Engine cylinder cover defect detection method and system based on machine vision

Abstract

The invention relates to the technical field of engine cylinder cover detection, in particular to an engine cylinder cover defect detection method and system based on machine vision, and the method comprises the steps: 1, calculating a sample fluctuation value in a sample data set, and determining a stable signal; 2, acquiring a sample product, performing comprehensive defect detection on the sample product by utilizing detection equipment, determining a defect occurrence rate, determining an inertial defect based on the defect occurrence rate, identifying the position of the inertial defect, and obtaining a key detection area; step 3, comparing pixel points in the background image and the comprehensive image to obtain a determined position of the to-be-detected product, acquiring an inertial defect and a corresponding key detection area, and performing position retrieval and image acquisition by detection equipment to obtain a real-time detection image; 4, performing defect analysis on the real-time detection image; through the processing process, the key part of the product can be quickly and accurately focused, unnecessary detection is reduced, and the detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of engine cylinder head detection, and particularly to a method and system for detecting defects of an engine cylinder head based on machine vision. Background Art

[0002] The cylinder head is installed on top of the cylinder block, sealing the cylinder from above and forming a combustion chamber. The cylinder head of a water-cooled engine is internally provided with a cooling water jacket, and the cooling water holes on the lower end face of the cylinder head communicate with the cooling water holes of the cylinder block, using circulating water to cool high-temperature parts such as the combustion chamber.

[0003] The prior art CN115855953A discloses a three-dimensional defect detection method and device for an engine cylinder head, including: Step 1: Grabbing the engine cylinder head to be detected by a first industrial robot and sending it to a designated detection area; Step 2: Cooperating the first industrial robot with a second industrial robot, and using a three-dimensional camera on the second industrial robot to obtain three-dimensional images of six faces, namely 1, 2, 3, 4, A, and B, on the engine cylinder to be detected; Step 3: Performing point cloud registration on the obtained three-dimensional images of the engine cylinder head; Step 4: Performing three-dimensional modeling on the detected point cloud registration data and comparing the shape difference with the three-dimensional model of the standard engine cylinder head; Step 5: Comparing the obtained shape difference with a set value; However, during the production process of the engine cylinder head, when comprehensively detecting the production products in real time, during the detection process, the analysis and positioning of defects are not accurate and efficient enough, lacking effective identification and targeted detection of inertial defects, and unable to quickly and accurately determine the key detection areas, resulting in high detection costs and poor effects. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems in the background art, and a method and system for detecting defects of an engine cylinder head based on machine vision are proposed.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: A method for detecting defects of an engine cylinder head based on machine vision, the method specifically includes the following steps: Step 1: Collect real-time electrical data of the production equipment, set a sample data group, calculate the sample fluctuation values in the sample data group, and determine a stable signal according to the sample fluctuation values; Step 2: When a stable signal is detected, switch the production equipment to the production state, obtain a sample product, perform a comprehensive defect detection on the sample product using a detection device to obtain a defect image, and classify the defect image to obtain a defect set; Count and calculate the number of images in the defect set to obtain the defect incidence rate. Based on the defect incidence rate, determine the inertial defects, and then identify the positions of the inertial defects to obtain the key detection areas; Step 3: Transmit the product to be inspected to the area to be inspected, and collect comprehensive images. Based on the comparison of pixel points in the background image and the comprehensive image, obtain the determined position of the product to be inspected; Obtain the inertial defects and the corresponding key detection areas, input the regional position features of the key detection areas into the detection device, and the detection device performs position retrieval and image acquisition to obtain real-time detection images; Step 4: Obtain the real-time detection images, use the convolutional neural network algorithm to analyze the defects in the real-time detection images, and then complete the image defect detection of the product.

