A water pump casing image processing method and system for quality inspection

Abstract

The present invention relates to a water pump casing image processing method and system for quality inspection. The present invention obtains a production control parameter deviation vector matrix by comparing the production expected control parameters and production monitoring control parameters of the water pump casing; retrieves high-frequency water pump casing defects and low-frequency water pump casing defects that meet the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold from historical samples of the water pump casing; establishes a high-frequency feature extractor and a low-frequency feature extractor based on a predefined defect feature extractor calibration table; performs feature extraction on a water pump casing image that meets the conditions to obtain an image feature extraction result; and achieves the technical effect of saving inspection resources and improving inspection efficiency.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, and in particular to a water pump casing image processing method and system for quality detection. Background Art

[0002] In traditional water pump casing production processes, defect monitoring and image feature extraction solutions usually adopt a fixed process, using the same inspection process for all casing types. The inspection process cannot be dynamically adjusted according to changes in various links and factors in production. As a result, a large amount of inspection resources are wasted on inspection items with low importance, and it is easy to ignore the inspection parameters that really need to be paid attention to, resulting in a waste of inspection resources and low efficiency. Summary of the Invention

[0003] The present invention aims to solve the technical problems of waste of inspection resources and low inspection efficiency in the prior art by providing an image processing method for water pump casing for quality inspection.

[0004] The technical solution of the present invention to solve the above technical problems is as follows:

[0005] In a first aspect, the present invention provides a water pump casing image processing method for quality inspection, comprising:

[0006] Compare the expected production control parameters and production monitoring control parameters of the water pump casing to obtain the production control parameter deviation vector matrix; retrieve high-frequency water pump casing defects and low-frequency water pump casing defects that meet the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to the trigger frequency threshold from historical samples of the water pump casing; based on a predefined defect feature extractor calibration table, process the high-frequency water pump casing defects and the low-frequency water pump casing defects to obtain a high-frequency feature extractor and a low-frequency feature extractor; call the high-frequency feature extractor to perform feature extraction on all water pump casing images that meet the production control parameter deviation vector matrix, call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that meet the production control parameter deviation vector matrix, and obtain image feature extraction results.

[0007] Optionally, the production expected control parameters and the production monitoring control parameters of the water pump casing are compared to obtain a production control parameter deviation vector matrix, including:

[0008] Normalizing the production expectation control parameter and the production monitoring control parameter to obtain a production expectation control characteristic value and a production monitoring control characteristic value;

[0009] The absolute deviation values ​​of the same attributes in the same domain are calculated for the production expected control characteristic value and the production monitoring control characteristic value to obtain the first attribute deviation time series information until the Nth attribute deviation time series information, wherein the time series covers the entire production time zone, and the matrix element configuration is empty when the attribute is not set at the corresponding time.

[0010] Optionally, high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold are retrieved from historical samples of the water pump casing, where the trigger frequency threshold includes a first trigger frequency threshold and a second trigger frequency threshold, and the first trigger frequency threshold is greater than the second trigger frequency threshold, including:

[0011] Extracting a sample production control parameter deviation vector matrix from the water pump housing historical samples;

[0012] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix; if the matrix similarity is greater than or equal to a matrix similarity threshold, adding the sample to the selected water pump casing history sample; otherwise, updating the water pump casing history sample;

[0013] When the selected water pump housing historical samples are greater than or equal to the set number, the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency, wherein the set number is ≥ 10,000 and the historical samples all belong to the set time window;

[0014] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extracting defects whose attribute trigger frequency is greater than or equal to the first trigger frequency threshold and setting them as the high-frequency water pump casing defects;

[0015] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extract defects whose attribute trigger frequency is greater than or equal to the second trigger frequency threshold and less than the first trigger frequency threshold, and set them as the low-frequency water pump casing defects;

[0016] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects whose attribute trigger frequency is less than the second trigger frequency threshold are extracted and set not to participate in the detection of water pump casing defects.

[0017] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix includes:

[0018] Through the user end, configure the first attribute deviation threshold up to the Nth attribute deviation threshold;

[0019] Based on the first attribute deviation threshold up to the Nth attribute deviation threshold, combined with the historical samples of the water pump housing, the frequency of simultaneous triggering of the defect and the first attribute when only the first attribute is in a deviation state and the other attributes are in a normal state is counted, and the frequency is set as the first attribute defect support;

[0020] Until the Nth attribute defect support is obtained;

[0021] Calculate the ratio of the defect support of the first attribute to the defect support of the Nth attribute to the sum of the defect support, and obtain the weight distribution vector of the first attribute to the Nth attribute;

[0022] Based on the weight distribution vector, weighting the production control parameter deviation vector matrix and the sample production control parameter deviation vector matrix respectively to obtain a production control parameter deviation vector update matrix and a sample production control parameter deviation vector update matrix;

[0023] Calculate the matrix similarity between the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix.

[0024] The trigger frequency threshold includes a first trigger frequency threshold and a second trigger frequency threshold, and the first trigger frequency threshold is greater than the second trigger frequency threshold; the trigger frequency threshold includes a second first trigger frequency threshold and a second second trigger frequency threshold, and the second first trigger frequency threshold is greater than the second second trigger frequency threshold; the first trigger frequency threshold is greater than the second first trigger frequency threshold, including:

[0025] When the defect attribute belongs to a failure defect, the second-class first trigger frequency threshold and the second-class second trigger frequency threshold are used;

[0026] When the defect belongs to a non-failure defect, the first trigger frequency threshold and the second trigger frequency threshold are used.

[0027] Optionally, calling the low-frequency feature extractor to perform feature extraction on a set proportion water pump casing image that satisfies the production control parameter deviation vector matrix includes:

[0028] Calculate the ratio of the number of occurrence samples of the trigger frequency of the low-frequency water pump casing defect to the total number of historical samples of the water pump casing, and set it as the sampling rate threshold;

[0029] The sampling rate threshold is used as the minimum sampling rate, the low-frequency feature extractor is called, and feature extraction is performed on the set proportion of water pump casing images that meet the production control parameter deviation vector matrix.

