areas and image quality characteristics per probe The step of analyzing, and Based on the analyzed results Therefore, each Ultrasound images by probe By adjusting at least one image quality control factor among brightness, contrast, resolution, and gain, between probes Ultrasound imaging A step for uniformly correcting the image quality and The above-mentioned corrected ultrasound images by probe and image data of the overlapping regions are fused. so Generates a single, integrated inspection image ...and includes the step of outputting the generated inspection image to a display device. Effects of the invention

[0018] The multi-probe-based ultrasonic inspection system of the present invention has the effect of maintaining uniform contrast and resolution of the entire image by simultaneously transmitting ultrasonic signals from multiple probes and receiving reflected signals to generate images for each probe, and then precisely correcting overlapping areas and image quality non-uniformity occurring at the boundaries between probes. In particular, since the processor automatically analyzes and adaptively adjusts parameters such as brightness, contrast, resolution, and gain of each probe image, image quality degradation caused by sensitivity differences or interference between probes can be minimized. Accordingly, a unified high-resolution image can be secured across the entire area of ​​the sample, the reliability of inspection results is improved, and the inspection speed is dramatically improved compared to a single-probe method.

[0019] In addition, various other additional effects may be achieved by various embodiments of the present invention. These various effects of the present invention are described in detail in each embodiment, or the description of effects that are easily understood by those skilled in the art is omitted. Brief explanation of the drawing

[0020] The following drawings attached to this specification illustrate preferred embodiments of the present invention and serve to further enhance understanding of the technical concept of the present invention together with the detailed description of the invention provided below; therefore, the present invention should not be interpreted as being limited only to the matters described in such drawings. FIG. 1 is a drawing showing a system according to one embodiment of the present invention. FIGS. 2 to 6 are drawings illustrating an example of an ultrasound examination method by virtual line-based image synthesis performed by the system of FIG. 1. FIG. 7 is a flowchart illustrating an ultrasound examination method based on virtual line-based image synthesis according to an embodiment of the present invention. Figures 8 and 9 are diagrams illustrating an example of a multi-probe-based ultrasound examination method performed by the system of Figure 1. FIG. 10 is a flowchart illustrating a multi-probe-based ultrasound examination method according to one embodiment of the present invention. Specific details for implementing the invention

[0021] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. Prior to this, terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, and should be interpreted in a meaning and concept consistent with the technical spirit of the present invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.

[0022] Therefore, the embodiments described in this specification and the configurations illustrated in the drawings are merely the most preferred embodiments of the present invention and do not represent all of the technical ideas of the present invention; thus, it should be understood that various equivalents and modifications that can replace them may exist at the time of filing this application.

[0023] FIG. 1 is a drawing showing a system (100) according to one embodiment of the present invention, and FIG. 2 to 6 are drawings showing an example of an ultrasound inspection method by virtual line-based image synthesis performed by the system (100) of FIG. 1.

[0024] Referring to FIGS. 1 to 6, a system (100) according to one embodiment of the present invention may include a memory (110) and a processor (120).

[0025] The memory (110) above can be configured to store instructions.

[0026] The processor (120) may be configured to perform the operations described below by executing instructions. Memory (110) may be configured to store instructions of a computer program or application, and the processor (120) may execute the program or application by executing instructions stored in memory (110). For example, memory (110) may be implemented as non-volatile memory such as ROM, PROM, EPROM, EEPROM, flash memory, PRAM, MRAM, RRAM, FRAM, etc., or volatile memory such as DRAM, SRAM, SDRAM, PRAM, RRAM, FeRAM, etc., and may be implemented in the form of HDD, SSD, SD, Micro-SD, etc., or a combination thereof. The processor (120) may be implemented as an array of logic gates, a microprocessor, a CPU, a GPU, an AP, or a combination thereof.

[0027] Specifically, the processor (120) can acquire image data corresponding to each scanning line by controlling the probe to scan the interior or surface of the sample in units of multiple scanning lines. Here, a scanning line refers to a data line in which a reflected signal is collected while an ultrasonic probe moves along a certain section of the sample. As shown in FIG. 2, the probe moves in the direction of the X-axis or Y-axis and scans multiple lines sequentially, and images the reflected signal for each line. In conventional technology, since all lines had to be scanned continuously, the inspection time was prolonged, and there was a problem in that the inspection efficiency decreased rapidly as the pitch interval became narrower.

[0028] Additionally, the processor (120) can generate an image of a virtual line corresponding between two or more adjacent scanning lines by using image data obtained from two or more adjacent scanning lines among a plurality of scanning lines, through interpolation or weighted average calculation. As illustrated in FIGS. 3 and 4, for example, an image of virtual line 2 can be generated using data from scanning line 1 and scanning line 3, and virtual line 4 can be generated using data from scanning lines 3 and 5 in the same way. In this case, the virtual line is an area that is not physically scanned, but is a composite image calculated based on signal distribution, reflection intensity, and phase information between two adjacent actual lines. This weighted average calculation is performed by considering signal strength, reflection pattern, noise ratio, etc., and can minimize image distortion through non-linear correction or pixel-level adaptive weight calculation beyond simple linear interpolation.

[0029] Additionally, the processor (120) can construct an overall inspection image of the sample by synthesizing the scanning line image and the virtual line image. As illustrated in FIG. 4, the processor can generate a corresponding virtual line using information from an adjacent line whenever each scanning line data is acquired, and output a composite image in real time. In addition, as another embodiment, the final composite image can be completed by sequentially interpolating the virtual line image in a post-processing step after all scanning line data has been acquired (see FIG. 5 and FIG. 6). In this process of constructing a composite image, the processor (120) can maximize inspection efficiency by minimizing the number of virtual lines in parts of the entire sample where defects are likely to occur (e.g., edge areas, joint areas, pattern-dense areas) to prioritize the reflection of actual scan data, and by configuring multiple virtual lines in homogeneous areas. That is, the system (100) of the present invention includes a structure that can dynamically adjust the density or number of generation of virtual lines according to the importance of the inspection area.

[0030] In addition, the processor (120) can automatically adjust weights during interpolation operations based on the signal reliability and defect sensitivity of each scanning line or virtual line. For example, high weights can be applied to edge regions or heterogeneous joints of materials that are highly likely to have defects to induce detailed data reflection, and low weights can be applied to homogeneous internal regions to increase processing speed. Through such weight control, boundary distortion that may occur during image synthesis can be prevented, and the reliability of defect detection can be increased.

[0031] Additionally, the processor (120) may be configured to determine whether there are defects or abnormalities in the sample based on the entire inspection image.

[0032] As illustrated in FIGS. 4 and 5, the synthesized final image includes both a scanning line image and a virtual line image, so that differences in image contrast can be clearly distinguished when internal defects such as voids, cracks, or delamination are present. The defect detection algorithm can be performed based on the rate of change of pixel intensity, the time difference of the reflection signal, the difference in frequency response, etc., and can visualize and output the location, size, and shape of the defect.

[0033] According to an embodiment of the present invention, by generating a virtual line image using data between an actual scanning line and an adjacent line and synthesizing it, the limitations of the conventional ultrasonic inspection method, which requires sequential scanning of all lines, can be overcome, thereby dramatically improving the inspection speed. Furthermore, by processing signals between adjacent lines using weighted averaging or interpolation operations, the continuity and resolution of the image can be maintained, and the reliability of defect detection can be secured while reducing the number of actual scanning lines. Moreover, the number and density of virtual lines can be variably set according to the defect risk or the shape conditions of the sample, thereby maximizing inspection efficiency. Through this, technical effects such as shortening the inspection time of the production line, homogenizing image quality, and improving the accuracy of defect detection can be achieved simultaneously.

