An intelligent detection method for parts processed by CNC machine tools

Abstract

The present invention relates to the field of image processing technology, and specifically discloses an intelligent detection method for parts processed by CNC machine tools, comprising the following steps: acquiring an infrared image of the part image; dividing the part image into multiple local areas having the same temperature range based on the infrared image; if the part contains a coating and the temperature of one or more local areas is greater than or less than a standard temperature, correcting the initial edges of the one or more local areas to obtain corrected edges; calculating the roundness error of the part image based on the corrected edges of the part; and determining that the corresponding part has a defect when the roundness error of the part image is greater than an error threshold. The intelligent detection method for parts processed by CNC machine tools provided by the present invention makes the roundness error detection results of the parts more accurate and reasonable, preventing misjudgments.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an intelligent detection method for parts processed by a numerically controlled machine tool. Background Art

[0002] CNC machine tools are core equipment in modern manufacturing. Using computer numerical control technology to precisely control machine tools, they are capable of producing parts of various complex shapes. Machined parts may exhibit various defects, including roundness errors, burrs, and surface irregularities. To prevent unqualified parts from being shipped, parts must be inspected for defects before leaving the factory. For circular or annular parts, roundness testing is a crucial step in assessing whether the circular profile of the part meets design requirements. Roundness testing typically involves obtaining the actual contour of the part and comparing it to an ideal circle to determine the roundness of the part.

[0003] Traditional methods for detecting part defects rely primarily on manual visual inspection, which suffers from low efficiency, strong subjectivity, and low accuracy. With the development of automatic defect detection technology based on machine vision, its application to part defect detection is enabling highly efficient and accurate defect detection. For example, edge detection technology can be used to determine the edge of a part, then automatically measure the roundness of the edge to determine whether the part is defective.

[0004] However, in the prior art, when performing roundness detection on parts, the effects of temperature and part coating on the roundness detection of parts are not taken into account. Since the parts processed by CNC machine tools are mostly made of metal materials such as steel, aluminum alloy, and copper alloy, these materials will expand and contract under the influence of temperature. In addition, some parts are coated with wear-resistant coatings, corrosion-resistant coatings, or oxidation-resistant coatings on their surfaces. Under the influence of temperature, the coatings of these parts are more likely to expand and contract, causing the contours of the parts to deform, which in turn causes the roundness error of circular or annular parts to increase. It cannot reflect whether the roundness error of the part is caused by processing reasons, and thus misjudgment occurs. Summary of the Invention

[0005] The present invention provides an intelligent detection method for parts processed by CNC machine tools, aiming to solve the technical problem in the prior art that the roundness error of parts becomes larger due to thermal expansion and contraction of parts, thereby causing misjudgment during detection.

[0006] The present invention provides an intelligent detection method for parts processed by a numerically controlled machine tool, comprising the following steps:

[0007] Acquire an infrared image of the part image;

[0008] Dividing the part image into a plurality of local areas having the same temperature range according to the infrared image;

[0009] If the part contains a coating and the temperature of one or more local areas is greater than or less than the standard temperature, the initial edge of the one or more local areas is corrected to obtain a corrected edge;

[0010] Calculate the roundness error of the part image based on the corrected edge of the part; when the roundness error of the part image is greater than the error threshold, it is determined that the corresponding part has a defect;

[0011] The acquisition of the corrected edge comprises the following steps:

[0012] Calculate the correction amount for each point on the initial edge; the correction amount is the product of the thermal expansion coefficient of the part and the absolute value of the temperature difference in the local area where the corresponding point on the initial edge is located; the temperature difference is the difference between the average temperature in the corresponding local area and the standard temperature; the direction of the correction amount is the perpendicular direction of the tangent line of the corresponding point on the initial edge;

[0013] The points on the initial edge are moved in the direction of the correction amount by a corresponding correction amount and connected in sequence to form the corrected edge.

[0014] In the above scheme, the corrected edge is obtained by combining the original edge and the correction amount of each point on it, which corrects the displacement caused by thermal expansion and contraction, making the roundness error calculated based on the corrected edge more reasonable and accurate, and preventing the occurrence of misjudgment during detection.