[0006] As a further solution of the present invention, the method for determining the sample fluctuation value includes: S1: Obtain the production equipment of the engine cylinder head, and detect the start signal of the production equipment. When the start signal is detected, based on a fixed time, collect the real-time electrical data of the production equipment, and mark the real-time electrical data as Ei, where i represents the serial number of different time periods. The real-time electrical data refers to the data collected according to the key parameters of the operation of the production equipment; S2: Obtain the real-time electrical data Ei, set consecutive n data as a group of data, and at the same time mark this group of data as the sample data group. Extract the real-time electrical data Ei in the sample data group, and use the formula to obtain the sample fluctuation value Eb, and Ep is the sample average value in the sample data group.

[0007] As a further solution of the present invention, the method for determining the stable signal includes: Compare the sample fluctuation value Eb with the fluctuation reference value X1. If Eb < X1, it means that the real-time electrical data Ei in the sample data group is stable data, and at this time, a stable signal is generated. On the contrary, when Eb ≥ X1, it means that the production equipment is in an unstable state at this time. Starting from the (n + 1)th data, continue to collect the real-time electrical data, re-mark the consecutive n real-time electrical data collected as the sample data group, and calculate and process the sample fluctuation value again until a stable signal is generated.

[0008] As a further solution of the present invention, when calculating the sample fluctuation value Eb, the sample average value Ep is within the normal operation range of the key parameters. If the sample average value Ep is not within the normal operation range of the key parameters, directly collect the real-time electrical data of the next sample data group at this time, and calculate and process the sample fluctuation value Eb again according to the above method.

[0009] As a further solution of the present invention, the method for obtaining defective images includes: When a stable signal is detected, switch the production equipment to the production state, produce the engine cylinder head, take continuous m products at the end of the production line, and mark them as sample products. Then, use the detection equipment to conduct a comprehensive defect detection on all sample products; Obtain the defects existing in each sample product, identify the positions of each defect in the sample product, and simultaneously obtain defective images; Use image processing technology to analyze the defect types in the defective images, and classify them according to the defect types to obtain several defect sets. Among them, one defect set corresponds to one defect type.

[0010] As a further solution of the present invention, the method for obtaining inertial defects includes: Arbitrarily select a defect set, count the number L of defective images in this defect set, divide the number L by the number m of products to obtain the defect incidence rate Fq, compare the defect incidence rate Fq with the occurrence threshold Fy. If Fq < Fy, mark the corresponding defect type as accidental defect. On the contrary, if Fq ≥ Fy, mark the corresponding defect type as inertial defect; Obtain the defect set corresponding to the inertial defect. According to the defective images in this defect set, identify the positions of the inertial defects, and mark the positions where the inertial defects are located as key detection areas. At the same time, record the regional position characteristics of each key detection area, where the regional position characteristics include the regional area and regional size of the key detection area and the specific position of the key detection area in the produced products.

[0011] As a further solution of the present invention, the method for obtaining real-time detection images includes: Set up an industrial detection area, and use the detection equipment to collect the background image of the area to be inspected. Among them, a detection equipment and an area to be inspected are set in the industrial detection area, and the detection equipment in the industrial detection area is set as a machine vision detection equipment; Obtain the terminal production products on the production line at this time, and mark them as products to be inspected. Transmit the products to be inspected to the area to be inspected in the industrial detection area in sequence, and collect comprehensive images; Based on the background image and the comprehensive image, determine the determined position of the product to be inspected in the area to be inspected, use the detection equipment to collect the three-dimensional visual image of the product to be inspected, then extract the regional position characteristics of the key detection area, and transmit the regional position characteristics as input data to the detection equipment. The detection equipment performs position retrieval in the three-dimensional visual image according to the input regional position characteristics; After the position retrieval is completed, determine the position of the key detection area on the product to be inspected, and then according to the determined position, collect the image of the determined position in real time to obtain real-time detection images As a further solution of the present invention, the method for determining the definite position of the product to be inspected in the inspection area to be inspected includes: Collect the overall image of the blank inspection area to be inspected and mark it as the background image. Herein, the blank inspection area to be inspected refers to the inspection area where there is no product to be inspected. When a product to be inspected is detected in the inspection area to be inspected, collect the overall image of the inspection area to be inspected again and mark it as the comprehensive image; Set a plane coordinate system in the background image and the comprehensive image respectively. Then, obtain the pixel points in the background image and the comprehensive image respectively, and based on the plane coordinate system, set the pixel points as coordinate positions. Then, obtain the pixel points with the same position coordinates respectively, and compare the pixel points with the same position coordinates. If the pixel values of the pixel points are the same, mark this pixel point as a background point in the comprehensive image. On the contrary, if the pixel values of the pixel points are different, mark this pixel point as an inspection point in the comprehensive image. Traverse the pixel points in the background image and the comprehensive image. At this time, divide the pixel points in the comprehensive image into background points and inspection points; Among them, the plane coordinate system in the background image and the plane coordinate system in the comprehensive image are the same plane coordinate system.