[0034] In one embodiment, the processor (120) may be configured to generate an image of a corresponding virtual line between adjacent first scanning lines and second scanning lines by using image data of the two lines and performing a weighted average or interpolation operation.

[0035] As shown in FIGS. 2 to 4, the processor (120) scans the interior or surface of a sample in units of multiple scanning lines and visualizes the ultrasonic reflection signal for each line.

[0036] At this time, the scanning line is a set of actual data formed sequentially along the scanning trajectory of the sample, and since there is structural or physical continuity between adjacent lines, a new intermediate line (virtual line) can be estimated based on the signal change between the two lines.

[0037] For example, when a first scanning line (L1) and a second scanning line (L2) exist as shown in FIG. 3, the processor (120) analyzes the reflection intensity, phase information, signal-to-noise ratio (SNR), and local structure pattern information at the same coordinate location between the two lines, and uses the results to generate a virtual line Lv corresponding to an intermediate location.

[0038] In this process, the processor (120) applies an interpolation or weighted averaging method to expand the discontinuous intervals of the actual data into a continuous signal space. The most basic form of weighted averaging operation can be expressed as shown in the following mathematical formula 1.

[0039] [Mathematical Formula 1]

[0040]

[0041] Here, is the pixel intensity at the coordinates (x,y) of the virtual line Lv, and represents the corresponding pixel intensity in the first and second scanning lines, respectively, and w1 and w2 represent weight factors.

[0042] Weights can be determined based on the distance between two scanning lines, the signal strength ratio, the local noise level, or the ultrasonic reflection coefficient. For example, if the distance Δd between lines is not constant, the following distance-based weighting model can be applied to assign higher weights to adjacent lines.

[0043] [Mathematical Formula 2]

[0044]

[0045] Here, p is the distance decay parameter, and ε is a very small positive constant to prevent convergence to 0.

[0046] This distance-weighted model can ensure stable interpolation performance even in situations where the spacing between lines is non-uniform or the probe's contact position changes slightly depending on the curved shape of the sample.

[0047] Additionally, the processor (120) may reflect various image characteristic factors in addition to simple distance factors when calculating weights. For example, (1) the average reflection intensity of each line (Imean), (2) the phase shift of the signal (Δ (3) local contrast gradient, (4) defect sensitivity index, etc. can be considered in combination.

[0048] In particular, in areas where defects are likely to occur (e.g., edges, junctions, pattern concentration areas), the signal difference between adjacent lines changes rapidly, so the processor can reflect even minute changes in reflection by assigning high sensitivity weights to these locations.

[0049] Conversely, in internal regions where the material is homogeneous, equal weighting or linear interpolation is applied to average the noise to improve image stability.

[0050] The processor (120) can also perform multidimensional interpolation that takes into account signal changes in the time axis or frequency axis.

[0051] For example, by aligning the reflection signal waveforms between adjacent lines at the same coordinate (x, y) along the time axis and then interpolating based on signal amplitude and phase difference, it is possible to reproduce the reflection signal of an intermediate line that does not actually exist. This multidimensional interpolation enhances the depth perception and signal continuity of the image, contributing to increasing the visual homogeneity of the final composite image.

[0052] Meanwhile, the processor (120) evaluates the reliability of the data during the computation process for generating virtual lines and assigns a reliability index at the specific pixel level. Pixels with low reliability are corrected by referencing data from surrounding pixels or other lines, thereby minimizing the occurrence of distortion in candidate defect areas within the image. Additionally, if the interpolation result falls outside the signal spectrum range of the original scanning data, the processor (120) performs an automatic clipping or normalization procedure to maintain image quality.

[0053] The series of steps illustrated in FIGS. 2 to 4 visually illustrates the process of generating such virtual lines. FIG. 2 illustrates a probe forming multiple scanning lines along the surface of a sample, and FIG. 3 illustrates the process of generating an intermediate virtual line (L2) from two adjacent scanning lines (L1, L3). FIG. 4 illustrates a sequential procedure in which the entire image is completed through interpolation and synthesis processes between multiple lines. With this configuration, the processor (120) of the present invention can dramatically improve inspection speed while maintaining spatial continuity of the image and without lowering the reliability of defect detection, while reducing the number of actual scan lines.

[0054] In one embodiment, the processor (120) may be configured to actually scan an odd-numbered scanning line and generate a virtual line image corresponding to an even-numbered line between them, or to widen the scanning interval of the scanning line to continuously generate a plurality of virtual lines.

[0055] As shown in FIGS. 2 to 4, the processor (120) scans the surface of a sample (e.g., semiconductor wafer, metal composite material, etc.) at regular intervals and forms a plurality of scanning lines. In this embodiment, all lines are not scanned sequentially, but only the scanning lines corresponding to odd numbers (e.g., 1, 3, 5, 7, etc.) are actually scanned, and the even lines (2, 4, 6, etc.) between them are set as virtual lines to generate image data through interpolation or weighted averaging operations.

[0056] This method provides a highly efficient structure that can maintain image continuity and detailed structural information while reducing the total number of scans to less than half.

[0057] FIG. 2 illustrates the basic arrangement concept between these scanning lines and virtual lines, and FIG. 3 shows the process of generating virtual lines corresponding to even lines from actual data of odd lines. For example, a virtual line corresponding to line 2 (L2) is generated using data from line 1 (L1) and line 3 (L3), and an image of line 4 (L4) is synthesized using data from line 3 (L3) and line 5 (L5) in the same way. At this time, the processor (120) analyzes the interval Δd between each scanning line and the signal correlation to calculate weights based on spatial position and determines the pixel value of the intermediate line based on this.

[0058] Each pixel value of the virtual line Li It can be calculated using the following mathematical formula 3.

[0059] [Mathematical Formula 3]

[0060]

[0061] Here, and are the pixel values ​​at the same coordinates on two adjacent odd scanning lines (e.g., L₁, L₃), respectively. is a distance-based or signal quality-based weight factor, is the pixel intensity of the virtual line corresponding to the even line between the two lines.

[0062] weight It can be variably set based on various factors, such as not only simple distance but also signal-to-noise ratio (SNR), reflection intensity, phase difference, and defect sensitivity. For example, in areas where the difference in reflection signal intensity is large, a larger weight is given to the scanning line with higher reliability, and when the signals between the two lines are similar, equal weighting is applied to ensure image homogeneity.

[0063] Distance-based weights can be defined by the following mathematical formula 4.

[0064] [Mathematical Formula 4]

[0065]

[0066] Here, is the distance between adjacent scanning lines, is the decay coefficient, is a very small constant for computational stability. This model provides stable interpolation results even when scanning intervals are non-uniform or there are height differences between lines due to the curved shape of the sample.

[0067] In addition, in this embodiment, the processor (120) can actively adjust the scanning interval of the scanning lines according to inspection conditions. For example, when the surface of the sample is homogeneous and the probability of defect occurrence is low, the spacing between scanning lines is widened, and multiple virtual lines are continuously generated between them to improve the overall inspection speed. Conversely, in areas where the probability of defect occurrence is high (e.g., joints, pattern dense areas, material boundaries, etc.), the spacing between scanning lines is narrowed to ensure precision based on actual lines, or the accuracy of generating virtual lines in the area can be increased by adjusting the weighting function.