[0030] Optionally, calling the high-frequency feature extractor to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix, calling the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix, and obtaining image feature extraction results, further comprising:

[0031] Counting the historical average computing power of the high-frequency feature extractor and the low-frequency feature extractor respectively to obtain the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction;

[0032] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, extract the first high-frequency feature extractor and the first low-frequency feature extractor whose required computing power is greater than or equal to the computing power configuration threshold, and perform cloud computing power scheduling configuration;

[0033] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, a second high-frequency feature extractor and a second low-frequency feature extractor whose required computing power is less than the computing power configuration threshold are extracted to perform edge computing power scheduling configuration.

[0034] In a second aspect, the present invention provides a water pump casing image processing system for quality inspection, comprising:

[0035] A deviation vector matrix extraction unit is used to compare the production expected control parameters and production monitoring control parameters of the water pump housing to obtain the production control parameter deviation vector matrix;

[0036] a defect type identification unit, configured to retrieve, from historical samples of water pump casings, high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold;

[0037] a defect feature processing unit, configured to process the high-frequency water pump casing defect and the low-frequency water pump casing defect based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor;

[0038] An image feature extraction unit is used to call the high-frequency feature extractor to perform feature extraction on all water pump casing images that meet the production control parameter deviation vector matrix, call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that meet the production control parameter deviation vector matrix, and obtain image feature extraction results.

[0039] By implementing the present invention, it is possible to compare the production expected control parameters and production monitoring control parameters of the water pump housing, obtain the production control parameter deviation vector matrix, and accurately measure the deviation value between the ideal production parameters and the actual production parameters;

[0040] By implementing the present invention, it is possible to retrieve high-frequency and low-frequency water pump casing defects that meet the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to the trigger frequency threshold from historical samples of water pump casings, set the trigger frequency threshold specifically according to different defects, reasonably feedback quality information, and improve the rationality and scientificity of inspection sample data;

[0041] By implementing the present invention, the high-frequency water pump casing defects and the low-frequency water pump casing defects can be processed based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor, thereby achieving targeted extraction of water pump casing defects with different occurrence frequencies, thereby saving inspection resources and improving extraction efficiency;

[0042] By implementing the present invention, it is possible to call the high-frequency feature extractor to perform feature extraction on all water pump casing images that meet the production control parameter deviation vector matrix, call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that meet the production control parameter deviation vector matrix, and obtain image feature extraction results, so as to achieve accurate extraction and feedback of various types of defect feature images according to inspection requirements under limited hardware conditions, thereby maximizing the inspection effect.

[0043] In summary, by implementing the present invention, the probability of occurrence of different shell defect types can be predicted and their importance can be evaluated through production parameters (such as casting pressure, mold temperature, and material batch), and detection resources can be dynamically adjusted to achieve the technical effect of saving inspection resources and improving inspection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic flow chart of a water pump casing image processing method for quality inspection provided by the present invention;

[0045] Figure 2 This is a structural schematic diagram of a water pump casing image processing system for quality inspection provided by the present invention.

[0046] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0047] Deviation vector matrix extraction unit 11, defect type identification unit 12, defect feature processing unit 13, image feature extraction unit 14. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0050] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0051] Example 1, as Figure 1 As shown, an embodiment of the present invention provides a water pump casing image processing method for quality inspection, comprising:

[0052] S100: Compare the production expected control parameters and production monitoring control parameters of the water pump casing to obtain a production control parameter deviation vector matrix;

[0053] S200: Retrieving high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold from historical samples of the water pump casing;

[0054] S300: Processing the high-frequency water pump housing defect and the low-frequency water pump housing defect based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor;

[0055] S400: Call the high-frequency feature extractor to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix, call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix, and obtain image feature extraction results.

[0056] In step S100 of the embodiment of the present application, the production expected control parameters and the production monitoring control parameters of the water pump housing are compared to obtain a production control parameter deviation vector matrix, including:

[0057] Normalizing the production expectation control parameter and the production monitoring control parameter to obtain a production expectation control characteristic value and a production monitoring control characteristic value;

[0058] The absolute deviation values ​​of the same attributes in the same domain are calculated for the production expected control characteristic value and the production monitoring control characteristic value to obtain the first attribute deviation time series information until the Nth attribute deviation time series information, wherein the time series covers the entire production time zone, and the matrix element configuration is empty when the attribute is not set at the corresponding time.

[0059] In the embodiment of the present application, the purpose of establishing a production control parameter deviation vector matrix is ​​to use it as a basis for screening and selecting historical samples of water pump casings, and to retrieve and classify high-frequency water pump casing defects and low-frequency water pump casing defects. Specifically, to establish a production control parameter deviation vector matrix, it is first necessary to calculate the actual deviation between the production expectation control parameters and the production monitoring control parameters. Among them, the production expectation control parameters refer to the preset parameters in the water pump casing production process, that is, the parameters expected to be achieved. However, in the actual production process, due to various reasons, the predetermined production expectation control parameters are often not achieved. Therefore, it is necessary to monitor the actual parameters in the trial production process as production monitoring control parameters.

