Intelligent visual analysis method applied to mold clamping

Abstract

The application provides an intelligent visual analysis method applied to mold clamping, and belongs to the technical field of mold intelligent analysis. The method solves the problem of low clamping efficiency of the existing manual detection. The intelligent visual analysis method applied to mold clamping comprises the following steps: collecting a fixed mold image; performing segmentation processing on the obtained fixed mold image to obtain a fixed mold region segmentation mask image, and then obtaining a fixed mold region; performing a red lead region analysis processing on the obtained fixed mold image to obtain a red lead region confidence map and a red lead region mask image; performing color space conversion processing on the obtained fixed mold image, and then sequentially obtaining a distance tensor map and a color depth tensor map; calculating the obtained color depth tensor map and the red lead region confidence map to obtain a uniformity value; when the distribution and depth of the fixed mold red lead are judged to not satisfy the mold clamping requirements according to the uniformity value, a processing suggestion value for mold polishing is generated; otherwise, the mold clamping is completed. The application can improve the clamping efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent mold analysis and relates to an intelligent visual analysis method applied to mold closing. Background Art

[0002] A mold is a form created in industrial production through methods such as injection molding, die casting, and stamping. It has a specific contour or internal cavity shape and is used for forming products. A mold typically consists of a movable mold and a fixed mold. During injection molding, these two molds close together to form the gating system and mold cavity. Mold closing is a critical step in the moldmaking process. Improper mold closing can lead to unstable panels and mislead panel designers during subsequent adjustments, resulting in unnecessary workload and wasted manpower and resources.

[0003] The current mold closing process involves installing the movable and fixed molds on a mold closing machine. The machine then closes the two molds, allowing the red lead (a type of fuel) evenly applied to the movable mold to spread onto the fixed mold through the closing process. The fixed mold red lead analysis phase determines the closure of the movable and fixed molds during mold closing based on the distribution and depth of the red lead on the fixed mold. The fixed mold uniformity analysis phase determines whether the fixed mold red lead analysis meets the requirements. If so, mold closing is complete. If not, the mold is polished based on the analysis results and the above process is repeated until the requirements are met. Currently, both fixed mold red lead analysis and fixed mold uniformity analysis rely on manual judgment. Due to differences in individual experience, it is impossible to form a stable, unified, and precisely quantifiable uniformity analysis result. This results in low analysis efficiency and accuracy, and also reduces the efficiency of subsequent polishing and mold closing. Summary of the Invention

[0004] The purpose of the present invention is to address the above-mentioned problems existing in the prior art and to propose an intelligent visual analysis method for mold closing. The technical problem to be solved is: how to improve the efficiency of mold closing.

[0005] The object of the present invention can be achieved by the following technical solution: An intelligent visual analysis method for mold closing, comprising the following steps:

[0006] A. Acquire the fixed mold and the fixed mold image of the area covered by red lead on the fixed mold;

[0007] B. Segmenting the obtained fixed mold image to obtain a fixed mold region segmentation mask image, thereby obtaining the fixed mold region;

[0008] C. Performing red lead region analysis on the obtained fixed model image to obtain a red lead region confidence map and a red lead region mask map;

[0009] D. Performing color space conversion on the obtained fixed model image to obtain a distance tensor map and a color depth tensor map in sequence;

[0010] E. Calculating a red lead color expectation map based on the obtained color depth tensor map and the red lead region confidence map, and then calculating a uniformity value based on the red lead color expectation map and the image width and image height of the red lead color expectation map;

[0011] F. Determine whether the distribution and depth of the fixed mold red lead meet the mold closing requirements based on the uniformity value. If the mold closing requirements are met, the mold closing is completed. If the mold closing requirements are not met, generate the mold polishing processing recommendation value based on the fixed mold area segmentation mask map and the red lead color expected map.