[0068] FIG. 4 illustrates an example in which the scanning interval of the scanning lines is extended to set lines 1, 4, and 7 as actual lines, and lines 2, 3, 5, and 6 in between as virtual lines. In this case, the processor (120) performs multi-step interpolation by considering the distance ratio between each line and the defect sensitivity index, and applies different weight parameters to each intermediate line.

[0069] For example, the section of lines 1 to 4 is By assigning weights at a ratio such as , the signal can be changed linearly, and quadratic interpolation can be applied in sections 4 to 7 to ensure curvilinear continuity of the signal.

[0070] In this way, by measuring odd-numbered lines and configuring the space between them as virtual lines, or by creating multiple virtual lines while maintaining the measured lines at regular intervals, the system of the present invention can reduce the total number of scanning lines while maintaining the detail of the image. Furthermore, even in a real-time data stream environment, the processor (120) can immediately interpolate and generate virtual lines using adjacent line information immediately after acquiring a scanning line, thereby shortening the scanning waiting time of subsequent lines and dramatically improving the overall inspection speed.

[0071] Therefore, the present embodiment provides a technical basis for maximizing scanning efficiency without loss of image quality even in high-speed inspection environments, and can be effectively applied to industrial fields where defect detection is important, such as semiconductor wafers, metal welds, and composite laminated structures.

[0072] In one embodiment, the processor (120) may be configured to set the number or density of virtual lines differently depending on the importance of each region of the sample, the possibility of defect occurrence, or the presence or absence of a pattern.

[0073] As illustrated in FIGS. 2 to 6, the processor (120) scans the interior or surface of a sample in units of multiple scanning lines to acquire data for each line, and then adaptively adjusts the generation density of virtual lines according to the characteristics of each region. Here, 'regional importance' refers to the degree to which relatively detailed inspection is required based on the physical structure, functional role, defect history, or inspection criteria of the sample. For example, in the case of a semiconductor wafer, the pattern forming area, wiring layer, and junction area are classified as regions with high importance, while the outer edge of the substrate or the homogeneous area has relatively low importance.

[0074] FIGS. 2 and FIGS. 3 conceptually illustrate the difference between a conventional sequential scanning method and a virtual line-based inspection method according to the present invention. In the conventional method, the inspection time is prolonged because all lines must be scanned at equal intervals, but in this embodiment, the overall inspection efficiency can be improved by densely arranging scanning lines and virtual lines in high-importance areas and intermittently setting virtual lines in low-importance areas.

[0075] The processor (120) has a defect risk corresponding to each location of the sample, as shown in FIG. 4. The density function of virtual lines can be determined based on this value. For example, in areas with a high risk of defects, the scanning interval can be narrowed or the weight set higher to enhance the interpolation precision of virtual lines, while in homogeneous areas, the virtual line interval can be widened to increase processing speed. This relationship can be expressed by the following mathematical equation 5.

[0076] [Mathematical Formula 5]

[0077]

[0078] Here, is the virtual line density at the corresponding coordinates, is the reference line density, is a risk-reflecting coefficient, is the defect risk of the relevant area, is the average risk of the entire area, is the maximum risk among the whole.

[0079] This equation is a nonlinear adjustment model that increases the frequency of virtual line placement as the probability of defects increases, and can precisely control the spatial distribution of image synthesis according to the non-uniform characteristics of the sample.

[0080] Additionally, the processor (120) can reflect the presence or absence of a pattern or texture information within the sample, set a high virtual line density in the area where the pattern exists, and apply a low density to the area without a pattern. For example, in a circuit section where a fine pattern exists, the change in local reflection intensity is rapid, so the signal difference between adjacent lines a reference value If it exceeds, you can set it to insert additional virtual lines in the corresponding section. This can be expressed mathematically as follows.

[0081] [Mathematical Formula 6]

[0082]

[0083] Here, is the number of virtual lines generated in the corresponding section, is the standard virtual line count, is the pattern sensitivity coefficient, is the difference in reflection intensity between adjacent lines.

[0084] Through such settings, the processor (120) can perform precise inspection by increasing the density of virtual lines in edges, joints, or defect-dominant areas within the sample, and reduce the number of virtual lines in homogeneous internal areas to shorten the inspection time. For example, only one virtual line can be interpolated in the central part of the sample (homogeneous area), and 2 to 3 virtual lines can be added within the same interval in the peripheral part (defect-sensitive area) to perform multilevel interpolation.

[0085] Additionally, the processor (120) can reset the virtual line density in real time by applying a composite weighting model that includes external variables such as the shape curvature of the sample, the rate of change of reflection intensity, and temperature correction data. For example, in the case of a curved sample, the intensity of the reflection signal decreases according to the surface slope θ, so the following correction weighting function can be applied.

[0086] [Mathematical Formula 7]

[0087]

[0088] Here is a stabilization constant, and is the inclination angle of the surface. This correction model enables uniform image synthesis even in curved samples or non-planar structures, and can minimize image distortion during defect detection.

[0089] Accordingly, the processor (120) according to the present embodiment can secure optimal image quality without acquiring unnecessary data by automatically adjusting the number and density of virtual lines by comprehensively considering various physical factors such as the importance of each region of the sample, the possibility of defect occurrence, pattern distribution, or curvature change. That is, in edge regions where the probability of defect occurrence is high, the density of virtual line generation is increased to enhance the detail of the image, and in planar regions where the probability of defect occurrence is low, virtual line generation is minimized, resulting in improved overall inspection speed and efficiency.

[0090] That is, according to the present embodiment, through features such as virtual line adaptive placement based on the structural and functional importance of the sample, (ii) real-time line density control based on defect risk, (iii) improvement of local interpolation precision reflecting the presence or absence of a pattern, and (iv) application of correction weights for curved and non-homogeneous samples, it provides a technical effect that simultaneously maximizes image quality and inspection efficiency compared to the existing uniform scanning method.

[0091] In one embodiment, the processor (120) may be configured to minimize the number of virtual lines in areas with a high risk of defects in the sample, and to configure multiple virtual lines in non-critical areas with a low risk of defects.

[0092] The above processor (120) is linked to the configuration shown in FIGS. 2 to 6, scans the surface or interior of a sample in units of multiple scanning lines, evaluates the risk level of defects in each area, and controls the number and arrangement of virtual lines accordingly.

[0093] The defect risk can be calculated by comprehensively considering the structural location of the sample, the non-uniformity of the reflected signal, the complexity of the surface pattern, or the frequency of defect occurrence in past inspection data.

[0094] For example, areas where defects frequently occur, such as the edges or junctions of a sample, are classified as high-risk areas, while internal homogeneous areas or flat areas without patterns are classified as low-risk areas.

[0095] FIGS. 2 and FIGS. 3 illustrate the conceptual difference in this risk classification. In conventional inspection methods, the inspection time is prolonged because all areas must be scanned at the same density, whereas in the embodiment according to the present invention, the overall inspection efficiency can be improved by setting the number of virtual lines differently according to the risk level. Specifically, the processor (120) controls the system to suppress signal distortion by interpolating only a minimum number of virtual lines based on actual scanning line data in areas with a high risk of defects, and conversely, to increase the inspection speed by continuously generating multiple virtual lines in areas with a low risk of defects.