[0012] An engine cylinder head defect detection system based on machine vision includes: A data acquisition module for collecting the real-time electrical data of the production equipment; A signal detection module for setting a sample data group according to the real-time electrical data of the production equipment, calculating the sample fluctuation values in the sample data group, and determining a stable signal based on the sample fluctuation values; A sample detection module for receiving the stable signal. When the stable signal is received, switch the production equipment to the production state, then obtain a sample product, and then use the detection equipment to conduct a comprehensive defect detection on the sample product to obtain a defect image; Classify the defect images to obtain a defect set, count and calculate the number of images in the defect set to obtain a defect incidence rate, and determine the inertial defects and the corresponding key detection areas based on the defect incidence rate; An industrial detection module for transporting the product to be inspected into the inspection area to be inspected. When a product to be inspected is detected in the inspection area to be inspected, determine the definite position of the product to be inspected, then obtain the inertial defects and the corresponding key detection areas, input the regional position characteristics of the key detection areas into the detection equipment, and the detection equipment conducts position retrieval and image acquisition to obtain a real-time detection image; A defect analysis module for obtaining the real-time detection image, using an image analysis method to conduct defect analysis on the real-time detection image, and thus completing the image defect detection of the product.

[0013] Compared with the existing technology, the advantages of the present invention are: The present invention determines a stable signal by collecting real-time electrical data of production equipment and calculating the sample fluctuation value, ensuring production and detection are carried out under the stable state of the equipment, effectively avoiding the interference of unstable factors of the equipment on product quality. Then, sample products are obtained under the production state for comprehensive defect detection, the defect images are classified to obtain a defect set, the defect incidence rate is calculated and inertial defects are determined, which can accurately locate the problems that repeatedly occur in production, thereby improving the pertinence and accuracy of defect detection; The present invention determines the position of the product to be inspected by comparing the pixel points of the background image and the comprehensive image, and at the same time determines the key detection area according to the inertial defects, can quickly and accurately focus on the key parts of the product, reduce unnecessary detections, improve the detection efficiency, and the detection equipment collects images according to the key detection area, providing accurate image data for subsequent defect analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Schematic structural diagram of the method flow of the present invention; Figure 2 It is a schematic structural diagram of the system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0016] Referring to Figure 1-2 , a method for detecting defects of an engine cylinder head based on machine vision, the method specifically includes the following steps: Step 1: Obtain the real-time electrical data of the production equipment, perform data analysis on the real-time electrical data, determine the stable operation state of the production equipment, and generate a stable signal according to the stable operation state. The specific method for generating the stable signal includes: S1: Obtain the production equipment of the engine cylinder head, and detect the start signal of the production equipment. When the start signal is detected, based on a fixed time, collect the real-time electrical data of the production equipment, and mark the real-time electrical data as Ei, where i represents the serial number of different time periods. The real-time electrical data refers to the data collected according to the key parameters of the production equipment operation. For example, for machine tool equipment, set the spindle speed as the key parameter for real-time electrical data collection, and for hydraulic drive equipment, set the hydraulic pressure as the key parameter for real-time electrical data collection; Further, the fixed time is a threshold, and the specific value of the fixed time is set by those skilled in the art according to big data experience; S2: Obtain the real-time electrical data Ei. Set consecutive n data as a group of data, and at the same time mark this group of data as a sample data group. Extract the real-time electrical data Ei in the sample data group, and use the formula to obtain the sample fluctuation value Eb, where Ep is the sample average value in the sample data group; Compare the sample fluctuation value Eb with the fluctuation