[0012] The principle of this intelligent visual analysis method for mold clamping is as follows: the clamping machine closes the fixed mold and the movable mold. At this time, the red lead evenly applied on the movable mold will be applied to the fixed mold through closing. This method is to analyze the red lead application on the fixed mold to determine whether it meets the mold clamping requirements. First, the fixed mold image is collected, and then the fixed mold image is processed separately to obtain the fixed mold area segmentation mask map, red lead area confidence map, red lead area mask map, distance tensor map and color depth tensor map. Then, the red lead color expectation map is obtained based on the color depth tensor map and the red lead area confidence map, and then the color depth tensor map is used to calculate the red lead color expectation map. The uniformity value is calculated based on the expected color map of red lead. Finally, the distribution and depth of the red lead in the fixed mold are judged based on the uniformity value to see whether they meet the mold closing requirements. If not, the processing recommendation value for mold polishing is generated based on the fixed mold area segmentation mask map and the expected color map of red lead. The mold is re-polished according to the processing recommendation value, and the polishing is more precise, which provides a guarantee for the success rate of re-closing. After the mold is polished, the movable mold and the fixed mold coated with red lead are re-closed, and the analysis and judgment of whether the mold closing requirements are met are re-based on this method. The application of this method effectively improves the efficiency of subsequent polishing and mold closing.

[0013] In the above-mentioned intelligent visual analysis method applied to mold closing, in step B, the operation of obtaining the fixed mold area segmentation mask image includes:

[0014] The obtained fixed pattern image is segmented using a fixed pattern region segmentation algorithm, which outputs pixel-by-pixel fixed pattern region segmentation results, thereby forming a fixed pattern region segmentation mask. The fixed pattern region segmentation algorithm can use a semantic segmentation algorithm to achieve pixel-level image classification. The fixed pattern region segmentation mask is an image that distinguishes the fixed pattern region from the non-fixed pattern region using black and white pixels.

[0015] In the above-mentioned intelligent visual analysis method applied to mold closing, in the shown step B, the operation of obtaining the fixed mold area includes: using an edge detection algorithm to obtain the edge contour of the black and white intersection in the fixed mold area segmentation mask image, and setting coordinate points at the edge contour, thereby obtaining the fixed mold area surrounded by the coordinate points.

[0016] In the above-mentioned intelligent visual analysis method applied to mold closing, in step C, the operation of obtaining the red lead area confidence map includes:

[0017] First, the fixed image is processed using the red lead region segmentation algorithm, which outputs a pixel-by-pixel two-channel feature map. The softmax function is then used to calculate the two-channel feature map to obtain the red lead region confidence map. The red lead region segmentation algorithm uses a semantic segmentation algorithm based on a convolutional neural network. The softmax function is also called the normalized exponential function.

[0018] In the above-mentioned intelligent visual analysis method applied to mold closing, in step C, the operation of obtaining the red lead area mask image includes:

[0019] The argmax function is used to perform a pixel-by-pixel comparison of the two-channel feature maps output by the red lead region segmentation algorithm. When the result is the channel number corresponding to the red lead region category, the compared pixel is determined to belong to the red lead region, thereby obtaining a red lead region mask. The red lead region mask distinguishes the red lead region from the non-red lead region using black and white pixels, thereby determining the red lead region.

[0020] In the above-mentioned intelligent visual analysis method applied to mold closing, in step D, the operation of obtaining the distance tensor graph includes:

[0021] The fixed model image is converted into the HSV color space through color space conversion to obtain the color space tensor, and then the distance between the color space tensor and the set color space center point is calculated to obtain the distance tensor map.

[0022] In the above-mentioned intelligent visual analysis method applied to mold closing, in step D, the operation of obtaining the color depth tensor map includes:

[0023] Flip the distance tensor map to obtain the color depth tensor map.

[0024] In the above-mentioned intelligent visual analysis method applied to mold closing, in step E, the operation of obtaining the desired red lead color map includes:

[0025] The color depth tensor map and the red lead area confidence map are multiplied pixel by pixel to obtain the red lead color expectation map. The specific formula is as follows:

[0026] E(u,v)=p(u,v)*x(u,v)

[0027] Among them, u and v represent the coordinates in the pixel coordinate system, P represents the red lead area confidence map; x represents the color depth tensor map; and E represents the red lead color expectation map.

[0028] In the above-mentioned intelligent visual analysis method applied to mold closing, in step E, the red lead color expectation image, image width and image height are calculated using a uniformity calculation formula to obtain a uniformity value. The uniformity calculation formula is:

[0029]

[0030] Where Q represents the uniformity value; M represents the image width; and N represents the image height.