[0060] Exemplarily, the desired production control parameters and production monitoring control parameters include, but are not limited to, casting pressure, mold temperature, cooling rate, etc. For example, if the desired production control parameters set the casting pressure at a certain time domain to 60 MPa, the mold temperature to 230°C, and the cooling rate to 70°C / h, while the production monitoring control parameters show the actual casting pressure to be 62 MPa, the mold temperature to be 227°C, and the cooling rate to be 60°C / h,

[0061] In the embodiment of the present application, in order to facilitate calculation, it is necessary to normalize the production expectation control parameters and the production monitoring control parameters respectively. Exemplarily, the maximum-minimum value normalization method can be adopted, that is, the normalized value = (parameter to be normalized - parameter upper limit) / (parameter upper limit - parameter lower limit). Among them, the parameter upper limit and the parameter lower limit are the maximum and minimum values ​​that the parameter to be normalized can reach under normal circumstances, which can be summarized from all production and processing data within a historical time (such as 3 months). If the range of all production and processing data of the casting pressure within the historical time is 50-70MPa, the production expectation control value is 60MPa, and the production monitoring control value is 62MPa, then the normalized value of the production expectation control value is (60-50) / (70-50)=0.5, and the normalized value of the production monitoring control value is (62-50) / (70-50)=0.6. This method can be used to normalize different production control parameters and production monitoring control parameters, such as casting pressure, mold temperature, and cooling rate. The resulting normalized values ​​are the production control characteristic values ​​and production monitoring control characteristic values. For example, if only the casting pressure, mold temperature, and cooling rate characteristics are calculated, the production control characteristic values ​​and production monitoring control characteristic values ​​can be three-dimensional feature vectors in the format of (0.6, 0.3, 0.8) and (0.4, 0.4, 0.6).

[0062] Exemplarily, the absolute value deviation of the production expected control characteristic value and the production monitoring control characteristic value can be calculated as absolute value deviation = |production expected control characteristic value - production monitoring control characteristic value|. For example, when at a certain moment, the production expected control characteristic value and the production monitoring control characteristic value corresponding to the casting pressure, mold temperature, and cooling rate are (0.6, 0.3, 0.8) and (0.4, 0.4, 0.6), respectively, the absolute deviation of the production expected control characteristic value and the production monitoring control characteristic value of the casting pressure can be |0.6-0.4|=0.2, the absolute deviation of the production expected control characteristic value and the production monitoring control characteristic value of the mold temperature can be |0.3-0.4|=0.1, and the absolute deviation of the production expected control characteristic value and the production monitoring control characteristic value of the cooling rate can be |0.8-0.6|=0.2. The corresponding production control parameter deviation vector is (0.2, 0.1, 0.2). By collecting the production control parameter deviation vectors at multiple time points, the production control parameter deviation vector matrix can be obtained. The above method can be used to calculate the deviation time series information corresponding to each parameter with the same attribute at each moment, that is, the first attribute deviation time series information to the Nth attribute deviation time series information. Among them, N represents the number of production expectation control parameters and production monitoring control parameters of different attributes. For example, if only the values ​​of casting pressure, mold temperature, and cooling rate are monitored, then N is 3. Among them, the collection frequency of time series parameter information can be determined according to actual needs. For example, it can be selected to collect production expectation control parameters and production monitoring control parameters every 10 minutes. The total parameter collection time range (that is, the time domain covered by the time series) is the production process of the entire pump casing (that is, the entire production time zone). For example, the production process can be 10 hours. If the parameter of a certain attribute, such as casting pressure, is no longer set after casting is completed in the actual production process, the position of the characteristic value corresponding to the parameter of the attribute can be vacant.

[0063] In step S200 of this embodiment, high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold are retrieved from water pump casing historical samples. The trigger frequency threshold includes a first trigger frequency threshold and a second trigger frequency threshold, and the first trigger frequency threshold is greater than the second trigger frequency threshold, including:

[0064] Extracting a sample production control parameter deviation vector matrix from the water pump housing historical samples;

[0065] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix; if the matrix similarity is greater than or equal to a matrix similarity threshold, adding the sample to the selected water pump casing history sample; otherwise, updating the water pump casing history sample;

[0066] When the selected water pump housing historical samples are greater than or equal to the set number, the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency, wherein the set number is ≥ 10,000 and the historical samples all belong to the set time window;

[0067] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extracting defects whose attribute trigger frequency is greater than or equal to the first trigger frequency threshold and setting them as the high-frequency water pump casing defects;

[0068] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extract defects whose attribute trigger frequency is greater than or equal to the second trigger frequency threshold and less than the first trigger frequency threshold, and set them as the low-frequency water pump casing defects;

[0069] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects whose attribute trigger frequency is less than the second trigger frequency threshold are extracted and set not to participate in the detection of water pump casing defects.

[0070] In this embodiment of the present application, a sample production control parameter deviation vector matrix for distinguishing high-frequency water pump casing defects from low-frequency water pump casing defects can be calculated using the method described in S100. Next, a set of desired production control parameters and production monitoring control parameters for water pump casing production over a historical period (e.g., three months) is collected, and a sample production control parameter deviation vector matrix is ​​extracted from these. Multiple sample production control parameter deviation vector matrices are calculated to serve as historical samples of water pump casings.

[0071] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix includes:

[0072] Through the user end, configure the first attribute deviation threshold up to the Nth attribute deviation threshold;

[0073] Based on the first attribute deviation threshold up to the Nth attribute deviation threshold, combined with the historical samples of the water pump housing, the frequency of simultaneous triggering of the defect and the first attribute when only the first attribute is in a deviation state and the other attributes are in a normal state is counted, and the frequency is set as the first attribute defect support;

[0074] Until the Nth attribute defect support is obtained;

[0075] Calculate the ratio of the defect support of the first attribute to the defect support of the Nth attribute to the sum of the defect support, and obtain the weight distribution vector of the first attribute to the Nth attribute;

[0076] Based on the weight distribution vector, weighting the production control parameter deviation vector matrix and the sample production control parameter deviation vector matrix respectively to obtain a production control parameter deviation vector update matrix and a sample production control parameter deviation vector update matrix;

[0077] Calculate the matrix similarity between the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix.