[0096] For example, the processor (120) has a defect risk value for each area of ​​the sample Calculate the number of virtual lines in the corresponding area accordingly. can be determined by the following mathematical formula 8.

[0097] [Mathematical Formula 8]

[0098] )

[0099] Here, is the number of virtual lines generated at that location, is the maximum number of virtual lines that can be generated within a given interval, is the defect risk of the relevant area, is the highest risk among all areas, is a small constant to prevent the denominator from converging to zero.

[0100] According to this formula, The higher the As the number decreases, the number of virtual lines is reduced, and a relatively large number of virtual lines are generated as the risk level decreases. That is, the processor (120) adjusts the number of virtual lines inversely proportional to the defect risk level, thereby maintaining actual signal-based precision inspection in high-risk areas and maximizing processing efficiency through virtual synthesis in low-risk areas.

[0101] In addition, the processor (120) can dynamically adjust the weights by considering factors such as time stability, local contrast, and phase deviation of the reflected signal, in addition to the simple risk value. To this end, the processor has a composite weight index for each region Calculate and the final number of virtual lines can be calculated according to the following mathematical formula 9.

[0102] [Mathematical Formula 9]

[0103]

[0104] Here, is the final number of virtual lines with weight correction applied, is the weight influence factor, is a weighted index representing the possibility of signal distortion or noise density.

[0105] At this time It can be derived from various measured characteristics such as the sample's surface reflectance, ultrasonic attenuation, signal-to-noise ratio (SNR), and pattern periodicity. For example, in regions where surface reflectance changes rapidly or noise is high As the value increases, the processor (120) suppresses the generation of virtual lines in the corresponding area to prevent degradation of image quality. Conversely, in a homogeneous area with low noise and stable signal, Since the value is low, multiple virtual lines can be generated to improve inspection speed.

[0106] Additionally, the processor (120) can dynamically readjust virtual line generation rules while updating the defect risk distribution by region in real time. For example, if signs of defects are detected from some scanning data acquired during the initial inspection, the risk level of that section can be increased to limit virtual line generation and increase the actual scanning frequency. Conversely, if defect signals are not continuously detected in a certain area, the risk level of that area can be lowered to allow more virtual lines during subsequent scanning, thereby maximizing inspection efficiency.

[0107] With such a configuration, the processor (120) in this embodiment performs high-precision inspection centered on actual scanning in areas with a high probability of defect occurrence, and in non-critical areas with a low probability of defect occurrence, multiple virtual lines are configured to reduce data throughput while maintaining image continuity. Ultimately, the present invention resolves the trade-off between inspection speed and image precision, and provides the technical effect of maximizing inspection efficiency while maintaining the reliability of defect detection.

[0108] In one embodiment, the processor (120) may be configured to correct the virtual line image by setting different weights for each virtual line according to the defect sensitivity of the sample, giving a high weight to areas with a high probability of void or crack occurrence and giving a low weight to homogeneous areas.

[0109] The processor (120) scans the surface or interior of a sample in units of multiple scanning lines, as shown in the series of processes illustrated in FIGS. 2 to 6, and then analyzes the signal change rate, reflection intensity, phase difference, etc. between adjacent lines to interpolate and generate a virtual line.

[0110] In this embodiment, instead of simply assigning equal weights based on the distance between adjacent lines or signal strength, the weights for each virtual line are set differently based on the defect sensitivity of the sample, thereby simultaneously improving the accuracy of defect detection and the effective resolution of the image.

[0111] Defect sensitivity is a value that quantifies the degree to which the probability of defect occurrence is relatively high based on the structural characteristics or inspection history of a sample. For example, areas with narrow wiring pattern spacing on semiconductor wafers, regions of non-uniform penetration in metal welds, and bonding interfaces of composite laminates exhibit high sensitivity.

[0112] The processor (120) has these defect sensitivity indices After calculating for each region, the weight for each virtual line can be determined by the following mathematical formula 10.

[0113] [Mathematical Formula 10]

[0114]

[0115] Here, is the weight assigned to the virtual line at the corresponding coordinates, is the reference weight value, is the sensitivity scaling factor, is the defect sensitivity value, is the maximum sensitivity among the entire range, is a small positive constant to prevent convergence to 0.

[0116] This formula is set so that the weight increases proportionally as sensitivity increases, so in areas where the probability of defects is high, the reflection ratio of actual scanning data increases, and the weight of interpolation by virtual lines is relatively reduced.

[0117] On the other hand, in homogeneous regions with low defect sensitivity, the weight of virtual lines is set low, and the focus is placed on ensuring image continuity through multi-line interpolation.

[0118] For example, in the illustrated examples of FIGS. 3 and 4, the processor (120) generates virtual lines corresponding to lines 2 and 4 based on actual data obtained from scanning lines 1, 3, and 5, but in areas of joints or pattern edges with high defect sensitivity, the weight of the scanning line data (actual signal) is set to 0.8 or higher, and the interpolation weight of the virtual line is maintained at 0.2 or lower. On the other hand, in homogeneous internal planar areas, the weight ratio is adjusted to 0.5:0.5 or 0.4:0.6 to more actively reflect the interpolation result of the virtual line.

[0119] In addition, the processor (120) has, in addition to defect sensitivity, the signal-to-noise ratio (SNR) of the image and the reflection phase difference (Δ The weights can be adjusted multidimensionally by considering factors such as local contrast, or the rate of change in surface curvature (∂²z / ∂²) as auxiliary factors. For example, since there is a high possibility of signal distortion in regions where the surface of the sample has a curved shape or the reflection phase is unstable, the weights of the virtual lines are attenuated to suppress the propagation of noise. This correction logic simultaneously improves the stability of image quality and the reliability of defect detection.

[0120] The processor (120) may adaptively respond to changes in the quality of scanning line data by updating the weight distribution according to defect sensitivity in real time. For example, if a void signal (internal void) or a crack signal (crack waveform) in a specific area is detected during the initial inspection phase, the defect sensitivity of that area is automatically increased, and the weight of the virtual line corresponding to it is adjusted upward to increase the weight of the actual signal in additional synthesis operations. Conversely, in a stable section where no defect signal is detected, the weight is relaxed to ensure interpolation efficiency.

[0121] Weighting based on defect sensitivity can be applied to the following generalized mathematical formula 11 during image synthesis.

[0122] [Mathematical Formula 11]

[0123]

[0124] Here, is the final synthesized pixel intensity, is actual scanning line data, is virtual line interpolation data.

[0125] According to this formula, the higher the defect sensitivity of the region, the The value increases The influence of increases, and in the homogeneous region The influence of increases.

[0126] Such defect sensitivity-based weighting not only improves image synthesis quality but also enhances the contrast of defect signals, enabling precise differentiation of the locations of minute voids, cracks, delamination, and the like.

[0127] Accordingly, the processor (120) according to the present embodiment (i) quantifies defect sensitivity and dynamically sets line-specific weights, (ii) limits interpolation based on actual data in areas with a high probability of defects, (iii) improves overall processing speed by increasing interpolation efficiency in stable areas, and (iv) simultaneously secures inspection accuracy and efficiency by performing correction functions for noise and phase distortion.

[0128] Ultimately, the present invention provides a high-precision ultrasonic inspection system capable of optimizing inspection speed while maintaining image reliability through defect sensitivity-based weight control.