reference value X1. If Eb < X1, it means that the real-time electrical data Ei in the sample data group is stable data, and at this time, a stable signal is generated. On the contrary, when Eb ≥ X1, it means that the production equipment is in an unstable state at this time. Starting from the (n + 1)-th data, continue to collect the real-time electrical data, re-mark the consecutive n real-time electrical data collected as a sample data group, and process it again according to the above method until a stable signal is generated; In another embodiment of the present invention, when calculating the sample fluctuation value Eb, if the sample average value Ep is within the normal operating range of the key parameters, and if the sample average value Ep is not within the normal operating range of the key parameters, directly collect the real-time electrical data of the next sample data group at this time, and calculate and process the sample fluctuation value Eb again according to the above method. Further, the normal operating range of the key parameters is determined by those skilled in the art according to the parameters of the actual production equipment; Step 2: Based on the stable signal, first select a production sample. According to the selected sample product, determine the inertial defects in the current production process. The specific method for determining the inertial defects includes: SS1: When a stable signal is detected, switch the production equipment to the production state and produce the engine cylinder head. Take consecutive m products at the end of the production line and mark them as sample products. Then use the detection equipment to conduct a comprehensive defect detection on all sample products. In this embodiment, the detection equipment is set as a laser detection equipment, and the specific value of m is set by those skilled in the art according to big data experience; Then obtain the defects existing in each sample product, identify the positions of each defect in the sample product, and at the same time obtain the defect images; Use image processing technology to analyze the defect types in the defect images, and classify them according to the defect types to obtain several defect sets. Among them, one defect set corresponds to one defect type. In this embodiment, the convolutional neural network algorithm is selected for the image processing technology, and the specific processing process is the prior art and will not be elaborated here; SS2: Arbitrarily select a set of defects. Taking this set of defects as an example, first count the number L of defective images in this set of defects, divide the number L by the number m of products to obtain the defect occurrence rate Fq, and compare the defect occurrence rate Fq with the occurrence threshold Fy. If Fq < Fy, mark the corresponding defect type as accidental defect; conversely, if Fq ≥ Fy, mark the corresponding defect type as inertial defect. Among them, the specific value of the occurrence threshold Fy is obtained by those skilled in the art based on big data operations; Then process all the remaining defect sets according to the method in the above step SS2 to determine all inertial defects; Obtain the set of defects corresponding to the inertial defects. Based on the defective images in this set of defects, identify the positions of the inertial defects, and mark the positions where the inertial defects are located as key detection areas. At the same time, record the regional position characteristics of each key detection area. Among them, the regional position characteristics include the area and size of the key detection area and the specific position of the key detection area in the produced product. For example, set the area contour line 30mm away from the product edge line as the key detection area, and then select the key detection area according to the area and size. Further, the specific values of the area and size are set by those skilled in the art based on the size of the inertial defects and combined with big data experience; Step three: Set up an industrial detection area. Among them, a detection device and a to-be-inspected area are set in the industrial detection area. In this embodiment, the detection device in the industrial detection area is set as a machine vision detection device; Obtain the terminal produced products on the production line at this time and mark them as to-be-inspected products. Transmit the to-be-inspected products to the to-be-inspected area in the industrial detection area in sequence. Then obtain the key detection areas, and determine the acquisition positions of the detection device according to the positions of the key detection areas on the to-be-inspected products. The specific method for determining the acquisition positions includes: ST1: First collect the overall image of the blank to-be-inspected area and mark it as the background image. Among them, the blank to-be-inspected area refers to the to-be-inspected area without to-be-inspected products. When a to-be-inspected product is detected in the to-be-inspected area, collect the overall image of the