[0015] Preferably, the local area with the same temperature range is an area in which the temperature difference between positions in the part image does not exceed the temperature threshold, and the temperature gradient at each position does not exceed the temperature gradient threshold.

[0016] In the above scheme, the areas with similar temperatures and small temperature changes in the part image are divided into the same local area according to the temperature difference and temperature gradient, which facilitates the calculation of the temperature change in the local area, and then facilitates the correction of the initial edge of the same local area in the subsequent steps.

[0017] Preferably, when the correction amount is a positive value, the corresponding point on the initial edge moves toward the inside of the initial edge along the perpendicular direction of its tangent; when the correction amount is a negative value, the corresponding point on the initial edge moves toward the outside of the initial edge along the perpendicular direction of its tangent.

[0018] In the above scheme, the displacement of each point on the initial edge caused by thermal expansion and contraction is corrected by moving the corresponding correction amount, so that the corrected edge can eliminate the interference of temperature and can truly reflect the roundness error of the part caused by processing reasons.

[0019] Preferably, the infrared image of the part image is acquired using an infrared thermal imager.

[0020] In the above solution, the infrared thermal imager can realize non-contact measurement, obtain the temperature at each position of the part, and intuitively reflect the temperature distribution.

[0021] Preferably, the initial edge of the part image is obtained by a Sobel edge detection algorithm or a Canny edge detection algorithm.

[0022] Among the above schemes, the Sobel edge detection algorithm has the advantages of being simple and easy to implement and being robust to noise. The Canny edge detection algorithm has the advantages of being high in accuracy and being robust to noise.

[0023] Preferably, the standard temperature is the average temperature of each component under room temperature conditions.

[0024] Preferably, the roundness error of the part image is calculated using a least squares method.

[0025] In the above scheme, the least square method is used to calculate the roundness error, which has the advantages of high accuracy, strong robustness, and simple calculation.

[0026] Preferably, the roundness error of the part image is calculated using the least squares method, comprising the following steps:

[0027] Constructing an error function, where the error function is the sum of the squares of the radial deviations of all points on the corrected edge from the ideal circle; the radial deviation is the difference between the distance between the corresponding point on the corrected edge and the assumed center of the ideal circle and the assumed radius of the ideal circle;

[0028] Minimize the error function to obtain the true center and true radius of the ideal circle;

[0029] Calculate the radial deviation of each point on the corrected edge based on the true center and true radius of the ideal circle;

[0030] The difference between the maximum and minimum radial deviations of each point on the corrected edge is taken as the roundness error of the corresponding part image.

[0031] Preferably, the part image is an image that has been filtered and denoised, wherein the filtering method is one of Gaussian filtering, median filtering and mean filtering.

[0032] In the above scheme, filtering and denoising can prevent noise from interfering with the part image, making the calculation results more accurate.

[0033] Preferably, the shape of the part is circular or ring-shaped.

[0034] The beneficial effects are:

[0035] The solution of the present invention comprehensively considers the impact of a part's temperature distribution and coating condition on its roundness error. A corrected edge is obtained by combining the original edge of the part image with the corrections made to each point on it. This is used to calculate the roundness error of the part image. This corrects for the displacement of each point on the original edge of the part caused by thermal expansion and contraction, making the calculation of the part's roundness error more reasonable and accurate, and preventing misjudgments during detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flowchart of the steps of an intelligent detection method for parts processed by a CNC machine tool according to an embodiment of the present invention;

[0037] Figure 2 This is a flow chart of the steps for obtaining a corrected edge according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention, but should not be construed as limiting the present invention.

[0039] The present invention provides an intelligent detection method for parts processed by CNC machine tools. The detection object of the present invention is a circular part, which is hereinafter referred to as parts for simplicity. The processing defects of parts can be detected by roundness detection, wherein roundness is a geometric property that describes a circular object, which reflects the degree to which the cross section of the object is close to an ideal circle. The parts in the present invention are circular, annular, and parts that are close to circular or annular. For example, if the roundness error meets the requirements, it means that the part has no processing defects. If the roundness error does not meet the requirements, it means that the part has defects. These defects include protrusions, depressions, or missing corners on the edge of the part. However, in actual scenarios, during the circulation of parts, the surface temperature distribution of the parts may be uneven due to changes in ambient temperature, touch, friction, and other reasons.