[0129] FIG. 7 is a flowchart illustrating an ultrasound examination method based on virtual line-based image synthesis according to an embodiment of the present invention.

[0130] Referring to FIG. 7, an ultrasound examination method by virtual line-based image synthesis according to one embodiment of the present invention may include the following steps S1 to S4. However, it is not limited thereto, and other general steps may be further included in the ultrasound examination method by virtual line-based image synthesis.

[0131] The above-described ultrasonic examination method based on virtual line-based image synthesis may consist of steps processed sequentially in a system (100) that performs the above-described ultrasonic examination method based on virtual line-based image synthesis. Therefore, even if details are omitted below, the description above regarding the system (100) that performs the ultrasonic examination method based on virtual line-based image synthesis may be equally applicable to the ultrasonic examination method based on virtual line-based image synthesis described later.

[0132] In the above S1 step, the processor (120) can acquire image data corresponding to each scanning line by controlling a probe to scan the interior or surface of the sample in units of multiple scanning lines.

[0133] In the above S2 step, the processor (120) can generate an image of a virtual line corresponding between them by interpolation or weighted average operation using image data obtained from two or more adjacent scanning lines among a plurality of scanning lines.

[0134] In the above S3 step, the processor (120) can synthesize the scanning line image and the virtual line image to form an overall inspection image of the sample.

[0135] In the above S4 step, the processor (120) can determine whether there are defects or abnormalities in the sample based on the entire inspection image.

[0136] FIGS. 8 and FIGS. 9 are drawings illustrating an example of a multi-probe based ultrasound examination method performed by the system (100) of FIG. 1.

[0137] Referring to FIGS. 8 and 9, the aforementioned system (100) can also be applied to a multi-probe based ultrasound examination method. Additionally, the system (100) according to the present embodiment can also be applied to the system (100) according to FIGS. 1 to 7 described above.

[0138] Specifically, the processor (120) can transmit ultrasonic signals to the surface of a sample through a plurality of probes and receive ultrasonic echo signals reflected from the sample to acquire ultrasonic data for each probe. Here, the plurality of probes may be arranged at a constant pitch interval in the X-axis and Y-axis directions, for example as shown in FIG. 8, and each probe can significantly improve the inspection speed by independently scanning different areas of the sample simultaneously. In this multi-probe scanning structure, the irradiation areas between adjacent probes overlap, so an overlapping area corresponding to the boundary surface between probes is formed. In this overlapping area, since multiple ultrasonic reflection signals are detected from the same sample location, image distortion or brightness non-uniformity may occur due to differences in sensitivity or signal phase between probes. Therefore, the processor (120) distinguishes and records the echo signals received from each probe and manages the ultrasonic data for each probe individually, thereby enabling subsequent alignment and correction processes.

[0139] Additionally, the processor (120) can generate probe-specific ultrasound images based on ultrasound data acquired from multiple probes. The probe-specific images generated here may have different brightness, contrast, resolution, and gain characteristics even when capturing the same sample, depending on hardware factors such as the frequency characteristics, sensitivity, and focal length of the probes. Accordingly, the processor (120) uses digital ultrasound data provided by the signal data processing unit to form independent two-dimensional image frames for each probe through the image generation unit, and aligns and stores them in a form that can be integrated in a subsequent step.

[0140] Additionally, the processor (120) can analyze image data of overlapping regions existing between ultrasound images of each probe and differences in image quality characteristics of images of each probe. Specifically, the processor (120) detects overlapping regions of images generated from a plurality of probes as illustrated in FIG. 8, and calculates relative deviations between probes by comparing signal strength, average luminance value, contrast ratio, etc., for each overlapping region. In this process, the overlapping regions are not simply distinguished based on boundary lines, but weights may be assigned in pixel units within a certain range according to the data correlation between probes. These analysis results are input into a subsequent image quality correction algorithm and utilized for the uniformization of images of each probe.

[0141] Additionally, the processor (120) can uniformly correct the quality of ultrasound images between probes by adjusting at least one quality control factor among brightness, contrast, resolution, and gain of the ultrasound image for each probe based on the analyzed results. Referring to FIG. 9, the image control device equipped in the processor (100) displays parameter values ​​corresponding to the image for each probe on a level from 0 to 100, and the processor (120) controls these values ​​to ensure overall uniformity of the image. For example, different parameter levels may exist, such as when the brightness of probe #1 is 100, the contrast is 98, the resolution is 95, and the gain is 96. The processor (120) automatically corrects all probe images to have the same level of contrast and resolution by comparing the quality control factor for each probe and correcting it to match the average or reference probe value. In this case, the processor can perform weight-based adaptive calibration rather than simple averaging, and calculates optimal calibration parameters using characteristic values ​​such as inter-probe reliability, signal-to-noise ratio (SNR), and reflection intensity.

[0142] Additionally, the processor (120) may be configured to fuse the corrected ultrasound images for each probe and the image data of the overlapping region to generate a single integrated inspection image, and to output the generated inspection image to a display device. In the fusion process, to minimize visual seam lines occurring at the boundary lines of each probe image, the processor (120) interpolates pixel values ​​between adjacent images or applies a blending algorithm so that the boundaries between images are naturally connected. Through this, a continuous and uniform high-resolution image can be generated, as if it were taken with a single probe, even though the images were obtained from multiple probes.

[0143] FIG. 8 is a diagram illustrating the conceptual configuration of a multi-probe image processing method according to the present invention, and is intended to explain the overlapping area between probes and the image adjustment process. As shown in the figure, as a plurality of probes scan the surface of a sample, there are overlapping sections of images acquired by each probe, and these overlapping areas are a major cause of image quality degradation. A processor (120) according to the present invention automatically recognizes these overlapping areas and generates a clean image by equalizing brightness and contrast through image adjustment parameters.

[0144] In addition, Figure 9 is a diagram showing an example of the operation of an image control device for each probe, showing that while the brightness, contrast, resolution, and gain values ​​of each probe image differ from one another in the original image stage on the left, all parameters are normalized to 100 by the control of the processor (120) in the optimized image stage on the right, thereby generating a uniform image.

[0145] In this way, the processor (120) according to the present invention can precisely process overlapping areas using data acquired from a plurality of probes and correct image quality adjustment factors in real time, thereby improving inspection speed compared to the existing single-probe method while preventing degradation of image quality. Accordingly, the ultrasonic inspection system of the present invention overcomes the physical limitations of multiple probes and enables high-precision defect detection and stable image quality.

[0146] The multi-probe based ultrasonic inspection system (100) according to the present invention has the effect of maintaining uniform brightness and resolution of the entire image by simultaneously transmitting ultrasonic signals from a plurality of probes and receiving reflected signals to generate images for each probe, and then precisely correcting overlapping areas and non-uniformity in image quality occurring at the boundaries between probes. In particular, since the processor (120) automatically analyzes parameters such as brightness, brightness, resolution, and gain of each probe image and adaptively adjusts them, image quality degradation caused by sensitivity differences or interference between probes can be minimized. Accordingly, a high-resolution integrated image can be obtained across the entire area of ​​the sample, the reliability of the inspection results is improved, and the inspection speed is dramatically improved compared to a single probe method.

[0147] In one embodiment, the processor (120) may be configured to correct the quality difference between probes by controlling at least one quality control factor among brightness, contrast, resolution, and gain of the ultrasound image for each probe according to the parameter level calculated from the ultrasound image for each probe.

