Image processing method, processing device and detection device for bonding energy detection

Abstract

The present invention provides an image processing method, processing device, and detection device for bond detection. The method includes: acquiring a target image including a blade and a crack generated by the insertion of the blade; identifying the target image to determine a first edge of the front end of the blade, and acquiring first coordinate data of the location of the first edge; identifying the target image to determine a second curved edge of the area where the crack is located away from one side of the blade, and acquiring second coordinate data of the location of the vertex of the second edge; and calculating the target distance between the first edge and the second edge based on the first coordinate data and the second coordinate data. Based on the technical solution of the present invention, the width of the crack area in the expansion direction can be automatically obtained through image recognition, thereby automatically calculating and evaluating the bonding strength of the wafer bonding structure, realizing automated detection, and solving the problems of low efficiency and poor measurement accuracy in manual detection operations.

Description

Technical Field

[0001] The present invention relates to the technical field of wafer bonding, and in particular to an image processing method, a processing device, and a detection device for bonding energy detection. Background Art

[0002] Bond energy testing is a means of testing the bond strength of bonded wafer pairs. A commonly used test method is the blade insertion method, which involves inserting a blade of a certain thickness into the bonding area in the middle of the wafer pair. An infrared camera is used to see through the crack caused by the blade insertion, and the distance from the blade to the top of the crack arc is calculated. This distance value represents the depth of the crack into the bond structure. Substituting this value into the calculation formula, the bond strength can be calculated.

[0003] Currently, crack measurements are manually taken after taking a photo of the crack. This method relies on a comparison method: based on the known actual dimensions of the blade, the image is measured to determine the proportional relationship between the blade size and the crack length, thereby calculating the actual distance from the crack to the blade. However, this method is inefficient and the accuracy of the measurement results is significantly affected by human factors, which can easily lead to large measurement errors.

[0004] Therefore, the present invention proposes an image analysis and processing method, which automatically calculates the crack length based on automatic analysis and processing of images, thereby avoiding interference of human factors on the measurement results. Summary of the Invention

[0005] In order to solve the problems of low efficiency and poor measurement accuracy in the current process of detecting bonding energy using the insert blade method, which involves manually measuring the detection image, the present application proposes an image processing method, processing device and detection device for bonding energy detection.

[0006] In a first aspect, the present invention provides an image processing method for bonding energy detection, comprising:

[0007] Acquire a target image including a blade and a crack generated by insertion of the blade;

[0008] Identifying the target image to determine a first edge of the front end of the blade, and acquiring first coordinate data of a position of the first edge;

[0009] Identify the target image to determine that the area where the crack is located is away from the curved second edge on one side of the blade, and obtain second coordinate data of the position of the vertex of the second edge;

[0010] A target distance between the first edge and the second edge is calculated according to the first coordinate data and the second coordinate data.

[0011] In one embodiment, recognizing the target image to determine the first edge of the front end of the blade includes:

[0012] performing edge detection on a pattern at a position where the blade is located on the target image to obtain a blade edge image of the blade;

[0013] According to the orientation, a straight line segment formed by the lowest pixel points is extracted from the blade edge image as the first edge, and the remaining portion of the blade edge image is eliminated.

[0014] In one embodiment, obtaining first coordinate data of the location of the first edge includes:

[0015] According to the straight line segment corresponding to the first edge, first coordinate data of two endpoints at both ends of the straight line segment are determined by using Hough transform.

[0016] In one embodiment, recognizing the target image to determine that the crack region is away from the curved second edge of one side of the blade includes:

[0017] Performing edge detection on a pattern in the target image where the crack is located to obtain a crack edge image of the crack area where the crack is located;

[0018] According to the orientation, a curve segment formed by the uppermost pixel points is extracted from the crack edge image as a second edge and the remaining portion of the crack edge image is eliminated.

[0019] In one embodiment, identifying the target image to determine that the crack region is away from the curved second edge of one side of the blade further comprises:

[0020] Determining whether the target image has a semiconductor pattern;

[0021] If there is a semiconductor pattern on the target image, performing two edge detections with different detection parameters on the pattern in the area where the crack is located on the target image to obtain two edge detection result images;

[0022] The two edge detection result images are superimposed, and the overlapping parts of the two patterns are eliminated, and the resulting image is the crack edge image.

[0023] In one embodiment, it further includes:

[0024] A curve segment consisting of the topmost pixel points is extracted from the crack edge image as the front and / or back of the second edge, and small connected domains with an area smaller than a preset threshold in the image are eliminated.