[0040] When a part is heated, the molecules or atoms inside it gain energy and move faster, causing the average distance between molecules or atoms to increase, thereby expanding the part's volume. Conversely, when a part is cooled, the movement of molecules or atoms slows down, the average distance decreases, and the part's volume shrinks.

[0041] When the temperature distribution on a part's surface is uneven, hotter areas expand due to heating, while cooler areas contract due to cooling. This means that different areas of the part expand or contract, causing the deformation to vary in size and direction. This in turn causes the part's roundness to change, leading to a larger roundness error.

[0042] Therefore, increasing roundness errors caused by thermal expansion and contraction can interfere with the roundness inspection process, leading to errors in test results and misjudgments. In other words, some parts identified as defective may be due to thermal expansion and contraction rather than processing. However, once these parts return to their original shape after a uniform temperature distribution, their roundness errors meet the required standards. Therefore, accounting for temperature interference when inspecting for part defects can ensure more accurate and reasonable test results and avoid misjudgments.

[0043] like Figure 1 As shown, the intelligent detection method for parts processed by CNC machine tools provided by the present invention includes the following steps:

[0044] S1. Obtain an infrared image of a part image.

[0045] In this step, the temperature at each location in the part image is determined by capturing an infrared image of the part. This infrared image can be acquired using an infrared thermal imager. An infrared thermal imager is an instrument capable of detecting and displaying the temperature distribution on an object's surface, suitable for non-contact measurement. It not only measures the temperature at each location in the part image, but also captures the infrared radiation emitted by the object and converts it into a visual image, allowing users to visually visualize temperature differences at various locations. Acquiring an infrared image facilitates subsequent processing of the part image.

[0046] In order to reduce the interference of noise on the part image, the part image is an image after filtering and denoising, wherein the filtering method is one of Gaussian filtering, median filtering and mean filtering. Of course, other filtering and denoising methods can also be used and can be selected according to needs.

[0047] S2. Divide the part image into a plurality of local areas having the same temperature range according to the infrared image.

[0048] In this step, the local area with the same temperature range is an area in which the temperature difference between each position in the part image does not exceed the temperature threshold, and the temperature gradient at each position does not exceed the temperature gradient threshold. The temperature threshold and the temperature gradient threshold can be determined by experiments. Among them, the temperature gradient characterizes the temperature change of the corresponding part at each position. The temperature gradient refers to the rate of change of temperature in space, which is expressed as the ratio of the difference between the temperature at the corresponding position and its adjacent position in the part image to the Euclidean distance between the two positions. It describes how fast the temperature changes with position. The direction of the temperature gradient points to the direction in which the temperature increases fastest, and its magnitude represents the temperature change per unit distance. According to the temperature at each position of the part image, the temperature gradient at the corresponding position can be obtained. According to the temperature difference between each position in the part image and the temperature gradient at each position, the part image is divided into multiple local areas.

[0049] In one embodiment, a region growing algorithm can also be used to divide the part image into multiple local regions. This region growing algorithm is a state-of-the-art technique. Its basic concept is to select a number of pixels from the image as seed points and, based on a similarity criterion, gradually merge surrounding pixels similar to the seed points into the same region until no more pixels can be merged, thereby obtaining local regions. The similarity criterion is the similarity of the grayscale values ​​of the pixels in the part image. Because the temperature at different locations on the part is different, the grayscale values ​​of the corresponding pixels in the part image are also different. The higher the temperature, the larger the grayscale value of the pixel at the corresponding location. Therefore, the region growing algorithm can divide regions with similar temperatures into the same local region by comparing the similarity of their grayscale values.

[0050] In other embodiments, a watershed algorithm may also be used, which belongs to the prior art and will not be described in detail here.

[0051] In step S2, the areas with similar temperatures and small temperature changes in the part image are divided into the same local area, which is convenient for calculating the temperature change in the local area and further facilitating the correction of the initial edge of the same local area in subsequent steps.

[0052] S3. If the part has a coating and the temperature of one or more local areas is greater than or less than the standard temperature, the initial edges of the one or more local areas are corrected to obtain corrected edges.