[0078] In the embodiment of the present application, the first attribute deviation threshold up to the Nth attribute deviation threshold are used as standards for determining whether the parameters of the attribute are in a deviation state. The threshold can be set according to actual production needs. For example, the absolute deviation value threshold of the production expected control characteristic value and the production monitoring control characteristic value of the mold temperature can be set to 0.2. When the actual deviation value is greater than 0.2, the deviation state of the attribute (mold temperature) can be determined as deviation. If the actual deviation value is less than the threshold (0.2), the deviation state of the attribute is determined as non-deviation. In the same way, the first attribute deviation threshold up to the Nth attribute deviation threshold can be determined, and whether the corresponding attribute is in a deviation state can be determined.

[0079] In an embodiment of the present application, when the first attribute is in a deviation state and the other attributes are in a non-deviation state, the frequency of simultaneous triggering of defects and the first attribute specifically refers to the frequency of ultimately producing defective pump casings only when the deviation state of the first attribute (such as mold temperature) is deviation and the deviation state of the other attributes (such as casting pressure, etc.) is non-deviation in historical data. For example, within the data of a historical time (such as 3 months), when the deviation state of the first attribute (such as mold temperature) is deviation and the deviation state of the other attributes (such as casting pressure, etc.) is non-deviation, the number of water pump casings produced is 100, and the number of water pump casings with defects (such as air holes, shrinkage, etc.) among them is 10, then the value of the simultaneous triggering frequency of the defect and the first attribute is 10 / 100=0.1=10%, and this value can be set as the defect support of the first attribute to indicate the degree of correlation between the attribute and the water pump casing defect.

[0080] By using the above method, the first attribute defect support, the second attribute defect support, and the Nth attribute defect support can be calculated until the Nth attribute defect support is obtained, where N is the number of parameters of different attributes (such as casting pressure, mold temperature, and cooling rate).

[0081] After obtaining the defect support of the first attribute up to the defect support of the Nth attribute, the distribution vectors of the first attribute up to the Nth attribute can be weighted according to the defect support to obtain the weight distribution vectors of the first attribute up to the Nth attribute. For example, if there are three attributes of casting pressure, mold temperature, and cooling rate, assuming that the defect support of the three attributes calculated by the above steps is 10%, 20%, and 20% respectively, then the weighted values ​​of the distribution vectors calculated according to the defect support of the above three attributes are 10% / (10%+20%+20%)=0.2, 20% / (10%+20%+30%)=0.4, and 20% / (10%+20%+30%)=0.4 respectively. It is not difficult to see that the sum of the weighted values ​​is 1. After obtaining the weighted values ​​of each attribute, it is necessary to weight the production control parameter deviation vector matrix and the sample production control parameter deviation vector matrix. Through the above calculation method, the weight distribution vectors of the first attribute up to the Nth attribute can be calculated. Where N is the number of parameters with different properties (such as casting pressure, mold temperature, cooling rate).

[0082] For example, assuming there are three parameters: casting pressure, mold temperature, and cooling rate, if a deviation vector in the production control parameter deviation vector matrix or the sample production control parameter deviation vector matrix is ​​(0.6, 0.3, 0.8), then the corresponding values ​​of the three parameters in the vector need to be multiplied by the weighted values ​​of the weight distribution vector calculated above. For example, when the weighted values ​​are When is 0.2, 0.4, and 0.4, the weighted vector is By weighting all vectors in the deviation vector matrix, the deviation vector update matrix can be obtained. In this way, the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix can be calculated.

[0083] Finally, it is necessary to calculate the matrix similarity between the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix. When monitoring the three parameter data of casting pressure, mold temperature, and cooling rate, the above two matrices are both matrices with 3 columns and M rows, where 3 represents the three monitoring parameters and M is the number of sampling times of the above three parameter data in the aforementioned full production time zone (such as 20 times).

[0084] The specific matrix similarity comparison method can be Dynamic Time Warping (DTW) combined with multi-dimensional parameter weighted averaging. The specific steps are to calculate the DTW distance for each parameter column (casting pressure, mold temperature, cooling rate) of the comparison matrix separately, and then combine them according to the weight. It is suitable for scenarios where the physical meaning of each parameter is independent and requires differentiated weights. For example, the DTW distance Dtotal can be defined as 0.4D1 (casting pressure DTW distance). 0.3D2 (mold temperature DTW distance) 0.3D3 (cooling rate DTW distance). The DTW distance Dtotal is then mapped to a range of 0 to 1 to obtain a similarity value. For example, similarity is defined as 1 / (1+Dtotal). The closer this similarity value is to 1, the more similar the control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix are.

[0085] Furthermore, a matrix similarity threshold, such as 0.8, can be set. When the similarity between the control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix is ​​greater than the threshold, it indicates that the two are extremely similar, and the sample production control parameter deviation vector update matrix can be used as the selected water pump casing historical sample. When it is less than the threshold, it indicates that the similarity between the two is not high, and the water pump casing historical sample needs to be updated so that more samples with similarities greater than or equal to the threshold (such as 0.8) can be selected from the updated historical samples as the selected water pump casing historical samples. When the number of selected water pump casing historical samples is greater than or equal to a set number (in the embodiment of the present application, the number can be ≥10,000), the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency, and the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency.

[0086] Q represents the number of pump casing defect types detected during the inspection of the selected historical pump casing samples. For example, if porosity, cracks, shrinkage, burrs, and potholes were detected during the inspection of the selected historical pump casing samples, the Q value would be 5. The defect attribute trigger frequency represents the ratio of the number of samples containing a particular defect in the selected historical pump casing samples to the total number of samples in the selected historical pump casing samples. For example, if the total number of selected historical pump casing samples is 10,000, and 6,000 samples contained burrs during inspection, the trigger frequency of this defect attribute (burrs) is 6,000 / 10,000 = 0.6 = 60%. Using this method, the trigger frequencies of the first defect attribute up to the Qth defect attribute trigger frequency can be calculated.