[0025] In one embodiment, obtaining second coordinate data of the position of the vertex of the second edge includes:

[0026] Randomly selecting a preset number of pixel points on the curve segment and determining the coordinates of the pixel points;

[0027] Performing curve fitting according to the pixel points and determining a function expression of the fitted curve according to the coordinates of the pixel points;

[0028] The coordinates of the extreme point on the curve are calculated according to the function expression of the curve, and the coordinates of the extreme point are the second coordinate data of the position of the vertex of the second edge.

[0029] In one embodiment, it further includes:

[0030] According to the function expression of the curve, determining whether the graph of the curve opens upward and has only one minimum extreme value point;

[0031] If so, determining the coordinates of the extreme points on the curve according to the function expression of the curve;

[0032] If not, a preset number of pixel points are randomly selected on the curve segment again to perform curve fitting and determine the function expression.

[0033] In one embodiment, it further includes:

[0034] Repeating the process from selecting pixel points to determining a curve function expression and calculating the coordinates of extreme value points on the curve for a preset number of times to obtain a preset number of extreme value point coordinate results;

[0035] Eliminate the extreme point coordinates with abnormal coordinate data from the preset number of extreme point coordinate results;

[0036] The average value of the remaining extreme point coordinates in the extreme point coordinate results is calculated, and the average value is used as the final coordinate of the extreme point.

[0037] In one embodiment, it further includes:

[0038] A portion of the curve segment as the second edge that is within the x-coordinate range of the first coordinate data is intercepted and the remaining portion is discarded.

[0039] In one embodiment, before identifying the target image to determine the second edge, the method further includes:

[0040] The target image is cropped according to the y-coordinate data of the first coordinate data to crop the image of the blade portion on the target image.

[0041] In one embodiment, it further includes:

[0042] Before identifying the target image, image parameters of the target image are adjusted, where the image parameters include contrast and brightness.

[0043] In a second aspect, the present invention provides an image processing device for bond energy detection, comprising:

[0044] An image acquisition module, configured to acquire a target image including a blade and a crack generated by the insertion of the blade;

[0045] an image recognition module, configured to recognize the target image and obtain an edge image of a target edge position of the blade and the crack on the target image;

[0046] an image processing module, configured to process the edge image to obtain coordinate data of the target edge of the blade and the crack;

[0047] A data calculation module is used to calculate a target distance between a target edge of the blade and a target edge of the crack according to the coordinate data.

[0048] In a third aspect, the present invention provides a bonding energy detection device, comprising the above-mentioned image processing device.

[0049] The above technical features can be combined in various suitable ways or replaced by equivalent technical features, as long as the purpose of the present invention can be achieved.

[0050] The image processing method, processing device, and detection device provided by the present invention for bond energy detection have at least the following beneficial effects compared with the prior art:

[0051] The present invention provides an image processing method, processing device and detection device for bonding energy detection, which can automatically obtain the width of the crack area in the expansion direction through image recognition, thereby automatically calculating and evaluating the bonding strength of the wafer bonding structure, realizing automated detection, and solving the problems of low efficiency and poor measurement accuracy in manual detection operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention will be described in more detail below based on embodiments and with reference to the accompanying drawings, wherein:

[0053] Figure 1 Shows the main flow chart of the method of the present invention;

[0054] Figure 2 A schematic diagram showing a target image without a semiconductor pattern obtained by the method of the present invention;

[0055] Figure 3 Shows Figure 2 The target image shown is a schematic diagram after the contrast and other parameters are adjusted;

[0056] Figure 4 Shows Figure 3 The image shown is a schematic diagram of a blade edge image obtained after edge detection;

[0057] Figure 5 Shows the extraction Figure 4 Schematic diagram showing that the bottommost pixel points of the image form a straight line segment as a first edge;

[0058] Figure 6 A schematic diagram showing a target image with a semiconductor pattern obtained by the method of the present invention after adjustment of parameters such as contrast;

[0059] Figure 7 Shows Figure 6 The image shown is a schematic diagram of a blade edge image obtained after edge detection;

[0060] Figure 8 Shows the extraction Figure 7 Schematic diagram showing that the bottommost pixel points of the image form a straight line segment as a first edge;

[0061] Figure 9 Shows the Figure 5 The coordinates of the straight line segment shown are used to intercept Figure 2 Schematic diagram behind the blade portion at the top of the target image shown;

[0062] Figure 10 Shows Figure 9 The image shown is a schematic diagram after adjustment of parameters such as contrast;

[0063] Figure 11 Shows Figure 10 The image shown is a schematic diagram of a crack edge image obtained after edge detection and small connected domain elimination;

[0064] Figure 12 Shows the extraction Figure 11 The top pixel of the image shown forms a curve segment as the second edge. Figure 5 A schematic diagram of the portion within the x-coordinate range of the straight line segment shown;