to-be-inspected area again and mark it as the comprehensive image; Set up a plane coordinate system in the background image and the comprehensive image respectively. Then obtain the pixel points in the background image and the comprehensive image respectively, and set the pixel points as coordinate positions based on the plane coordinate system. Then obtain the pixel points with the same position coordinates respectively, and compare the pixel points with the same position coordinates. If the pixel values of the pixel points are the same, mark this pixel point as a background point in the comprehensive image; conversely, if the pixel values of the pixel points are different, mark this pixel point as an inspection point in the comprehensive image. Traverse the pixel points in the background image and the comprehensive image. At this time, divide the pixel points in the comprehensive image into background points and inspection points; It should be further noted that the plane coordinate system in the background image and the plane coordinate system in the composite image are the same plane coordinate system, that is, the origin positions and unit lengths of the plane coordinate system in the background image and the plane coordinate system in the composite image are the same; ST2: Obtain the inspection points in the composite image, recombine the inspection points to obtain the independent image of the product to be inspected and the determined position of the product to be inspected in the inspection area; Obtain the key detection area and its area position characteristics. At the same time, use the detection device to collect the three-dimensional visual image of the product to be inspected, then extract the area position characteristics of the key detection area, and transmit the area position characteristics as input data to the detection device. The detection device performs position retrieval in the three-dimensional visual image according to the input area position characteristics; After the position retrieval is completed, determine the position of the key detection area on the product to be inspected, and then according to the determined position, collect the image of the determined position in real time to obtain the real-time detection image; Step 4: Obtain the real-time detection image, use the image analysis method to analyze the defects in the real-time detection image, and then complete the image defect detection of the product. Among them, the image analysis method in this embodiment is set as the convolutional neural network algorithm, and the specific algorithm processing process belongs to the prior art and will not be elaborated here; An engine cylinder head defect detection system based on machine vision, which includes: A data acquisition module, which is used to acquire the real-time electrical data of the production equipment and transmit it to the signal detection module; A signal detection module, which is used to set continuous n data as a sample data group according to the real-time electrical data of the production equipment, then calculate the sample fluctuation value in the sample data group, and determine the stable signal based on the sample fluctuation value. Then the signal detection module transmits the stable signal to the sample detection module; A sample detection module, which is used to receive the stable signal. When the stable signal is received, switch the production equipment to the production state, then obtain m consecutive products at the production line terminal and mark them as sample products, and then use the detection device to perform a comprehensive defect detection on the sample products to obtain defect images, classify the defect images to obtain a defect set, then count and calculate the number of images in the defect set to obtain the defect incidence rate, determine the inertial defect based on the defect incidence rate, and then identify the position of the inertial defect to obtain the key detection area. The sample detection module transmits the inertial defect and the key detection area to the industrial detection module; Industrial detection module, which is used to mark the products produced in real time as products to be inspected, transmit the products to be inspected to the area to be inspected. When the area to be inspected detects a product to be inspected, first compare the pixel points in the background image and the comprehensive image to obtain the determined position of the product to be inspected, then obtain the inertial defects and the corresponding key detection areas, input the regional position features of the key detection areas into the detection device, and the detection device performs position retrieval and image acquisition to obtain a real-time detection image. After that, the industrial detection module transmits the real-time detection image to the defect analysis module; Defect analysis module, which is used to obtain the real-time detection image, perform defect analysis on the real-time detection image by using an image analysis method, and thus complete the image defect detection of the product.