[0031] In the above-mentioned intelligent visual analysis method applied to mold closing, in step F, the operation of judging whether the distribution and depth of the fixed mold red lead meet the mold closing requirements according to the uniformity value includes:

[0032] Set the uniformity standard value;

[0033] The uniformity value is compared with the uniformity standard value. When the uniformity value is greater than the uniformity standard value, the mold closing completion step is entered; when the uniformity value is less than the uniformity standard value, the processing suggestion value generation step is entered.

[0034] In the above-mentioned intelligent visual analysis method applied to mold clamping, in step F, the operation of processing the step of generating the recommended value includes:

[0035] The red lead color expectation map is flipped pixel by pixel to form a flipped expectation map;

[0036] Using the fixed mold area segmentation mask image to perform mask processing on the expected value of red lead color to form a fixed mold area flip expected image;

[0037] Connect the clustered areas in the fixed mold area flip expectation map and use the average of the flip expectation values ​​of each clustered area as the processing recommendation value. The processing recommendation value is obtained by the following calculation formula: Where I represents a vector composed of pixels in a certain clustered area, x represents the vector subscript, and N represents the length of the vector.

[0038] Compared with existing technologies, this intelligent visual analysis method for mold closing has the following advantages:

[0039] 1. The present invention realizes the intelligent analysis of the distribution of fixed-mode red lead without human participation, thus effectively saving manpower.

[0040] 2. The visual analysis method adopted by the present invention can automatically complete the multi-dimensional and refined distribution analysis of the fixed mold red lead in each area, and the quantitative analysis of the depth of the red lead in each part, so that the analysis is more accurate. Furthermore, through the multi-dimensional and refined distribution analysis and the quantitative analysis of the depth, a stable, unified and accurately quantifiable uniformity analysis result can be obtained, and then the subsequent processing suggestions for the mold are given according to the analysis results, so that the processing personnel can polish the mold in a targeted manner so that it can meet the mold closing requirements more quickly. This method effectively improves the efficiency of subsequent polishing and mold closing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a control flow diagram of the present invention. DETAILED DESCRIPTION

[0042] To make the objectives, technical solutions, and advantages of the present invention more apparent, embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention, its application, or use. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0043] like Figure 1 As shown, when the intelligent visual analysis method for mold closing is applied, the images of the fixed mold and the area covered by red lead on the fixed mold are first collected to obtain a fixed mold image; then the fixed mold area segmentation step is entered, and the obtained fixed mold image is segmented to obtain a fixed mold area segmentation mask map, and then the fixed mold area is obtained. The specific operation is: the obtained fixed mold image is segmented by using a fixed mold area segmentation algorithm, and the fixed mold area segmentation result is output pixel by pixel, thereby forming a fixed mold area segmentation mask map. The fixed mold area segmentation algorithm can adopt a semantic segmentation algorithm, which is an existing algorithm. The semantic segmentation algorithm can realize pixel-level classification of the image, and distinguish the fixed mold area and the non-fixed mold area by black and white pixels to form a fixed mold area segmentation mask map. For the obtained fixed mold area segmentation mask map, an edge detection algorithm is used to determine the edge contour of the black and white boundary, and coordinate points are set at the edge contour to obtain the fixed mold area surrounded by the coordinate points, so as to achieve effective distinction between the fixed mold area and the non-fixed mold area. The edge detection algorithm used in this step is an existing algorithm.

[0044] At the same time, the red lead region analysis step performs red lead region analysis on the obtained fixed model image to obtain a red lead region confidence map and a red lead region mask map. The specific operations are as follows:

[0045] First, the obtained fixed image is processed using the red lead region segmentation algorithm to output a pixel-by-pixel two-channel feature map. The softmax function is then used to calculate the two-channel feature map to obtain a red lead region confidence map. The two-channel feature map output by the red lead region segmentation algorithm is compared pixel by pixel using the argmax function. When the result is the channel number corresponding to the red lead region category, the compared pixel is determined to belong to the red lead region, thereby obtaining a red lead region mask map. The red lead region segmentation algorithm uses a semantic segmentation algorithm based on a convolutional neural network. The softmax function is also known as the normalized exponential function. Semantic segmentation algorithms, softmax functions, and argmax functions are all existing technologies. During operation, assuming the fixed image is a tensor of [w, h, 3], the red lead region segmentation algorithm outputs a two-channel feature map (tensor) of [w, h, 2]. Where w represents the image width, h represents the image height, and 2 or 3 represents the channel dimension. The red lead region confidence map has the shape of [w, h, 1]. The value at each pixel represents the confidence (probability) that the pixel belongs to the red lead region.