[0065] Figure 13 Shows Figure 12 Schematic diagram of the position of the curve segment of the image shown in the target image without the semiconductor pattern;

[0066] Figure 14 A schematic diagram showing a target image with a semiconductor pattern obtained by the method of the present invention and then adjusted again with parameters such as contrast;

[0067] Figure 15 Shows Figure 14 A schematic diagram of an edge detection result image obtained after edge detection with a larger detection parameter on the rightmost image in the shown images;

[0068] Figure 16 Shows Figure 14 A schematic diagram of an edge detection result image obtained after edge detection with smaller detection parameters on the rightmost image in the shown images;

[0069] Figure 17 Shows Figure 15 The image shown is the same as Figure 16 The schematic diagram of the images after superposition and removal of overlapping patterns is shown;

[0070] Figure 18 Shows Figure 17 The image shown is a schematic diagram after removing small connected domains;

[0071] Figure 19 Shows the extraction Figure 17 A schematic diagram showing that the top pixel points of the image form a curved line segment serving as a second edge;

[0072] Figure 20 Shows Figure 19 The image shown is a schematic diagram after removing small connected domains;

[0073] Figure 21 Shows Figure 20 The curve segment of the image shown is Figure 8 A schematic diagram of the portion within the x-coordinate range of the straight line segment shown;

[0074] Figure 22 Shows Figure 21 A schematic diagram of the position of the curve segment of the image shown in the target image with the semiconductor pattern;

[0075] Figure 23 The schematic diagram of the structure of the image processing device of the present invention is shown.

[0076] In the drawings, like reference numerals are used for like parts, but the drawings are not necessarily true to scale. DETAILED DESCRIPTION

[0077] The present invention will be further described below with reference to the accompanying drawings and the following embodiments.

[0078] Example 1

[0079] This embodiment mainly describes the principle and process of image processing when the method of the present invention is applied to the bonding structure of a wafer without a semiconductor pattern.

[0080] An embodiment of the present invention provides an image processing method for bonding energy detection, comprising:

[0081] Step S100 : acquiring a target image including a blade and a crack generated by inserting the blade, wherein the target image corresponds to a bonding structure of a wafer without a semiconductor pattern.

[0082] Specifically, the target image acquired for the bonding structure of the wafer without semiconductor pattern is shown in FIG. Figure 2 When the blade insertion method is used to test the bonding energy, a blade of a certain thickness is inserted between two bonded wafers, and the bonding interface near the blade is peeled off from each other, forming a crack. From the direction perpendicular to the wafer surface, the crack has a fan-like shape extending from the edge to the center of the bonding interface. And due to the entry of air, there is a clear boundary between the area corresponding to the crack and the bonding interface that has not been peeled off, as shown in the figure. Figure 2 It can be clearly seen as shown above.

[0083] Therefore, this embodiment identifies the position of the blade's front edge and the edge of the crack region away from the blade, and calculates the maximum distance between them as an indicator of bond strength. Naturally, the greater the bond strength, the smaller the maximum distance between the blade's front edge and the crack region boundary should be, meaning the crack region extends less toward the center of the bond interface. Conversely, the greater the maximum distance between the blade's front edge and the crack region boundary should be.

[0084] Step S200: adjusting image parameters of a target image, identifying the target image to determine a first edge of the front end of the blade, and obtaining first coordinate data of a position of the first edge.

[0085] Step S210: After adjusting the image parameters of the target image, the target image is recognized, and edge detection is performed on the pattern at the position of the blade on the target image to obtain a blade edge image of the blade.

[0086] Specifically, in order to more clearly identify the blade, before identifying the target image, the contrast and / or brightness of the target image are adjusted (other image parameters such as saturation and sharpness can also be adjusted). Since the pixel value of the blade part is the lowest (darkest), the target image is adjusted until only the blade is basically visible, as shown in the attached figure. Figure 3 As shown in the figure, the blade boundary outline is clearer at this time, and the interference of the crack area pattern on blade recognition is eliminated. Figure 3 The image shown is subjected to canny edge detection, which is performed once using the default detection parameters to obtain the attached figure Figure 4 The blade edge image shown in the figure only retains the pixels of the outer edge of the entire blade.

[0087] Step S220: extracting the straight line segment formed by the lowest pixel points from the blade edge image as the first edge according to the orientation and eliminating the remaining portion of the blade edge image.

[0088] Specifically, the front end direction of the blade can be determined according to the orientation, that is, Figure 4 The bottom of the image is the front end of the blade, that is, the first edge. Therefore, the straight line segment formed by the bottom pixel points of the blade edge image is directly extracted as the first edge of the blade front end. The extraction result is shown in the figure Figure 5 shown.