[0017] As mentioned above, the above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for detecting defects in an engine cylinder head based on machine vision, characterized in that, The method specifically includes the following steps: Step 1: Collect the real-time electrical data of the production equipment, set a sample data group, calculate the sample fluctuation values in the sample data group, and determine the stable signal based on the sample fluctuation values. Step 2: When the stable signal is detected, switch the production equipment to the production state, obtain sample products, comprehensively detect the sample products using the detection equipment to obtain defect images, classify the defect images, and obtain a defect set. Statistically calculate the number of images in the defect set to obtain the defect incidence rate. Based on the defect incidence rate, determine the inertial defects, and then identify the positions of the inertial defects to obtain the key detection areas. Step 3: Transmit the product to be inspected to the inspection area, collect comprehensive images, and obtain the determined position of the product to be inspected based on the comparison of pixel points in the background image and the comprehensive image. Obtain the inertial defects and the corresponding key detection areas, input the regional position features of the key detection areas into the detection equipment, and the detection equipment performs position retrieval and image collection to obtain real-time detection images. Step 4: Obtain the real-time detection images, use the convolutional neural network algorithm to analyze the defects in the real-time detection images, and thus complete the image defect detection of the products.

2. The method for detecting defects of an engine cylinder head based on machine vision according to claim 1, wherein The method for determining the sample fluctuation values includes: S1: Obtain the production equipment of the engine cylinder head, detect the start signal of the production equipment. When the start signal is detected, collect the real-time electrical data of the production equipment based on a fixed time, and mark the real-time electrical data as Ei, where i represents the serial number of different time periods. The real-time electrical data refers to the data collected according to the key parameters of the production equipment operation. S2: Obtain the real-time electrical data Ei. Set every consecutive n data as a group of data, and at the same time label this group of data as a sample data group. Extract the real-time electrical data Ei in the sample data group, and use the formula to obtain the sample fluctuation value Eb, where Ep is the sample average value in the sample data group.

3. The method and system for detecting defects of an engine cylinder head based on machine vision according to claim 2, wherein, The method for determining the stable signal includes: Compare the sample fluctuation value Eb with the fluctuation reference value X1. If Eb < X1, it means that the real-time electrical data Ei in the sample data group is stable data, and at this time, a stable signal is generated. On the contrary, when Eb ≥ X1, it means that the production equipment is in an unstable state at this time. Starting from the (n + 1)-th data, continue to collect the real-time electrical data, re-mark the continuously collected n real-time electrical data as the sample data group, and calculate and process the sample fluctuation value again until a stable signal is generated.

4. The method for detecting defects of an engine cylinder head based on machine vision according to claim 3, wherein When calculating the sample fluctuation value Eb, if the sample average value Ep is within the normal operating range of the key parameters, if the sample average value Ep is not within the normal operating range of the key parameters, directly collect the real-time electrical data of the next sample data group at this time, and calculate and process the sample fluctuation value Eb again according to the above method.

5. A method for detecting defects in an engine cylinder head based on machine vision according to claim 1, characterized in that, The method for obtaining the defect images includes: When the stable signal is detected, switch the production equipment to the production state, produce the engine cylinder head, take m consecutive products at the production line terminal and mark them as sample products, and then comprehensively detect all sample products using the detection equipment. Obtain the defects existing in each sample product, identify the positions of each defect in the sample product, and at the same time obtain the defect images. Using image processing technology, analyze the defect types in the defect images, and classify them according to the defect types to obtain several defect sets, where one defect set corresponds to one defect type.

6. The method for detecting defects of an engine cylinder head based on machine vision according to claim 5, characterized in that, The method for obtaining inertial defects includes: Arbitrarily select a defect set, count the number L of defect images in this defect set, divide the number L by the number m of products to obtain the defect occurrence rate Fq, compare the defect occurrence rate Fq with the occurrence threshold Fy. If Fq < Fy, mark the corresponding defect type as an accidental defect; otherwise, if Fq ≥ Fy, mark the corresponding defect type as an inertial defect. Obtain the defect set corresponding to the inertial defect. Based on the defect images in this defect set, identify the positions of the inertial defects, mark the positions where the inertial defects are located as key detection areas, and record the regional position features of each key detection area at the same time. The regional position features include the area and size of the key detection area and the specific position of the key detection area in the produced products.