[0053] This step comprehensively considers the roundness error of a part based on two factors: the difference between the local temperature and the standard temperature, as well as the coating condition of the part. The standard temperature is the average temperature of all parts at room temperature. If the temperature of a local area is higher or lower than the standard temperature, it indicates uneven temperature distribution in the corresponding part, causing thermal expansion and contraction. However, since most parts are made of metal and are relatively hard, the impact of temperature alone on the roundness error is not significant.

[0054] Therefore, the present invention also considers the influence of whether a part contains a coating. A coating is a layer of material applied to a part's surface to protect it, enhance its performance, or alter its appearance. Examples include wear-resistant coatings, corrosion-resistant coatings, and insulating coatings. Because coatings on parts are typically thin, they are more susceptible to deformation due to thermal expansion and contraction. Therefore, when calculating a part's roundness error, it is necessary to consider not only the effects of temperature but also the presence or absence of a coating.

[0055] The present invention comprehensively considers the influence of two factors, temperature and coating of the part, on the roundness error of the part, and calculates the roundness error of the corresponding part based on the corrected edge in the part image, so that the calculation result of the roundness error of the part is more accurate and reasonable, thereby avoiding misjudgment during part detection.

[0056] The present invention first illustrates the situation where two factors are simultaneously met: a coated part and one or more localized areas with a temperature greater than or less than the reference temperature. Specifically, when the temperature of a localized area is greater than or less than the reference temperature, it indicates that the corresponding part has an area with uneven temperature distribution. In this case, if the part is coated, the coated part is more likely to deform due to thermal expansion and contraction under the influence of temperature, thereby affecting the image recognition results. In other words, areas with uneven temperature distribution may increase the roundness error of the part due to thermal expansion and contraction.

[0057] Whether a part contains a coating is known before processing. Therefore, when determining whether a part contains a coating, the presence or absence of the coating on the part can be directly input manually, which can save time and improve detection efficiency.

[0058] In step S3, if the part image temperature and the coating condition of the corresponding part meet the requirements, the roundness error of the part image is calculated based on the corrected edge of the part image. The corrected edge is the edge obtained by correcting the initial edge for expansion or contraction due to thermal expansion and contraction. This corrected edge prevents uneven deformation of the part due to uneven temperature distribution, which in turn leads to increased roundness error.

[0059] like Figure 2 As shown, the acquisition of the corrected edge includes the following steps:

[0060] S31. Calculate the correction amount of each point on the initial edge.

[0061] In this step, the initial edge refers to the edge of the part's outer contour, which can reflect the part's outline shape. The initial edge can be obtained using the Sobel edge detection algorithm, which has advantages such as simplicity, ease of implementation, and robustness to noise. Alternatively, the Canny edge detection algorithm can be used, which offers advantages such as high accuracy and robustness to noise. Both the Sobel and Canny edge detection algorithms are prior art and will not be described in detail in this invention.

[0062] The magnitude of the correction amount is the product of the thermal expansion coefficient of the part and the absolute value of the temperature difference of the local area where the corresponding point on the initial edge is located. The temperature difference is the difference between the average temperature in the corresponding local area and the standard temperature. The direction of the correction amount is the perpendicular direction of the tangent of the corresponding point on the initial edge. This is because the thermal expansion coefficient represents the change in the length of an object under a unit temperature change. Therefore, due to the temperature change, each point on the initial edge will cause the corresponding point to expand toward the outside of the initial edge or contract toward the inside. Specifically, when the temperature of the local area is greater than the standard temperature, the corresponding point will expand toward the outside of the initial edge due to heat. When the temperature of the local area is lower than the standard temperature, the corresponding point will contract toward the inside of the initial edge due to cooling. The displacement of the corresponding point to expand outward or contract inward is the correction amount corresponding to the point.

[0063] Specifically, no. The size of the correction amount for the points on the initial edge of the local area for:

[0064] ;

[0065] Where, is the thermal expansion coefficient of the material of the part, For the The absolute value of the difference between the mean temperature of each local area and the standard temperature.