[0089] Step S230: Determine first coordinate data of two endpoints of the straight line segment corresponding to the first edge by using Hough transform.

[0090] Specifically, the Hough transform is used to determine the coordinates of the two endpoints P1 and P2 at both ends of the straight line segment, namely P1 (x1, y) and P2 (x2, y). The straight line segment of the first edge is horizontal.

[0091] Step S240: cropping the target image according to the y-coordinate data of the first coordinate data to crop the image of the blade portion on the target image.

[0092] Specifically, according to the y coordinate of the first coordinate data of the straight line segment of the first edge determined in step S230, the image of the upper blade portion of the original target image obtained in step S100 is cut off, as shown in FIG. Figure 9 As shown, this can directly avoid the image of the blade part from interfering in the crack recognition process. Figure 9 The lower part of the original target image is also cut off in order to better show the image of the crack part. In actual operation, the lower part has less interference and may not be cut off depending on the situation.

[0093] It should be noted that step S240 is not a necessary step. The purpose of cropping the original target image is to fundamentally reduce the interference of other parts of the image on crack identification. If the original target image has good imaging effect, with clear contrast and distinct boundaries between parts, step S240 can be omitted. Of course, to accommodate most situations, step S240 is preferably selected.

[0094] Step S300: adjusting image parameters of the target image, identifying the target image to determine that the area where the crack is located is away from the curved second edge on one side of the blade, and obtaining second coordinate data of the vertex position of the second edge.

[0095] Step S310: determining whether there is a semiconductor pattern on the target image.

[0096] Step S320: There is no semiconductor pattern on the target image. After adjusting the image parameters of the target image, the target image is identified. Edge detection is performed on the pattern in the crack area on the target image to obtain a crack edge image of the crack area.

[0097] Specifically, the same as step S210, the image parameters are adjusted first, and the adjustment result is shown in the attached figure. Figure 10 As shown in the accompanying drawings Figure 9 In comparison, in the image after image parameter adjustment, the contrast between the crack area and other parts of the bonding interface is stronger, and the boundary is clearer, which is conducive to the accuracy of recognition. Figure 10 The image shown is subjected to Canny edge detection, which is performed once using the default detection parameters. The crack edge image includes the pixel points of the outer edge of the entire crack area, as well as the pixel points of the edge of the bonding interface and the outer edge of the bubble inside it.

[0098] Step S330: Eliminate small connected domains in the crack edge image whose areas are smaller than a preset threshold.

[0099] Specifically, after removing the small connected domains from the crack edge image, it may be possible to directly exclude all other interfering pixels according to the connected domain of the original image, and directly obtain a similar image as shown in the figure. Figure 12 But in most cases, you will get the image of the curve segment of the second edge as shown in the figure. Figure 11 The crack edge image shown still has some interference, which requires further interference elimination.

[0100] Step S340: extracting the curve segment formed by the topmost pixel points from the crack edge image as the second edge according to the orientation and eliminating the rest of the crack edge image.

[0101] Specifically, the side of the crack away from the blade can be determined according to the orientation, i.e. Figure 11 The top of the image is the second edge. To address possible bubbles or interference that cannot be removed (the crack area is above the lifted and peeled part, so there is no interference above, and only the crack area may have interference such as bubbles). Therefore, the curve segment formed by the top pixel points of the crack edge image is directly extracted as the second edge of the crack area.

[0102] Step S350: intercepting a portion of the curve segment serving as the second edge that is within the x-coordinate range of the first coordinate data and discarding the remaining portion.

[0103] Specifically, referring to the range from the endpoints P1 to P2 of the straight line segment, the portion of the curve segment within the range [x1, x2] is intercepted. The interception result is shown in the attached figure. Figure 12 As shown in the attached figure Figure 13 It can also be seen that the relative position of the intercepted curve segment on the original target image, that is, the edge of the crack area corresponds to the part within the range of [x1, x2].

[0104] Step S360: randomly selecting a preset number of pixel points on the curve segment and determining the coordinates of the pixel points.

[0105] Specifically, after obtaining the attached drawings Figure 12 After finding the curve segment shown, it is necessary to determine the function expression of the curve corresponding to the curve segment. This is to accurately determine the coordinates of the vertex of the second edge. A predetermined number of pixels on the curve segment are selected and their coordinates are determined. The predetermined number can be a ratio of the total number of pixels, for example, 1 / 10 of the total number of pixels.

[0106] Step S370: performing curve fitting according to the pixel points and determining a function expression of the fitted curve according to the coordinates of the pixel points.