7. A method for detecting defects of an engine cylinder head based on machine vision according to claim 1, characterized in that, The method for obtaining real-time detection images includes: Set up an industrial detection area, and use the detection equipment to collect the background image of the area to be inspected. Among them, a detection equipment and an area to be inspected are set in the industrial detection area, and the detection equipment in the industrial detection area is set as a machine vision detection equipment. Obtain the terminal produced products on the production line at this time and mark them as products to be inspected. Transmit the products to be inspected to the area to be inspected in the industrial detection area in sequence and collect comprehensive images. Based on the background image and the comprehensive image, determine the determined position of the product to be inspected in the area to be inspected. Use the detection equipment to collect the three-dimensional vision image of the product to be inspected, and then extract the regional position features of the key detection area. Transmit the regional position features as input data to the detection equipment, and the detection equipment performs position retrieval in the three-dimensional vision image according to the input regional position features. After the position retrieval is completed, determine the position of the key detection area on the product to be inspected, and then according to the determined position, collect the image of the determined position in real time to obtain a real-time detection image.

8. A method for detecting defects in an engine cylinder head based on machine vision according to claim 7, characterized in that, The method for determining the determined position of the product to be inspected in the area to be inspected includes: Collect the overall image of the blank area to be inspected and mark it as the background image. The blank area to be inspected refers to the area to be inspected where there is no product to be inspected. When a product to be inspected is detected in the area to be inspected, collect the overall image of the area to be inspected again and mark it as the comprehensive image. Set up a plane coordinate system in the background image and the comprehensive image respectively, then obtain the pixel points in the background image and the comprehensive image respectively, and based on the plane coordinate system, set the pixel points as coordinate positions. Then obtain the pixel points with the same position coordinates respectively, and compare the pixel points with the same position coordinates. If the pixel values of the pixel points are the same, mark this pixel point as a background point in the comprehensive image; otherwise, if the pixel values of the pixel points are different, mark this pixel point as an inspection point in the comprehensive image. Traverse the pixel points in the background image and the comprehensive image. At this time, the pixel points in the comprehensive image are divided into background points and inspection points. Among them, the plane coordinate system in the background image and the plane coordinate system in the comprehensive image are the same plane coordinate system.

9. An engine cylinder head defect detection system based on machine vision, the defect detection system adopting the engine cylinder head defect detection method described in any one of the above-mentioned claims 1-8, characterized in that, Include: A data acquisition module for collecting real-time electrical data of production equipment; A signal detection module for setting a sample data group according to the real-time electrical data of production equipment, calculating the sample fluctuation values in the sample data group, and determining a stable signal based on the sample fluctuation values; A sample detection module for receiving the stable signal, switching the production equipment to the production state when the stable signal is received, then obtaining sample products, and then using detection equipment to conduct a comprehensive defect detection on the sample products to obtain defect images; Classify the defect images to obtain a defect set, count and calculate the number of images in the defect set to obtain a defect incidence rate, and determine inertial defects and corresponding key detection areas based on the defect incidence rate; An industrial detection module for transporting a product to be inspected to an area to be inspected. When the area to be inspected detects the product to be inspected, determine the determined position of the product to be inspected, then obtain the inertial defects and corresponding key detection areas, input the regional position features of the key detection areas into the detection equipment, and the detection equipment conducts position retrieval and image acquisition to obtain real-time detection images; A defect analysis module for obtaining the real-time detection images, using image analysis methods to analyze the defects in the real-time detection images, and thus completing the image defect detection of the products.

Images