[0087] The trigger frequency threshold includes a first trigger frequency threshold and a second trigger frequency threshold, and the first trigger frequency threshold is greater than the second trigger frequency threshold;

[0088] The trigger frequency threshold includes two types of first trigger frequency thresholds and two types of second trigger frequency thresholds, and the two types of first trigger frequency thresholds are greater than the two types of second trigger frequency thresholds;

[0089] The first trigger frequency threshold of the first category is greater than the first trigger frequency threshold of the second category, including:

[0090] When the defect attribute belongs to a failure defect, the second-class first trigger frequency threshold and the second-class second trigger frequency threshold are used;

[0091] When the defect belongs to a non-failure defect, the first trigger frequency threshold and the second trigger frequency threshold are used.

[0092] In the embodiment of the present application, since different image processing strategies need to be adopted for different defect attributes, it is necessary to divide the defect attributes. Among them, failure defects refer to relatively serious defects that may cause the failure of the water pump housing, including cracks, shrinkage, etc. Non-failure defects are relatively mild defects that will not cause the failure of the water pump housing, including burrs, potholes, etc. The division of failure defects and non-failure defects can be done by first counting all Q-type defect attributes detected in the selected water pump housing historical samples, and then the technicians analyze the previous causes of water pump housing failure to evaluate the possibility of each defect causing the water pump housing to fail during use. Based on this, all Q-type defect attributes are divided into failure defects and non-failure defects, and a defect classification index table is established based on the division results. By inputting the defect name, the defect category can be output. This analysis method and classification index table establishment method are prior art and will not be repeated here.

[0093] Furthermore, for more serious failure-type defects, the second-class first trigger frequency threshold and the second-class second trigger frequency threshold should be smaller to detect these serious defects in a timely manner. For example, the second-class first trigger frequency threshold can be set to 40% and the second-class second trigger frequency threshold can be set to 30%.

[0094] For less severe, non-failure defects that won't cause pump casing failure, the Class 1 trigger frequency threshold (Class 1) and Class 1 trigger frequency threshold (Class 1) should be lower. For example, the Class 1 trigger frequency threshold (Class 1) can be set to 60%, and the Class 1 trigger frequency threshold (Class 1) can be set to 50%. This reduces unnecessary defect identification. Furthermore, the Class 1 trigger frequency threshold (Class 1) should be higher than the Class 2 trigger frequency threshold (Class 2).

[0095] Next, defects with attribute trigger frequencies greater than or equal to the first trigger frequency threshold (e.g., 40%) are extracted and set as high-frequency water pump casing defects. Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects with attribute trigger frequencies greater than or equal to the second trigger frequency threshold (e.g., 20%) and less than the first trigger frequency threshold (e.g., 40%) are extracted and set as low-frequency water pump casing defects. This facilitates the development of different inspection strategies based on low-frequency and high-frequency defects. Specifically, based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects with attribute trigger frequencies less than the second trigger frequency threshold (20%) are extracted and excluded from water pump casing defect detection.

[0096] In step S300 of the embodiment of the present application, the high-frequency water pump housing defect and the low-frequency water pump housing defect are processed based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor.

[0097] The defect feature extractor calibration table contains the feature information of all Q-type water pump casing defects determined by the above method, including whether the defect belongs to a high-frequency water pump casing defect, a low-frequency water pump casing defect, or a water pump casing defect that does not participate in the detection, and whether the defect belongs to a failure type defect or a non-failure type defect, etc. Then, according to the feature information of a specific defect, the corresponding high-frequency feature extractor or low-frequency feature extractor is matched. The feature extraction logic of the high-frequency feature extractor and the low-frequency feature extractor are different. For the defect type matched to the high-frequency feature extractor, the inspection strategy adopted is to extract the image feature information of all samples to prevent the omission of high-frequency water pump casing defects; and for the defect type matched to the low-frequency feature extractor, the inspection strategy adopted is to extract the image feature information of some samples for random inspection to save inspection resources. The specific random inspection method is described in step S400.

[0098] The high-frequency feature extractor and the low-frequency feature extractor can identify the image of the water pump casing, identify defects such as pores, cracks, shrinkage, burrs, potholes, etc. contained in the water pump casing image, and submit them to technical personnel for review.

[0099] The high-frequency and low-frequency feature extractors have identical structures, differing only in the training data used for high-frequency and low-frequency defect types. A modified U-Net architecture based on multi-scale feature fusion can be used, along with an attention mechanism and data augmentation strategies. Specifically, ResNet-50 is used as the encoder, whose residual structure effectively extracts deep features while preventing vanishing gradients. The decoder adopts the symmetric U-Net architecture. Several improved modules are implemented, including a multi-scale feature pyramid (FPN), added after the outputs of encoder stages 2-5. This module uses 1x1 convolution to unify the number of channels before upsampling and fusion, enhancing sensitivity to subtle defects such as glitches. A channel attention module (SE Block) is also introduced before each skip connection in the decoder layer to dynamically adjust feature channel weights and suppress background interference. Deformable Conv replaces the 3x3 convolution in the third stage of the backbone network to accommodate irregular shapes such as cracks and shrinkage.

[0100] The feature extractor optimizer uses AdamW, with an initial learning rate of 0.001 and weight decay of 0.01. The loss function uses Focal Loss + Dice, with α = 0.8 and γ = 2.0 to mitigate class imbalance (a high proportion of pores). The batch size is set to 16. Cosine annealing is used for learning rate scheduling, with a 20-epoch period and a minimum learning rate of 0.0001. Training is terminated if the validation set mAP does not improve after 10 consecutive epochs. The gradient clipping threshold is set to 5.0 to prevent gradient explosion.