[0107] Specifically, based on the previous research on the edge shape of the crack area, it was concluded through verification and analysis that the edge shape of the crack area can be approximately described as a quartic function, that is, y = ax 4 +bx 3 +cx 2 +dx+e. Therefore, the coordinates of the acquired pixel points are directly substituted into the curve fitting, and the coefficients of the quartic function are solved using the least squares method to obtain the function expression of the fitted curve.

[0108] The reason why a preset number of pixels are randomly selected for solution is that if all pixels are used for least squares calculation, since the least squares method has a unique optimal solution, it may cause errors and misunderstandings in the presence of interfering pixels, so a certain amount of randomness is added.

[0109] Step S371: According to the function expression of the curve, determine whether the image of the curve opens upward and has only one minimum extreme point.

[0110] Specifically, because the curve segment obtained through image analysis does not necessarily exclude all interfering pixels, the curve segment pattern obtained in step S350 may contain interfering pixels that do not belong to the curve segment. Once an interfering pixel is randomly selected, the fitted curve will be incorrect and will not match the original curve segment. Therefore, the judgment is made based on the function expression of the curve. The shape of the curve that correctly matches the original curve segment should be similar to a parabola opening upward, and the judgment criteria are set as opening upward and having one and only one minimum extreme point.

[0111] Step S372: If yes, determine the coordinates of the extreme points on the curve according to the function expression of the curve.

[0112] Specifically, if the curve fitting result is judged to be correct according to the function expression of the curve, the coordinates of the extreme point on the curve (the minimum extreme point at the vertex of the curve segment) are determined according to the function expression.

[0113] Step S373: If not, randomly select a preset number of pixel points on the curve segment again to perform curve fitting and determine the function expression.

[0114] Specifically, if it is determined that the result of the curve fitting is incorrect according to the function expression of the curve, the process returns to step S370 and performs the determination again.

[0115] Step S380: Calculate the coordinates of the extreme point on the curve according to the function expression of the curve. The coordinates of the extreme point are the second coordinate data of the position of the vertex of the second edge.

[0116] Specifically, if the curve fit is successful, the coordinates of a single extreme point can be directly determined based on the curve's functional expression. This is the coordinates of the minimum extreme point at the lowest vertex of the curve's coordinate image. This point is mapped to the curve segment of the second edge obtained in step S350, namely, the point on this curve segment with the greatest vertical distance from the straight line segment of the first edge. This point is placed in the coordinate system of the straight line segment of the first edge, and its coordinates are P3(x3, y').

[0117] Step S381: Repeat the process from step S360 to step S380 (including step S371 to step S373) for a preset number of times to obtain a preset number of extreme point coordinate results.

[0118] Step S382: removing the extreme point coordinates with abnormal coordinate data from a preset number of extreme point coordinate results.

[0119] Step S383: Calculate the average value of the remaining extreme point coordinates in the extreme point coordinate results, and use the average value as the final coordinate of the extreme point.

[0120] Specifically, the method of averaging multiple results after eliminating abnormal data can further eliminate the interference of abnormal data on the detection results.

[0121] Step S400 : Calculating a target distance between a first edge and a second edge according to the first coordinate data and the second coordinate data.

[0122] Specifically, based on the coordinates of the endpoints of the straight line segment and the coordinates of the extreme points of the curve segment, namely P1(x1, y), P2(x2, y) and P3(x3, y'), the target distance between the first edge and the second edge can be directly calculated as |y'-y| (pixel unit). This target distance reflects the extent of the crack expansion toward the interior of the bonding interface.

[0123] Example 2

[0124] This embodiment mainly describes the principle and process of image processing when the method of the present invention is applied to the bonding structure of a wafer with a semiconductor pattern. Some of the same contents refer to Example 1 and will not be repeated in this embodiment.

[0125] An embodiment of the present invention provides an image processing method for bonding energy detection, comprising:

[0126] Step S100 : acquiring a target image including a blade and a crack generated by inserting the blade, wherein the target image corresponds to a bonding structure of a wafer with a semiconductor pattern.

[0127] Step S200: adjusting image parameters of a target image, identifying the target image to determine a first edge of the front end of the blade, and obtaining first coordinate data of a position of the first edge.

[0128] Step S210: After adjusting the image parameters of the target image, the target image is recognized, and edge detection is performed on the pattern at the position of the blade on the target image to obtain a blade edge image of the blade.

[0129] Specifically, in order to more clearly identify the blade, the image parameters of the target image are adjusted before identifying the target image. Since there is a semiconductor pattern on the wafer, the interference will be more serious. When adjusting the image parameters such as contrast, the corresponding adjustment values ​​need to be set. The purpose is to adjust the target image until only the blade can be seen, as shown in the attached figure. Figure 6 As shown in the figure, the blade boundary contour is clearer at this time, and the interference of the crack area image on the blade recognition is eliminated. Figure 6 The image shown is subjected to canny edge detection, which is performed once using the default detection parameters to obtain the attached figure Figure 7 Image of blade edge shown.