[0148] More specifically, the processor (120) digitizes ultrasound data acquired from each of the multiple probes and then forms individual image frames through the probe-specific image generation unit. Each probe-specific image may have different image quality characteristics due to the reflection intensity of the ultrasound signal, the curvature of the sample surface, the acoustic impedance difference, etc. For example, even when photographing the same sample, the brightness or contrast of the image may vary depending on the probe's center axis, irradiation angle, sensitivity deviation, etc., and this may reduce the uniformity of the entire composite image.

[0149] Accordingly, the processor (120) extracts a parameter level corresponding to an image quality control factor from an image for each probe. Here, the parameter level can be calculated from the average luminance value, histogram distribution, spatial frequency response, signal-to-noise ratio (SNR), etc. of the image, and each element is combined with different weights.

[0150] For example, the processor, for probe i, uses the image quality adjustment factor according to the following mathematical formula 12. It can produce.

[0151] [Mathematical Formula 12]

[0152]

[0153] Here is brightness, is contrast, is resolution, corresponds to gain, and represents the weight factor for each factor.

[0154] These weights can be set differently depending on the inspection purpose and sample characteristics, for example, when high-resolution detection is required. When assigning a relatively large value, or when signal contrast is important You can set it high.

[0155] For each probe calculated in this way The value is compared with the reference probe or the overall average value, and the correction coefficient ...is determined. The correction factor can be calculated according to Equation 13.

[0156] [Mathematical Formula 13]

[0157]

[0158] Here is the image quality factor value of the reference probe, and is a coefficient representing the correction strength, where 0 < It can be dynamically adjusted within the range ≤ 1. At this time, the processor (120) calculates Using the value, at least one of the brightness, contrast, resolution, or gain of the image per probe is corrected. For example, if the average luminance value of probe #1 is lower than the reference, the processor can automatically increase the gain of that probe to balance the luminance of the entire image.

[0159] Additionally, the processor (120) can perform statistical-based adaptive updates when fluctuations in parameter levels occur due to environmental changes or sample characteristics. For example, if a change in the amount of ultrasonic attenuation occurs due to temperature changes or differences in the density of the acoustic medium during inspection, the processor can perform real-time updated The parameters are readjusted by recalculating the values. This allows for the maintenance of consistent image quality even under disturbance conditions.

[0160] Referring to FIG. 9, the processor (120) analyzes the image parameter levels for each probe input from the image control device and performs an adjustment process to equalize them. In the original image on the left, the brightness, contrast, resolution, and gain values ​​differ for each probe, but in the optimized image on the right, where the processor's correction is performed, the levels of all factors are normalized to 100, thereby ensuring uniform image quality.

[0161] As such, the processor (120) according to the present embodiment can automatically analyze the parameter levels of images acquired from a plurality of probes and apply a weight-based correction algorithm to each factor, thereby minimizing image quality deviation between probes and maximizing image uniformity. In addition, since these correction results are used as reference values ​​for the subsequent steps of overlapping region fusion and single image generation, the stability of image quality for the entire system can be ensured.

[0162] In one embodiment, the processor (120) may be configured to identify overlapping image data in boundary regions between probes, calculate a reliability weight for each ultrasound image based on the difference in image quality adjustment factors within the overlapping regions, and then fuse the corrected image data to which the weight has been applied using an interpolation or seamless blending method.

[0163] Specifically, when multiple ultrasonic probes scan the surface of a sample, the irradiation areas between adjacent probes overlap at a certain ratio, resulting in different probes acquiring overlapping ultrasonic data for the same sample area. Since the signal strength or image brightness of this overlapping data varies depending on the irradiation angle, reflection angle, sensitivity, and phase error of the ultrasound, discontinuity in image brightness or seam line phenomena may occur in boundary regions. Therefore, the processor (120) refers to the coordinate grid and scan path information between probes within the entire image frame to automatically identify the overlapping area and extract pixel-unit data of the area.

[0164] The processor (120) analyzes the difference in image quality control factors—e.g., brightness, contrast, resolution, and gain—for each probe in the identified overlapping area and calculates a reliability weight for each probe based on this difference. The reliability weight is a quantitative indicator of how accurately each probe image reflects the actual sample characteristics and can be defined by the following mathematical formula 14.

[0165] [Mathematical Formula 14]

[0166]

[0167] Here is the confidence weight of probe i, and is the image quality control factor value of the corresponding probe (expressed as a combination of at least one of brightness, contrast, resolution, and gain), is the same parameter value as the reference probe, is a minimum constant to prevent the denominator from converging to zero. In other words, the smaller the difference from the reference value, the higher the reliability of the probe is evaluated, and conversely, if the difference is large, a low weight is assigned.

[0168] Subsequently, the processor (120) determines each pixel value within the overlapping area The following mathematical formula 15 is used to calculate weight-based corrected image data.

[0169] [Mathematical Formula 15]

[0170]

[0171] Here is the pixel value corresponding to the coordinates (x, y) obtained from probe i, and N represents the number of probes involved in the overlapping area.

[0172] Equation 15 is in the form of a weighted average, and minimizes image quality imbalance between probes by correcting pixel values ​​in proportion to the reliability of each probe.

[0173] The correction image data of the overlapping region obtained through such operations is processed to be naturally continuous with the boundary portion of the adjacent probe image, and the processor (120) rearranges this data into the entire image frame to form a single integrated image. During the fusion process, the processor can selectively apply two modes.

[0174] First, in the interpolation method, pixel values ​​within the boundary region between probes are linearly or quadratically interpolated according to the spatial distance (distance ratio) or intensity difference to form a smooth transition region. For example, the center of the overlapping region is adjusted to be close to the average value of both probes, and the boundary region is adjusted to reflect the intensity of the adjacent probe more, thereby eliminating discontinuities at the boundary.

[0175] Second, in the seamless blending method, a gradient function is generated based on the signal intensity distribution of overlapping regions, and this is used to naturally connect the hue and brightness of the image between probes. This seamless blending effectively eliminates visual boundaries at image transitions and ensures uniform visual quality, just as if the image were acquired with a single probe.

[0176] The processor (120) can also dynamically adjust the influence of the weights according to the shape or size of the overlapping region. For example, data corresponding to the central axis of the probe can be given a larger weight because it generally has high sensitivity and SNR, while data corresponding to the irradiation boundary of the probe can be given a relatively smaller weight because it has high signal distortion. In this case, the spatial distribution of the weights is a distance function within the overlapping region. It can be corrected according to this, thereby improving the uniformity of the entire image while maintaining spatial consistency.

[0177] As such, the processor (120) according to the present embodiment can simultaneously secure natural continuity and visual unity of the image by solving the problem of image imbalance between probes occurring in overlapping areas through reliability-based weighting and interpolation / blending algorithms. As a result, the image quality of the entire ultrasound image is maintained uniformly, the accuracy of identifying boundary defects is improved, and image distortion or seam lines caused by multi-probe driving are effectively eliminated.

[0178] In one embodiment, the processor (120) may be configured to set a reference probe and detect differences in brightness, contrast, and resolution of the non-reference probe image to perform adaptive calibration to minimize the error with the reference probe image.

[0179] Specifically, when multiple probes scan the surface of a sample simultaneously, each probe may generate images with different brightness, contrast, and resolution even when measuring the same sample area due to minute manufacturing variations, imbalances in driving voltage, reduced sensitivity, and differences in ultrasonic coupling states. Such inconsistencies in image characteristics between probes not only reduce the uniformity of the overall image quality but also cause seam lines at boundaries or defect detection errors.