[0101] The training data consists of images of various defects that have appeared over a period of time. Each defect image is pre-labeled with the defect type, and the number of training images for each defect is no less than 500. The data is divided into a validation set and a training set in a ratio of 2:8. Based on the aforementioned categories of high-frequency and low-frequency water pump casing defects, the data is divided into high-frequency feature extractor training data and low-frequency feature extractor training data, which are used to train the high-frequency feature extractor and the low-frequency feature extractor, respectively.

[0102] During training, pre-training is performed first, the initial learning rate is set to 0.001, and training is performed for 50 rounds; the model evaluation indicator can be the validation set mAP, calculated using the COCO standard, the IOU threshold range is [0.5:0.95], and weighted average is calculated by defect type. When the value is set to ≥0.92, the model converges and obtains the high-frequency feature extractor and the low-frequency feature extractor. By inputting the water pump housing image, the corresponding low-frequency defect features and high-frequency defect features can be extracted.

[0103] The water pump housing image includes a complete image of each surface of the water pump housing, which can be obtained through monitoring equipment at the production site. Such monitoring equipment is existing technology and will not be described in detail here.

[0104] In step S400 of the embodiment of the present application, the low-frequency feature extractor is called to perform feature extraction on the set proportion water pump housing image that satisfies the production control parameter deviation vector matrix, including:

[0105] Calculate the ratio of the number of occurrence samples of the trigger frequency of the low-frequency water pump casing defect to the total number of historical samples of the water pump casing, and set it as the sampling rate threshold;

[0106] The sampling rate threshold is used as the minimum sampling rate, the low-frequency feature extractor is called, and feature extraction is performed on the set proportion of water pump casing images that meet the production control parameter deviation vector matrix.

[0107] In one embodiment, the ratio of the number of occurrence samples (e.g., 2000) of the trigger frequency of a low-frequency water pump casing defect (e.g., surface potholes) to the total number of selected historical samples of the water pump casing (e.g., 10,000) can be calculated (e.g., 2000 / 10,000=0.2), and set as the sampling rate threshold, that is, the sampling rate for the low-frequency water pump casing defect during the sampling process is not less than 0.2 (20%) of the total production quantity. Using the trigger frequency of the low-frequency water pump casing defect as the sampling rate threshold can ensure that as many defects as possible are found while saving inspection resources during the sampling process; then it is necessary to call a low-frequency feature extractor for extracting the corresponding low-frequency feature image, and perform feature extraction on the water pump casing image that meets the set ratio of the production control parameter deviation vector matrix (the set ratio is greater than or equal to the trigger frequency of the low-frequency water pump casing defect, such as greater than or equal to 20%).

[0108] In step S400 of the embodiment of the present application, the high-frequency feature extractor is called to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix, and the low-frequency feature extractor is called to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix to obtain image feature extraction results, further comprising:

[0109] Counting the historical average computing power of the high-frequency feature extractor and the low-frequency feature extractor respectively to obtain the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction;

[0110] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, extract the first high-frequency feature extractor and the first low-frequency feature extractor whose required computing power is greater than or equal to the computing power configuration threshold, and perform cloud computing power scheduling configuration;

[0111] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, a second high-frequency feature extractor and a second low-frequency feature extractor whose required computing power is less than the computing power configuration threshold are extracted to perform edge computing power scheduling configuration.

[0112] In the embodiment of the present application, in order to improve inspection efficiency, it is also necessary to optimize the computing power configuration of the high-frequency feature extractor and the low-frequency feature extractor. The optimization method is to collect statistics on the historical computing power requirements of the high-frequency feature extractor and the low-frequency feature extractor based on historical data, and to schedule computing power based on the computing power requirements to support the operation of the high-frequency feature extractor and the low-frequency feature extractor to ensure inspection efficiency.

[0113] First, you need to obtain the average computing power consumed by the high-frequency feature extractor and the low-frequency feature extractor when performing feature extraction tasks in the historical period (such as the last 100 feature extraction tasks). Then, compare this average computing power with the computing power configuration threshold to formulate a computing power scheduling strategy accordingly.

[0114] Among them, the computing power configuration threshold is the maximum computing power that the terminal system can provide (such as 2TFLOPS), which depends on the hardware facilities of the terminal system. When the computing power required for high-frequency feature extraction and the computing power required for low-frequency feature extraction (such as 3TFLOPS) are greater than the threshold (such as 2TFLOPS), it means that the terminal system cannot provide sufficient computing power support for the computing power required for high-frequency feature extraction and low-frequency feature extraction. In order to ensure the normal progress of feature extraction, additional cloud computing power needs to be scheduled for the first high-frequency feature extractor and the first low-frequency feature extractor to support the normal operation of the first high-frequency feature extractor and the first low-frequency feature extractor. Among them, cloud computing power can be obtained through the computing power leasing service provided by a third-party cloud platform, or it can be provided by remote computing units deployed in other systems, which will not be repeated here.

[0115] When the computing power required for high-frequency feature extraction and the computing power required for low-frequency feature extraction are less than the computing power configuration threshold, it means that the terminal system still has redundant computing power reserves, which can provide sufficient computing power support for the computing power required for high-frequency feature extraction and low-frequency feature extraction. It is only necessary to schedule redundant computing power within the terminal system to support the first high-frequency feature extractor or the first low-frequency feature extractor that is performing the operation, and perform edge computing power scheduling configuration. The specific edge computing power scheduling configuration method is an existing technology and will not be repeated here.

[0116] Example 2, as Figure 2 As shown, based on the same inventive concept as the water pump casing image processing method for quality inspection provided in the first embodiment, the embodiment of the present invention further provides a water pump casing image processing system for quality inspection, comprising:

[0117] The deviation vector matrix extraction unit 11 is used to compare the production expected control parameters and the production monitoring control parameters of the water pump housing to obtain the production control parameter deviation vector matrix;

[0118] a defect type identification unit 12, configured to retrieve, from historical samples of water pump casings, high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold;

[0119] The defect feature processing unit 13 is used to process the high-frequency water pump housing defect and the low-frequency water pump housing defect based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor;

[0120] The image feature extraction unit 14 is used to call the high-frequency feature extractor to perform feature extraction on all water pump casing images that meet the production control parameter deviation vector matrix, and call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that meet the production control parameter deviation vector matrix to obtain image feature extraction results.