[0130] Step S220: extracting the straight line segment formed by the lowest pixel points from the blade edge image as the first edge according to the orientation and eliminating the remaining portion of the blade edge image.

[0131] Specifically, the front end direction of the blade can be determined according to the orientation, that is, Figure 7The bottom of the image is the front end of the blade, that is, the first edge. Therefore, the straight line segment formed by the bottom pixel points of the blade edge image is directly extracted as the first edge of the blade front end. The extraction result is shown in the figure Figure 8 shown.

[0132] Step S230: Determine first coordinate data of two endpoints of the straight line segment corresponding to the first edge by using Hough transform.

[0133] Specifically, the Hough transform is used to determine the coordinates of the two endpoints P1 and P2 at both ends of the straight line segment, namely P1 (x1, y) and P2 (x2, y). The straight line segment of the first edge is horizontal.

[0134] It should be noted that after step S230 of this embodiment, the image cropping step of step S240 can be performed as in embodiment 1. However, since step S240 is not a necessary step, in order to demonstrate applications in different situations, this embodiment does not perform the image cropping step, but adopts other methods to eliminate interference.

[0135] Step S300: adjusting image parameters of the target image, identifying the target image to determine that the area where the crack is located is away from the curved second edge on one side of the blade, and obtaining second coordinate data of the vertex position of the second edge.

[0136] Step S310: determining whether there is a semiconductor pattern on the target image.

[0137] Step S320: There is a semiconductor pattern on the target image. After adjusting the image parameters of the target image, the target image is identified. The pattern in the crack area of ​​the target image is subjected to two edge detections with different detection parameters to obtain two edge detection result images.

[0138] Specifically, the same as step S210, first adjust the image parameters such as contrast, and the adjustment result is shown in the attached figure. Figure 14 As shown in the attached figure Figure 14 The following images show three different adjustment results, corresponding to different parameter adjustment values. Figure 14 The rightmost image has the best adjustment effect, that is, the grayscale of the crack region is close to that of the semiconductor pattern located therein, the contrast between the inside and outside of the crack region is large, and the edge boundary of the crack region is clear.

[0139] Then, based on the attached drawings Figure 14 The original target image on the right is subjected to two canny edge detections with different detection parameters. The detection parameters of the two detections are customized. In this embodiment, the parameters of the two detections are set to 0.5 and 0.2, and the image shown in the figure is obtained. Figure 15 and Figure 16The two edge detection result images shown in the figure. Among them, the larger parameter will only extract more obvious edges, while the smaller parameter will extract more detailed edges. Figure 15 The edge detection result image with detection parameter 0.5 is shown in the figure. Figure 16 This is the edge detection result image with detection parameter 0.2.

[0140] Step S330: superimposing the two edge detection result images, removing the overlapping portion of the two patterns, and outputting the result as the crack edge image.

[0141] Specifically, the edge detection result image with a larger parameter basically only contains the wafer edge and part of the blade edge that are very obvious in the original target image, while the edge detection result image with a smaller parameter also includes the crack area and the edge of the semiconductor pattern. The two edge detection result images are superimposed and the overlapping patterns of the two patterns are removed, that is, Figure 16 The pattern shown is subtracted Figure 15 The pattern shown in the figure can eliminate the interference of the blade part outside the wafer and the edge part of the wafer, and obtain Figure 17 The image shown. Figure 17 Small connected domains are eliminated from the image shown. The preset threshold of the small connected domain can be set in advance to eliminate some interfering pixels.

[0142] Step S340: Eliminate small connected domains in the crack edge image whose areas are smaller than a preset threshold.

[0143] Specifically, after removing the small connected domains from the crack edge image, it may be possible to directly exclude all other interfering pixels according to the connected domain of the original image, and obtain the following: Figure 18 The image shown.

[0144] Step S350: extracting the curve segment formed by the topmost pixel points from the crack edge image as the second edge according to the orientation and eliminating the rest of the crack edge image.

[0145] Specifically, the side of the crack away from the blade can be determined according to the orientation, i.e. Figure 18 The top of the image, i.e. the second edge, the initial extracted image corresponding to the extracted curve segment is shown in the attached figure Figure 19 shown.

[0146] Step S360: Eliminate small connected domains in the crack edge image whose areas are smaller than a preset threshold.