[0180] Accordingly, the processor (120) can set one reference probe among a plurality of probes by comprehensively considering the signal-to-noise ratio (SNR), center frequency response, reflection intensity, reference calibration value, etc. This reference probe may be a standard probe that exhibits the most stable sensitivity in the entire system or has completed precision calibration.

[0181] The processor (120) sets the reference image data obtained from the reference probe as a reference frame and analyzes the differences in luminance, contrast, and resolution by comparing it with the image data obtained from the non-reference probe. For example, the average luminance value of each probe image , contrast ratio , spatial frequency response Calculate and the corresponding value of the reference probe An error function can be defined using the difference from.

[0182] At this time, the processor (120) for the image of non-reference probe i has an image quality error (Error Function) according to the following mathematical formula 16. Calculate.

[0183] [Mathematical Formula 16]

[0184]

[0185] Here These are coefficients weighted according to the importance of luminance, contrast, and resolution, respectively, and can be dynamically changed depending on the inspection purpose. For example, if defect contour detection is important, the weight for contrast Set it high, and if fine structure detection is required, weight for resolution It can relatively increase.

[0186] Such an error function indicates how far each non-reference probe image deviates from the reference image, and the processor (120) calculates a correction parameter to minimize this. The processor (120) has a correction coefficient regarding brightness, contrast, and resolution control. Calculate the difference from the reference image according to the following mathematical formula 17.

[0187] [Mathematical Formula 17]

[0188]

[0189] Here is a constant representing the corrected learning rate (gain factor), and is the gradient of the error function, representing the direction of adjustment in the current frame. The processor (120) minimizes the difference from the reference image by gradually updating at least one factor among the luminance, brightness, and resolution of each probe image using the gradient of this error function.

[0190] This process is not performed only on a single frame, but can be performed in the form of an adaptive loop that reflects temporal variation between frames. That is, the processor (120) performs real-time calibration by dynamically updating the correction coefficient of the current frame by referring to the correction result of the previous frame. For example, even if the reflectance of the sample surface changes over time or the ultrasonic attenuation characteristics change due to temperature changes, the processor (120) Continuously calculating the value to create a new It automatically maintains uniform image quality by reflecting the value.

[0191] Additionally, the processor (120) may consider the correlation between each image quality factor when performing correction. For example, brightness and gain have a proportional relationship, so and Since the differences between them are adjusted simultaneously and resolution and contrast can have an inverse correlation, constraints can be set to prevent excessive adjustment of one factor from affecting the other. These mutual correction constraints contribute to maintaining the stability and consistency of image quality.

[0192] Meanwhile, the processor (120) may set an intermediate reference when the reliability of the reference probe is temporarily reduced during adaptive calibration. For example, when the ultrasonic output of the reference probe is reduced or attenuation occurs in a specific frequency band, the processor (120) selects the probe with the best signal quality among adjacent probes and uses it as an intermediate reference to maintain the calibration continuity of the entire system.

[0193] After undergoing this adaptive calibration procedure, all probe images are normalized to the same luminance distribution and resolution level; consequently, when images acquired from multiple probes are integrated into a single unified image, discontinuities at the boundaries or contrast imbalances are substantially eliminated.

[0194] The processor (120) according to the present embodiment analyzes the image characteristics of a non-reference probe relative to a reference probe in real time and performs an automatic adaptive correction function by defining and minimizing an error function based on the difference in image quality factors. Through this, the difference in sensitivity and image quality between probes can be quantitatively controlled, and the uniformity of the entire image and inspection reliability can be secured simultaneously.

[0195] In one embodiment, the processor (120) may be configured to automatically update a reference value of an image quality adjustment factor based on a statistical characteristic value of an overlapping area or the entire image, and to adaptively update an image correction parameter in response to environmental changes per probe.

[0196] Specifically, image quality control factors (brightness, contrast, resolution, gain) for correcting image deviation between probes in a multi-probe system are controlled around a constant reference value, and this reference value needs to be dynamically adjusted as the inspection environment or sample conditions change. For example, if the surface reflectance of the sample changes or the temperature, viscosity, or density of the ultrasonic transmission medium changes, the intensity and phase of the reflected signal received from each probe change, and accordingly, the average brightness and contrast distribution of the entire image may change. In such cases, since using a fixed reference value reduces the accuracy of image quality correction, the processor (120) is configured to automatically update the reference value by analyzing statistical characteristic values ​​in real time.

[0197] The processor (120) periodically collects image data generated for each probe and calculates statistical indicators in the entire image or overlapping areas. Here, the statistical indicators are, for example, mean luminance, ), luminance variance (variance, ), average contrast ratio (mean contrast, These can include resolution-related spatial frequency response (SFR). These metrics are accumulated at regular intervals or on a frame-by-frame basis and used to update reference values ​​for image quality control factors.

[0198] At this time, the processor (120) may assign weights to each statistical value by considering the reliability of the overlapping area. For example, since the overlapping area is a section observed simultaneously by multiple probes, it has high reliability in determining the stability of the image quality. Therefore, the processor (120) may reflect the statistical value of the overlapping area as higher than the overall image statistical value, and the weight can be defined by the following mathematical formula 18.

[0199] [Mathematical Formula 18]

[0200]

[0201] Here is an integrated statistic for calculating the standard value of image quality factors, is the statistical value of the entire image (global image), is the statistical value of the overlap region, and These are the confidence weights for the entire image and the overlapping area, respectively.

[0202] For example, if the reliability of the overlapping area is determined to be higher By setting it to this, the reference value can be configured to respond more sensitively to the actual environment.

[0203] The processor (120) has the integrated statistical value calculated in this way Using this, the reference values ​​of each image quality control factor (brightness, contrast, resolution, gain) are updated. The updating of reference values ​​can be performed reliably using statistical adaptive filters such as the cumulative average method or the Exponential Moving Average (EMA). For example, the processor the reference value can be updated for time t as shown in the following mathematical equation 19.

[0204] [Mathematical Formula 19]

[0205]

[0206] Here Is It is an adaptive learning rate, and a larger value can be applied to increase responsiveness when environmental changes are rapid, and a smaller value can be applied in a stable testing environment to mitigate fluctuations in reference values.

[0207] This reference value update process can be performed in real time even when the inspection environment is constantly changing. For example, if the temperature of the probe rises and the sensitivity decreases, the processor (120) detects that the average luminance value in the image of the probe is decreasing relative to the reference, and immediately reflects the new reference value updated by Equation 19 into the overall correction parameter. Conversely, if the reflectance of the sample increases or the density of the ultrasonic medium increases and the signal is strengthened, the average luminance value rises, and the processor (120) automatically adjusts the reference value upward to control the brightness of the entire image so that it does not increase excessively.

[0208] Additionally, the processor (120) can independently update probe-specific correction parameters to individually correct environmental imbalances among multiple probes. For example, if probe #1 has characteristics sensitive to temperature rise while probe #3 is susceptible to electrical noise, the processor (120) detects environmental change factors of each probe and selectively adjusts the gain and brightness factors of the corresponding probe in response. These probe-specific adaptive updates serve to offset individual hardware differences while maintaining the overall uniformity of the system.