[0066] In step S31 , the correction amount of each point on the initial edge is obtained according to the thermal expansion coefficient of the part material and the temperature difference of the corresponding local area, and the initial edge can be corrected by the correction amount.

[0067] S32 , connecting points obtained by moving each point on the initial edge along the direction of the correction amount by a corresponding correction amount in sequence to form the corrected edge.

[0068] In this step, when the correction amount is positive, the corresponding point on the initial edge moves inward along the perpendicular direction of its tangent toward the inside of the initial edge. When the correction amount is negative, the corresponding point on the initial edge moves outward along the perpendicular direction of its tangent toward the outside of the initial edge. This is because a positive correction amount indicates that the temperature difference in the local area is positive, that is, the temperature is increasing, and the points on the initial edge will expand outward due to heat, so it is necessary to move the corresponding points inward by the corresponding correction amount for correction. A negative correction amount indicates that the temperature difference in the local area is negative, that is, the temperature is decreasing, and the points on the initial edge will shrink inward due to cooling, so it is necessary to move the corresponding points outward by the corresponding correction amount for correction.

[0069] Specifically, a rectangular coordinate system is established on the part image to obtain the coordinates of each point on the initial edge. Based on the magnitude and direction of the correction, its components on the horizontal and vertical axes of the coordinate system are calculated. Therefore, the horizontal coordinate of each point on the corrected edge is the horizontal coordinate of the corresponding point on the initial edge plus the component of the corresponding correction on the horizontal axis. The vertical coordinate of each point on the corrected edge is the vertical coordinate of the corresponding point on the initial edge plus the component of the corresponding correction on the vertical axis.

[0070] In steps S31 and S32, corrections are calculated for each point on the initial edge based on the coefficient of thermal expansion and the temperature difference in the local area. Then, corrections are made to each point on the initial edge based on the magnitude and direction of the corrections to obtain the corresponding corrected edge. Because the corrected edge compensates for the displacement of each point on the initial edge due to thermal expansion and contraction, it prevents increased roundness error due to uneven temperature distribution within the part, making the roundness error obtained from the corrected edge more accurate and reasonable.

[0071] After obtaining the corrected edge of the part image in step S3, the process further includes step S33:

[0072] The roundness error of the corresponding part image is calculated based on the corrected edge.

[0073] The calculation of roundness error is a prior art. The present invention provides a method for calculating roundness error based on the least squares method. The method evaluates the deviation between the corrected edge and the ideal circle by fitting an ideal circle. Step S33 includes the following steps:

[0074] S331. Construct an error function, where the error function is the sum of squares of radial deviations between all points on the corrected edge and the ideal circle; the radial deviation is the difference between the distance between the corresponding point on the corrected edge and the assumed center of the ideal circle and the assumed radius of the ideal circle.

[0075] Therefore, the correction on the edge The radial deviation of a point from the ideal circle for:

[0076] ;

[0077] Where, To correct the edge The horizontal coordinate of the point, To correct the edge The vertical coordinate of the point, is the abscissa of the hypothetical center of the ideal circle, is the ordinate of the hypothetical center of the ideal circle, is the assumed radius of the ideal circle.

[0078] Error function for:

[0079] ;

[0080] Where, To correct the edge The radial deviation of a point from the ideal circle, is the total number of points on the corrected edge.

[0081] S332. Minimize the error function to obtain the true center and true radius of the ideal circle.

[0082] In order to facilitate the solution, , , then the error function for:

[0083] .

[0084] For the parameters in the error function above 、 、 Taking partial derivatives and setting the derivative to 0, we can get three equations. Solving the equations, we can get 、 、 The value of , and then the true center of the ideal circle is ( ), the true radius of the ideal circle is .

[0085] S333. Calculate the radial deviation of each point on the corrected edge based on the true center and true radius of the ideal circle.

[0086] Substituting the true center and true radius of the ideal circle into the radial deviation formula in step S331, the radial deviation of each point on the corrected edge can be obtained.

[0087] S334, taking the difference between the maximum and minimum radial deviations of each point on the corrected edge as the roundness error of the corresponding part image .

[0088] Therefore, the roundness error of the part image for:

[0089] ;

[0090] Where, To correct the maximum value of the radial deviation of each point on the edge, The minimum value of the radial deviation of each point on the corrected edge.