[0046] At the same time, the color calculation step performs color space conversion on the obtained fixed model image to obtain the distance tensor map and color depth tensor map in turn. The specific operations are as follows:

[0047] The fixed image is converted into the HSV color space through color space conversion to obtain the color space tensor, and then the distance between the color space tensor and the set color space center point is calculated to obtain the distance tensor map. The distance tensor map is flipped to obtain the color depth tensor map. For example, assuming that the fixed image is a tensor of [w,h,3], the shape of the tensor in the color space is [w,h,3]. If converted to the HSV color space, the 3 channels represent the three dimensions of H (hue), S (saturation), and V (lightness). The distance tensor map calculation step is to calculate the distance between the color space tensor and the color space center point. Assuming that the color space center point is (c1 c , c2 c , c3 c ), the shape of the color space tensor is [w,h,3], then the value corresponding to the point with coordinates (u,v) in the color space tensor is (c1 u,v , c2 u, , c3 u, ), the distance from that point to the center point can be calculated. This results in a pixel-by-pixel distance tensor with a shape of [w, h, 1]. The color depth tensor is calculated by flipping the distance tensor to a shape of [w, h, 1].

[0048] Then, the red lead color expectation map is calculated based on the obtained color depth tensor map and the red lead area confidence map, and then the uniformity value is calculated based on the red lead color expectation map and the image width and image height of the red lead color expectation map. The specific operation is:

[0049] The red lead color expectation map is obtained by multiplying the color depth tensor map and the red lead area confidence map pixel by pixel. The specific formula is as follows:

[0050] E(u,v)=p(u,v)*x(u,v)

[0051] Among them, u and v represent the coordinates in the pixel coordinate system, P represents the red lead area confidence map; x represents the color depth tensor map; and E represents the red lead color expectation map.

[0052] The uniformity value is calculated by using the uniformity calculation formula to calculate the red lead color expectation image, image width and image height to obtain the uniformity value, wherein the uniformity calculation formula is:

[0053]

[0054] Where Q represents the uniformity value; M represents the image width; and N represents the image height.

[0055] Finally, the distribution and depth of the fixed mold red lead are judged based on the uniformity value to see whether they meet the mold closing requirements. If they meet the mold closing requirements, the mold closing is completed. If they do not meet the mold closing requirements, the processing recommendation value for mold polishing is generated based on the fixed mold area segmentation mask map and the red lead color expected map. The specific operation is: first set the uniformity standard value; then compare the uniformity value with the uniformity standard value, and when the uniformity value is greater than the uniformity standard value, enter the mold closing completion step; when the uniformity is less than the uniformity standard value, enter the processing recommendation value generation step: first flip the red lead color expected map pixel by pixel to form a flipped expected map; then use the fixed mold area segmentation mask map to mask the red lead color expected value to form a fixed mold area flipped expected map; finally, connect the clustered areas in the fixed mold area flipped expected map, and use the average of the flipped expected values ​​of each clustered area as the processing recommendation value. The processing recommendation value is calculated using the following calculation formula: Where I represents a vector of pixels in a clustered area, x represents the vector subscript, and N represents the vector length. Based on this recommended processing value, personnel can re-polish the mold in a targeted manner, achieving more precise polishing and improving efficiency during subsequent mold re-closing. Finally, this recommended processing value is integrated with the previously calculated uniformity value and stored in a database. This operation accumulates experience and data assets for the intelligent mold closing system and the Industrial Internet of Things system, helping to improve mold closing efficiency.

[0056] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.