[0209] Furthermore, the processor (120) may analyze the temporal trend to correct for long-term reference value fluctuations. That is, it estimates the average deviation of each image quality factor based on statistical data accumulated from multiple inspection sessions, and automatically performs a calibration mode when the deviation exceeds a certain threshold. This function is useful for ensuring consistent image quality despite system aging or changes in sensitivity.

[0210] As such, the processor (120) according to the present embodiment automatically updates reference values ​​by synthesizing statistical characteristic values ​​of the entire image and overlapping areas, and adaptively responds to environmental changes for each probe, thereby maintaining consistency in image quality and inspection reliability even under conditions of long-term continuous inspection or significant environmental changes. Accordingly, the system of the present invention is a high-precision multi-probe ultrasonic inspection device equipped with a real-time automatic correction function, and can be effectively utilized for real-time quality control and defect detection in various industrial environments.

[0211] FIG. 10 is a flowchart illustrating a multi-probe-based ultrasound examination method according to one embodiment of the present invention.

[0212] Referring to FIG. 10, a multi-probe based ultrasound examination method according to one embodiment of the present invention may include the following steps S1 to S5. However, it is not limited thereto, and other general steps may be further included in the multi-probe based ultrasound examination method.

[0213] The above multi-probe based ultrasound examination method may consist of steps processed in a time-series manner in a system (100) that performs the above-described multi-probe based ultrasound examination method. Therefore, even if details are omitted below, the description above regarding the system (100) that performs the multi-probe based ultrasound examination method may be equally applicable to the multi-probe based ultrasound examination method described below.

[0214] In the above S1 step, the processor (120) can transmit an ultrasonic signal through a plurality of probes to the surface of a sample and receive an ultrasonic echo signal reflected from the sample to obtain ultrasonic data for each probe.

[0215] In the above S2 step, the processor (120) can generate ultrasound images for each probe based on ultrasound data obtained from a plurality of probes.

[0216] In the above S3 step, the processor (120) can analyze the image data of the overlapping region existing between the ultrasound images for each probe and the difference in image quality characteristics of the images for each probe.

[0217] In the above S4 step, the processor (120) can uniformly correct the quality of the ultrasound images between probes by adjusting at least one quality control factor among brightness, contrast, resolution, and gain of the ultrasound images for each probe based on the analyzed results.

[0218] In the above S5 step, the processor (120) can fuse the corrected ultrasound images per probe and the image data of the overlapping area to generate a single integrated inspection image and output the generated inspection image to a display device.

[0219] As described above, although the present invention has been explained by limited embodiments and drawings, the present invention is not limited thereto, and it is obvious that various modifications and variations are possible within the scope of the technical spirit of the present invention and the equivalent scope of the claims described below by those skilled in the art to which the present invention belongs.

[0220] Meanwhile, although terms indicating directions such as up, down, left, right, front, and back have been used in this invention, these terms are used merely for convenience of explanation, and it is obvious to those skilled in the art that they may vary depending on the position of the object or the position of the observer. Explanation of the symbols

[0221] 100 : System 110 : Memory 120 : Processor

Claims

Claim 1 In an ultrasonic inspection system for inspecting a sample using multiple ultrasonic probes, a memory configured to store commands; The system includes a processor configured to perform the following operations by executing the above commands, wherein the processor transmits an ultrasonic signal to the surface of a sample through a plurality of probes, receives an ultrasonic echo signal reflected from the sample to acquire ultrasonic data for each probe, generates an ultrasonic image for each probe based on the ultrasonic data acquired from the plurality of probes, analyzes image data of overlapping regions existing between the ultrasonic images for each probe and the differences in image quality characteristics of the images for each probe, and, based on the analyzed results, adjusts at least one image quality control factor among brightness, contrast, resolution, and gain of the ultrasonic image for each probe to uniformly correct the image quality of the ultrasonic images between probes, fuses the corrected ultrasonic images for each probe and image data of the overlapping regions to generate a single integrated inspection image, and is configured to output the generated inspection image to a display device, wherein the processor identifies overlapping image data of boundary regions between probes, calculates a reliability weight for the ultrasonic image for each probe based on the difference in image quality control factors within the overlapping regions, and then fuses the corrected image data to which the weight is applied using an interpolation or seamless blending method, wherein the reliability weight is each probe It is a quantitative indicator representing how accurately an image reflects the actual sample characteristics, and is defined as shown in the following mathematical formula, [Mathematical Formula] (Here is the confidence weight of probe i, and is the image quality control factor value of the corresponding probe (expressed as a combination of at least one of brightness, contrast, resolution, and gain), is the same parameter value as the reference probe, (is an infinitesimal constant to prevent convergence of the denominator to zero.) The processor is configured to adjust the influence of the reliability weight according to the shape or size of the overlapping region, and the spatial distribution of the reliability weight is corrected according to a distance function within the overlapping region, in a multi-probe-based ultrasound inspection system. Claim 2 A multi-probe based ultrasound inspection system according to claim 1, wherein the processor is configured to correct image quality deviation between probes by controlling at least one image quality control factor among brightness, contrast, resolution, and gain of the ultrasound image for each probe according to a parameter level calculated from the ultrasound image for each probe. Claim 3 delete Claim 4 A multi-probe based ultrasound examination system according to claim 1, wherein the processor is configured to set a reference probe and detect differences in brightness, contrast, and resolution of non-reference probe images to perform adaptive calibration to minimize errors with the reference probe image. Claim 5 A multi-probe based ultrasound inspection system according to claim 1, wherein the processor is configured to automatically update reference values ​​of image quality control factors based on statistical characteristic values ​​of overlapping regions or the entire image, and to adaptively update image correction parameters in response to environmental changes per probe. Claim 6 A multi-probe based ultrasonic inspection method performed by an ultrasonic inspection system for inspecting a sample using multiple ultrasonic probes, comprising: a step of transmitting an ultrasonic signal through a plurality of probes to the surface of a sample and receiving an ultrasonic echo signal reflected from the sample to acquire ultrasonic data for each probe; a step of generating an ultrasonic image for each probe based on the ultrasonic data acquired from the plurality of probes; a step of analyzing image data of an overlapping region existing between the ultrasonic images for each probe and the difference in image quality characteristics of the images for each probe; and a step of uniformly correcting the image quality of the ultrasonic images between probes by adjusting at least one image quality control factor among brightness, contrast, resolution, and gain of the ultrasonic images for each probe based on the analyzed results. The method includes the step of fusing the corrected ultrasound images for each probe and the image data of the overlapping region to generate a single integrated inspection image, and outputting the generated inspection image to a display device. The ultrasound inspection system is configured to identify overlapping image data of the boundary region between probes, calculate a reliability weight for each ultrasound image based on the difference in image quality control factors within the overlapping region, and then fuse the corrected image data to which the weight has been applied using an interpolation or seamless blending method. The reliability weight is a quantitative indicator representing how accurately each probe image reflects the actual sample characteristics, and is defined as shown in the following mathematical formula. [Mathematical Formula] (Here is the confidence weight of probe i, and is the image quality control factor value of the corresponding probe (expressed as a combination of at least one of brightness, contrast, resolution, and gain), is the same parameter value as the reference probe, (is an infinitesimal constant to prevent convergence of the denominator to zero.) The above-described ultrasonic inspection system is configured to adjust the influence of the reliability weight according to the shape or size of the overlapping region, and the spatial distribution of the reliability weight is corrected according to a distance function within the overlapping region, in a multi-probe-based ultrasonic inspection method.

Images