[0147] Specifically, it can be seen that when the wafer has a semiconductor pattern, there are still some interfering pixels on the extracted edge image, as shown in the accompanying figure. Figure 19As shown, it is necessary to remove the small connected domain again to eliminate the interference again. The resulting image after removing the small connected domain is shown in the attached figure. Figure 20 The preset threshold value for this small connected domain elimination can be different from the preset threshold value in step S340 , and the preset threshold values ​​for the two small connected domain eliminations can be set separately.

[0148] Step S370: intercepting a portion of the curve segment serving as the second edge that is within the x-coordinate range of the first coordinate data and discarding the remaining portion.

[0149] Specifically, referring to the range from the endpoints P1 to P2 of the straight line segment, the portion of the curve segment within the range [x1, x2] is intercepted. The interception result is shown in the attached figure. Figure 21 As shown, it is a discontinuous curve segment. Figure 22 It can be seen that the relative position of the discontinuous curve segment on the original target image after fitting, that is, the edge of the crack area corresponds to the part within the range of [x1, x2].

[0150] Step S380: randomly selecting a preset number of pixel points on the curve segment and determining the coordinates of the pixel points.

[0151] Specifically, after obtaining the attached drawings Figure 21 After finding the curve segment shown, it is necessary to determine the function expression of the curve corresponding to the curve segment. This is to accurately determine the coordinates of the vertex of the second edge. A predetermined number of pixels on the curve segment are selected and their coordinates are determined. The predetermined number can be a ratio of the total number of pixels, for example, 1 / 10 of the total number of pixels.

[0152] Step S390: performing curve fitting according to the pixel points and determining a function expression of the fitted curve according to the coordinates of the pixel points.

[0153] Specifically, based on the previous research on the edge shape of the crack area, it was concluded through verification and analysis that the edge shape of the crack area can be approximately described as a quartic function, that is, y = ax 4 +bx 3 +cx 2 +dx+e. Therefore, the coordinates of the acquired pixel points are directly substituted into the curve fitting, and the coefficients of the quartic function are solved using the least squares method to obtain the function expression of the fitted curve.

[0154] Step S391: According to the function expression of the curve, determine whether the image of the curve opens upward and has only one minimum extreme point.

[0155] Step S392: If yes, determine the coordinates of the extreme points on the curve according to the function expression of the curve.

[0156] Step S393: If not, randomly select a preset number of pixel points on the curve segment again to perform curve fitting and determine the function expression.

[0157] Step S3100: Calculate the coordinates of the extreme point on the curve according to the function expression of the curve. The coordinates of the extreme point are the second coordinate data of the position of the vertex of the second edge.

[0158] Specifically, if the curve fitting is successful, the coordinates of the only extreme point, ie, the coordinates of the minimum extreme point at the bottom vertex of the curve coordinate image, P3(x3, y'), can be directly determined based on the function expression of the curve.

[0159] Step S3101: Repeat the process from step S380 to step S3100 (including step S391 to step S393) a preset number of times (10 times) to obtain a preset number (10) of extreme point coordinate results.

[0160] Step S3102: removing the extreme point coordinates with abnormal coordinate data from a preset number of extreme point coordinate results.

[0161] Step S3103: Calculate the average value of the remaining extreme point coordinates in the extreme point coordinate results, and use the average value as the final coordinate of the extreme point.

[0162] Specifically, the method of averaging multiple results after eliminating abnormal data can further eliminate the interference of abnormal data on the detection results.

[0163] Step S400 : Calculating a target distance between a first edge and a second edge according to the first coordinate data and the second coordinate data.

[0164] Specifically, based on the coordinates of the endpoints of the straight line segment and the coordinates of the extreme points of the curve segment, namely P1(x1, y), P2(x2, y) and P3(x3, y'), the target distance between the first edge and the second edge can be directly calculated as |y'-y| (pixel unit). This target distance reflects the extent of the crack expansion toward the interior of the bonding interface.

[0165] Example 3

[0166] An embodiment of the present invention provides an image processing device for detecting bonding energy, comprising:

[0167] An image acquisition module, configured to acquire a target image including a blade and a crack generated by the insertion of the blade;

[0168] An image recognition module is used to recognize a target image and obtain an edge image of a target edge position of a blade and a crack on the target image;

[0169] an image processing module for processing the edge image to obtain coordinate data of the target edge of the blade and the crack;

[0170] The data calculation module is used to calculate the target distance between the target edge of the blade and the target edge of the crack according to the coordinate data.

[0171] Example 4

[0172] An embodiment of the present invention provides a bonding energy detection device, including the aforementioned image processing device, and thus having all the technical effects thereof.

[0173] In the description of the present invention, it should be understood that the terms "upper", "lower", "bottom", "top", "front", "back", "inside", "outside", "left", "right", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.