Claims

1. An intelligent visual analysis method for mold closing, characterized in that: The intelligent analysis method includes the following steps: A. Acquire the fixed mold and the fixed mold image of the area covered by red lead on the fixed mold; B. Segmenting the obtained fixed mold image to obtain a fixed mold region segmentation mask image, thereby obtaining the fixed mold region; C. Performing red lead region analysis on the obtained fixed model image to obtain a red lead region confidence map and a red lead region mask map; D. Performing color space conversion on the obtained fixed model image to obtain a distance tensor map and a color depth tensor map in sequence; E. Calculating a red lead color expectation map based on the obtained color depth tensor map and the red lead region confidence map, and then calculating a uniformity value based on the red lead color expectation map and the image width and image height of the red lead color expectation map; F. Determine whether the distribution and depth of the fixed mold red lead meet the mold closing requirements based on the uniformity value. If the mold closing requirements are met, the mold closing is completed. If the mold closing requirements are not met, generate the mold polishing processing recommendation value based on the fixed mold area segmentation mask map and the red lead color expected map.

2. The intelligent visual analysis method for mold closing according to claim 1, characterized in that: In step B, the operation of obtaining the fixed mold area segmentation mask image includes: The obtained fixed mold image is segmented using a fixed mold region segmentation algorithm, and then a pixel-by-pixel fixed mold region segmentation result is output, thereby forming a fixed mold region segmentation mask image.

3. The intelligent visual analysis method for mold closing according to claim 2, characterized in that: In step B, the operation of obtaining the fixed mold area includes: using an edge detection algorithm to obtain the edge contour of the black and white boundary in the fixed mold area segmentation mask image, and setting coordinate points at the edge contour to obtain the fixed mold area surrounded by the coordinate points.

4. The intelligent visual analysis method for mold closing according to claim 1, characterized in that: In step C, the operation of obtaining the red lead area confidence map includes: First, the red lead region segmentation algorithm is used to process the obtained fixed model image, and a two-channel feature map is output pixel by pixel. Then, the softmax function is used to calculate the two-channel feature map to obtain the red lead region confidence map.

5. The intelligent visual analysis method for mold closing according to claim 4, characterized in that: In step C, the operation of obtaining the red lead area mask image includes: The argmax function is used to compare the two-channel feature maps output by the red lead area segmentation algorithm pixel by pixel. When the result is the channel number corresponding to the red lead area category, it is determined that the compared pixel belongs to the red lead area, thereby obtaining the red lead area mask map.

6. The intelligent visual analysis method for mold closing according to claim 4 or 5, characterized in that: In step D, the operations of obtaining the distance tensor map and the color depth tensor map include: The fixed model image is converted into the HSV color space through color space conversion to obtain the color space tensor, and then the distance between the color space tensor and the set color space center point is calculated to obtain the distance tensor map; Flip the distance tensor map to obtain the color depth tensor map.

7. The intelligent visual analysis method for mold closing according to claim 6, characterized in that: In step E, the operation of obtaining the desired red lead color diagram includes: The color depth tensor map and the red lead area confidence map are multiplied pixel by pixel to obtain the red lead color expectation map. The specific formula is as follows: Eu,v=pu,v*x(u,v) Among them, u and v represent the coordinates in the pixel coordinate system, P represents the red lead area confidence map; x represents the color depth tensor map; and E represents the red lead color expectation map.

8. The intelligent visual analysis method for mold closing according to claim 7, characterized in that: In step E, the uniformity value is obtained by calculating the red lead color desired image and the image width and image height of the red lead color desired image using a uniformity calculation formula. The uniformity calculation formula is: Where Q represents the uniformity value; M represents the image width; and N represents the image height.

9. The intelligent visual analysis method for mold closing according to claim 8, characterized in that: In step F, the operation of judging whether the distribution and depth of the fixed mold red lead meet the mold closing requirements according to the uniformity value includes: Set the uniformity standard value; The uniformity value is compared with the uniformity standard value. When the uniformity value is greater than the uniformity standard value, the mold closing completion step is entered; when the uniformity value is less than the uniformity standard value, the processing suggestion value generation step is entered.

10. The intelligent visual analysis method for mold closing according to claim 9, characterized in that: In step F, the operations of generating the recommended value include: The red lead color expectation map is flipped pixel by pixel to form a flipped expectation map; Using the fixed mold area segmentation mask image to perform mask processing on the expected value of red lead color to form a fixed mold area flip expected image; Connect the clustered areas in the fixed mold area flip expectation graph, and use the average of the flip expectation values ​​of each clustered area as the processing recommendation value.

Images