[0091] In step S33, the least square method is used to calculate the roundness error, which has the advantages of high accuracy, strong robustness, and simple calculation.

[0092] In some other embodiments, the minimum circumscribed circle method may be used to calculate the roundness error.

[0093] In the present invention, when the temperature and coating factors are not simultaneously met, the initial edge in the part image can be directly used to calculate the roundness error of the part, which can simplify the calculation steps and improve the detection efficiency. The calculation method of the roundness error is the same as the calculation method in step S33 and will not be repeated here.

[0094] S4. When the roundness error of the part image is greater than the error threshold, it is determined that the corresponding part has a defect.

[0095] In this step, if the roundness error of a part exceeds the error threshold, it indicates that the roundness of the corresponding part is poor and the corresponding part is defective. If a part is determined to be defective, an alarm or other measures can be taken to promptly remind staff to conduct inspections and prevent the release of unqualified products.

[0096] In the intelligent inspection method for parts processed by CNC machine tools, the roundness error of the corresponding part is calculated using the corrected edge of the part image when both the part's temperature and coating conditions meet the requirements. Because the corrected edge is the initial edge plus the corresponding correction amount, displacement of each point on the initial edge due to thermal expansion and contraction is corrected, the calculated roundness error of the part is more reasonable and accurate, preventing misjudgments.

[0097] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. An intelligent detection method for parts processed by CNC machine tools, characterized in that: The method comprises the following steps: obtaining an infrared image of a part image; Dividing the part image into a plurality of local areas having the same temperature range according to the infrared image; If the part contains a coating and the temperature of one or more local areas is greater than or less than the standard temperature, the initial edge of the one or more local areas is corrected to obtain a corrected edge; Calculate the roundness error of the part image based on the corrected edge of the part; when the roundness error of the part image is greater than the error threshold, it is determined that the corresponding part has a defect; The acquisition of the corrected edge comprises the following steps: Calculate the correction amount for each point on the initial edge; the correction amount is the product of the thermal expansion coefficient of the part and the absolute value of the temperature difference in the local area where the corresponding point on the initial edge is located; the temperature difference is the difference between the average temperature in the corresponding local area and the standard temperature; the direction of the correction amount is the perpendicular direction of the tangent line of the corresponding point on the initial edge; The points on the initial edge are moved in the direction of the correction amount by a corresponding correction amount and connected in sequence to form the corrected edge.

2. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The local area with the same temperature range is an area in which the temperature difference between positions in the part image does not exceed the temperature threshold, and the temperature gradient at each position does not exceed the temperature gradient threshold.

3. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: When the correction amount is a positive value, the corresponding point on the initial edge moves toward the inside of the initial edge along the perpendicular direction of its tangent; when the correction amount is a negative value, the corresponding point on the initial edge moves toward the outside of the initial edge along the perpendicular direction of its tangent.

4. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The infrared image of the part image is obtained using an infrared thermal imager.

5. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The initial edge of the part image is obtained by using a Sobel edge detection algorithm or a Canny edge detection algorithm.

6. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The standard temperature is the average temperature of each component under room temperature conditions.

7. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The roundness error of the part image is calculated using the least squares method.

8. The intelligent detection method for parts processed by CNC machine tools according to claim 7, characterized in that: The roundness error of the part image is calculated using the least squares method, which includes the following steps: Constructing an error function, where the error function is the sum of the squares of the radial deviations of all points on the corrected edge from the ideal circle; the radial deviation is the difference between the distance between the corresponding point on the corrected edge and the assumed center of the ideal circle and the assumed radius of the ideal circle; Minimize the error function to obtain the true center and true radius of the ideal circle; Calculate the radial deviation of each point on the corrected edge based on the true center and true radius of the ideal circle; The difference between the maximum and minimum radial deviations of each point on the corrected edge is taken as the roundness error of the corresponding part image.

9. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The part image is an image that has been filtered and denoised, wherein the filtering method is one of Gaussian filtering, median filtering and mean filtering.

10. The intelligent detection method for parts processed by CNC machine tools according to claim 1, characterized in that: The shape of the part is circular or ring-shaped.

Images