[0121] Furthermore, the deviation vector matrix extraction unit 11 is further configured to:

[0122] Normalizing the production expectation control parameter and the production monitoring control parameter to obtain a production expectation control characteristic value and a production monitoring control characteristic value;

[0123] The absolute deviation values ​​of the same attributes in the same domain are calculated for the production expected control characteristic value and the production monitoring control characteristic value to obtain the first attribute deviation time series information until the Nth attribute deviation time series information, wherein the time series covers the entire production time zone, and the matrix element configuration is empty when the attribute is not set at the corresponding time.

[0124] Furthermore, the defect type identification unit 12 is further configured to:

[0125] Extracting a sample production control parameter deviation vector matrix from the water pump housing historical samples;

[0126] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix; if the matrix similarity is greater than or equal to a matrix similarity threshold, adding the sample to the selected water pump casing history sample; otherwise, updating the water pump casing history sample;

[0127] When the selected water pump housing historical samples are greater than or equal to the set number, the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency, wherein the set number is ≥ 10,000 and the historical samples all belong to the set time window;

[0128] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extracting defects whose attribute trigger frequency is greater than or equal to the first trigger frequency threshold and setting them as the high-frequency water pump casing defects;

[0129] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extract defects whose attribute trigger frequency is greater than or equal to the second trigger frequency threshold and less than the first trigger frequency threshold, and set them as the low-frequency water pump casing defects;

[0130] Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects whose attribute trigger frequency is less than the second trigger frequency threshold are extracted and set not to participate in the detection of water pump casing defects.

[0131] Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix includes:

[0132] Through the user end, configure the first attribute deviation threshold up to the Nth attribute deviation threshold;

[0133] Based on the first attribute deviation threshold up to the Nth attribute deviation threshold, combined with the historical samples of the water pump housing, the frequency of simultaneous triggering of the defect and the first attribute when only the first attribute is in a deviation state and the other attributes are in a normal state is counted, and the frequency is set as the first attribute defect support;

[0134] Until the Nth attribute defect support is obtained;

[0135] Calculate the ratio of the defect support of the first attribute to the defect support of the Nth attribute to the sum of the defect support, and obtain the weight distribution vector of the first attribute to the Nth attribute;

[0136] Based on the weight distribution vector, weighting the production control parameter deviation vector matrix and the sample production control parameter deviation vector matrix respectively to obtain a production control parameter deviation vector update matrix and a sample production control parameter deviation vector update matrix;

[0137] Calculate the matrix similarity between the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix.

[0138] The trigger frequency threshold includes a first trigger frequency threshold of type I and a second trigger frequency threshold of type I, wherein the first trigger frequency threshold of type I is greater than the second trigger frequency threshold of type I; the trigger frequency threshold includes a first trigger frequency threshold of type II and a second trigger frequency threshold of type II, wherein the first trigger frequency threshold of type II is greater than the second trigger frequency threshold of type II; the first trigger frequency threshold of type I is greater than the first trigger frequency threshold of type II, including:

[0139] When the defect attribute belongs to a failure defect, the second-class first trigger frequency threshold and the second-class second trigger frequency threshold are used;

[0140] When the defect belongs to a non-failure defect, the first trigger frequency threshold and the second trigger frequency threshold are used.

[0141] Furthermore, the image feature extraction unit 14 is further configured to:

[0142] The low-frequency feature extractor is called to perform feature extraction on a set proportion water pump casing image that satisfies the production control parameter deviation vector matrix, including:

[0143] Calculate the ratio of the number of occurrence samples of the trigger frequency of the low-frequency water pump casing defect to the total number of historical samples of the water pump casing, and set it as the sampling rate threshold;

[0144] The sampling rate threshold is used as the minimum sampling rate, the low-frequency feature extractor is called, and feature extraction is performed on the set proportion of water pump casing images that meet the production control parameter deviation vector matrix.

[0145] The high-frequency feature extractor is used to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix, and the low-frequency feature extractor is used to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix to obtain image feature extraction results, further comprising:

[0146] Counting the historical average computing power of the high-frequency feature extractor and the low-frequency feature extractor respectively to obtain the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction;

[0147] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, extract the first high-frequency feature extractor and the first low-frequency feature extractor whose required computing power is greater than or equal to the computing power configuration threshold, and perform cloud computing power scheduling configuration;

[0148] Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, a second high-frequency feature extractor and a second low-frequency feature extractor whose required computing power is less than the computing power configuration threshold are extracted to perform edge computing power scheduling configuration.