[0174] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be employed in conjunction with other described embodiments.

Claims

1. An image processing method for bonding energy detection, characterized in that: include: Acquire a target image including a blade and a crack generated by insertion of the blade; Identifying the target image to determine a first edge of the front end of the blade, and acquiring first coordinate data of a position of the first edge; Identify the target image to determine that the area where the crack is located is away from the curved second edge on one side of the blade, and obtain second coordinate data of the position of the vertex of the second edge; Calculating a target distance between the first edge and the second edge according to the first coordinate data and the second coordinate data; Recognizing the target image to determine a first edge of the front end of the blade includes: performing edge detection on a pattern at a position where the blade is located on the target image to obtain a blade edge image of the blade; extracting a straight line segment formed by the lowest pixel points from the blade edge image as the first edge according to the orientation and eliminating the remaining portion of the blade edge image; Recognizing the target image to determine that the crack region is away from the curved second edge of one side of the blade comprises: Performing edge detection on a pattern in the target image where the crack is located to obtain a crack edge image of the crack area where the crack is located; extracting a curve segment formed by the topmost pixel points from the crack edge image as a second edge according to the orientation and eliminating the remaining portion of the crack edge image; Determining whether the target image has a semiconductor pattern; If there is a semiconductor pattern on the target image, performing two edge detections with different detection parameters on the pattern in the area where the crack is located on the target image to obtain two edge detection result images; The two edge detection result images are superimposed, and the overlapping parts of the two patterns are eliminated, and the resulting image is the crack edge image.

2. The image processing method for bonding energy detection according to claim 1, characterized in that: Acquiring first coordinate data of the position of the first edge includes: According to the straight line segment corresponding to the first edge, first coordinate data of two endpoints at both ends of the straight line segment are determined by using Hough transform.

3. The image processing method for bonding energy detection according to claim 1, characterized in that: Also includes: A curve segment consisting of the topmost pixel points is extracted from the crack edge image as the front and / or back of the second edge, and small connected domains with an area smaller than a preset threshold in the image are eliminated.

4. The image processing method for bonding energy detection according to claim 1, characterized in that: Obtaining second coordinate data of the vertex position of the second edge includes: Randomly selecting a preset number of pixel points on the curve segment and determining the coordinates of the pixel points; Performing curve fitting according to the pixel points and determining a function expression of the fitted curve according to the coordinates of the pixel points; The coordinates of the extreme point on the curve are calculated according to the function expression of the curve, and the coordinates of the extreme point are the second coordinate data of the position of the vertex of the second edge.

5. The image processing method for bonding energy detection according to claim 4, characterized in that: Also includes: According to the function expression of the curve, determining whether the graph of the curve opens upward and has only one minimum extreme value point; If so, determining the coordinates of the extreme points on the curve according to the function expression of the curve; If not, a preset number of pixel points are randomly selected on the curve segment again to perform curve fitting and determine the function expression.

6. The image processing method for bonding energy detection according to claim 4 or 5, characterized in that: Also includes: Repeating the process from selecting pixel points to determining a curve function expression and calculating the coordinates of extreme value points on the curve for a preset number of times to obtain a preset number of extreme value point coordinate results; Eliminate the extreme point coordinates with abnormal coordinate data from the preset number of extreme point coordinate results; The average value of the remaining extreme point coordinates in the extreme point coordinate results is calculated, and the average value is used as the final coordinate of the extreme point.

7. The image processing method for bonding energy detection according to claim 1, characterized in that: Also includes: A portion of the curve segment as the second edge that is within the x-coordinate range of the first coordinate data is intercepted and the remaining portion is discarded.

8. The image processing method for bonding energy detection according to claim 1, characterized in that: Before identifying the target image to determine the second edge, the method further includes: The target image is cropped according to the y-coordinate data of the first coordinate data to crop the image of the blade portion on the target image.

9. The image processing method for bonding energy detection according to claim 1, characterized in that: Also includes: Before identifying the target image, image parameters of the target image are adjusted, where the image parameters include contrast and brightness.

10. An image processing device for bonding energy detection that processes images according to the image processing method for bonding energy detection according to any one of claims 1 to 9, characterized in that: include: An image acquisition module, configured to acquire a target image including a blade and a crack generated by the insertion of the blade; an image recognition module, configured to recognize the target image and obtain an edge image of a target edge position of the blade and the crack on the target image; an image processing module, configured to process the edge image to obtain coordinate data of the target edge of the blade and the crack; A data calculation module is used to calculate a target distance between a target edge of the blade and a target edge of the crack according to the coordinate data.

11. A bonding energy detection device, characterized in that: The apparatus comprises the image processing device according to claim 10.

Images