[0149] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0150] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0154] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0155] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A water pump casing image processing method for quality inspection, characterized in that: include: Compare the production expected control parameters and production monitoring control parameters of the water pump casing to obtain the production control parameter deviation vector matrix; Retrieving high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and whose trigger frequency is greater than or equal to a trigger frequency threshold from historical samples of the water pump casing; Based on a predefined defect feature extractor calibration table, processing the high-frequency water pump housing defect and the low-frequency water pump housing defect to obtain a high-frequency feature extractor and a low-frequency feature extractor; The high-frequency feature extractor is called to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix; the low-frequency feature extractor is called to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix to obtain image feature extraction results; Among them, the production expected control parameters and production monitoring control parameters of the water pump housing are compared to obtain the production control parameter deviation vector matrix, including: Normalizing the production expectation control parameter and the production monitoring control parameter to obtain a production expectation control characteristic value and a production monitoring control characteristic value; Calculate the absolute deviation value of the same attribute in the same domain for the production expected control characteristic value and the production monitoring control characteristic value to obtain the first attribute deviation time series information to the Nth attribute deviation time series information, wherein the time series covers the entire production time zone, and the matrix element configuration is empty when the attribute is not set at the corresponding time; Among them, from the historical samples of the water pump casing, high-frequency water pump casing defects and low-frequency water pump casing defects that meet the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold are retrieved, and the trigger frequency threshold includes a first trigger frequency threshold and a second trigger frequency threshold, and the first trigger frequency threshold is greater than the second trigger frequency threshold, including: Extracting a sample production control parameter deviation vector matrix from the water pump housing historical samples; Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix; if the matrix similarity is greater than or equal to a matrix similarity threshold, adding the sample to the selected water pump casing history sample; otherwise, updating the water pump casing history sample; When the selected water pump housing historical samples are greater than or equal to the set number, the first defect attribute trigger frequency is counted until the Qth defect attribute trigger frequency, wherein the set number is ≥ 10,000 and the historical samples all belong to the set time window; Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extracting defects whose attribute trigger frequency is greater than or equal to the first trigger frequency threshold and setting them as the high-frequency water pump casing defects; Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, extract defects whose attribute trigger frequency is greater than or equal to the second trigger frequency threshold and less than the first trigger frequency threshold, and set them as the low-frequency water pump casing defects; Based on the first defect attribute trigger frequency up to the Qth defect attribute trigger frequency, defects whose attribute trigger frequency is less than the second trigger frequency threshold are extracted and set not to participate in the detection of water pump casing defects.

2. The method according to claim 1, wherein Calculating the matrix similarity between the sample production control parameter deviation vector matrix and the production control parameter deviation vector matrix includes: Through the user end, configure the first attribute deviation threshold up to the Nth attribute deviation threshold; Based on the first attribute deviation threshold to the Nth attribute deviation threshold, combined with the historical samples of the water pump housing, the frequency of defect triggering when only the first attribute is in a deviation state and the other attributes are in a normal state is counted, and the frequency is set as the first attribute defect support; Until the Nth attribute defect support is obtained; Calculate the ratio of the defect support of the first attribute to the defect support of the Nth attribute to the sum of the defect support, and obtain the weight distribution vector of the first attribute to the Nth attribute; Based on the weight distribution vector, weighting the production control parameter deviation vector matrix and the sample production control parameter deviation vector matrix respectively to obtain a production control parameter deviation vector update matrix and a sample production control parameter deviation vector update matrix; Calculate the matrix similarity between the production control parameter deviation vector update matrix and the sample production control parameter deviation vector update matrix.

3. The method according to claim 1, wherein The trigger frequency threshold includes a first trigger frequency threshold of type I and a second trigger frequency threshold of type I, wherein the first trigger frequency threshold of type I is greater than the second trigger frequency threshold of type I; the trigger frequency threshold includes a first trigger frequency threshold of type II and a second trigger frequency threshold of type II, wherein the first trigger frequency threshold of type II is greater than the second trigger frequency threshold of type II; the first trigger frequency threshold of type I is greater than the first trigger frequency threshold of type II, including: When the defect attribute belongs to a failure defect, the second-class first trigger frequency threshold and the second-class second trigger frequency threshold are used; When the defect belongs to a non-failure defect, the first trigger frequency threshold and the second trigger frequency threshold are used.

4. The method according to claim 1, wherein The low-frequency feature extractor is called to perform feature extraction on a set proportion water pump casing image that satisfies the production control parameter deviation vector matrix, including: Calculate the ratio of the number of occurrence samples of the trigger frequency of the low-frequency water pump casing defect to the total number of historical samples of the water pump casing, and set it as the sampling rate threshold; Taking the sampling rate threshold as the minimum sampling rate, the low-frequency feature extractor is called to perform feature extraction on the set proportion of water pump casing images that meet the production control parameter deviation vector matrix.

5. The method according to claim 1, wherein The high-frequency feature extractor is called to perform feature extraction on all water pump casing images that satisfy the production control parameter deviation vector matrix, and the low-frequency feature extractor is called to perform feature extraction on a set proportion of water pump casing images that satisfy the production control parameter deviation vector matrix to obtain image feature extraction results, further comprising: Counting the historical average computing power of the high-frequency feature extractor and the low-frequency feature extractor respectively to obtain the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction; Based on the high-frequency feature extraction demand computing power and the low-frequency feature extraction demand computing power, extract the first high-frequency feature extractor and the first low-frequency feature extractor whose demand computing power is greater than or equal to the computing power configuration threshold, and perform cloud computing power scheduling configuration; Based on the required computing power for high-frequency feature extraction and the required computing power for low-frequency feature extraction, a second high-frequency feature extractor and a second low-frequency feature extractor whose required computing power is less than the computing power configuration threshold are extracted to perform edge computing power scheduling configuration.

6. A water pump casing image processing system for quality inspection, characterized in that: Used to implement the method according to any one of claims 1 to 5, comprising: A deviation vector matrix extraction unit is used to compare the production expected control parameters and production monitoring control parameters of the water pump housing to obtain the production control parameter deviation vector matrix; a defect type identification unit, configured to retrieve, from historical samples of water pump casings, high-frequency water pump casing defects and low-frequency water pump casing defects that satisfy the production control parameter deviation vector matrix and have a trigger frequency greater than or equal to a trigger frequency threshold; a defect feature processing unit, configured to process the high-frequency water pump casing defect and the low-frequency water pump casing defect based on a predefined defect feature extractor calibration table to obtain a high-frequency feature extractor and a low-frequency feature extractor; An image feature extraction unit is used to call the high-frequency feature extractor to perform feature extraction on all water pump casing images that meet the production control parameter deviation vector matrix, call the low-frequency feature extractor to perform feature extraction on a set proportion of water pump casing images that meet the production control parameter deviation vector matrix, and obtain image feature extraction results.

Images