Hub intelligent identification method based on image identification

Abstract

The invention relates to an intelligent wheel hub identification method based on image identification, which combines an ultrasonic scanning technology, performs ultrasonic scanning atlas acquisition on a forged wheel hub, performs slicing operation, and identifies wheel hub forging image features by using an AI model so as to judge whether the internal structure of the forged wheel hub has defect features or not. And if the defect features exist, the corresponding defect feature type is output, forging quality judgment is conducted, and therefore whether the forged hub is qualified or not is judged according to the defect features. Forging quality detection is carried out through automatic ultrasonic spectrum scanning and intelligent defect feature recognition, high-efficiency forging detection can be carried out instead of manual work, and manpower is saved; meanwhile, according to the unified hub forging quality detection means and detection standards, quality inspection is conducted on the forged hub, defect recognition is conducted through an intelligent model, detection personnel can be liberated, the hub quality detection efficiency is greatly improved, the quality inspection entry requirement and difficulty are reduced, and the hub quality recognition workload is reduced.

Description

Technical Field

[0001] The present disclosure relates to the technical field of quality inspection, and in particular to a method, system and electronic device for intelligent wheel hub recognition based on image recognition. Background Art

[0002] The defects that may occur during the wheel forging process mainly include but are not limited to the following:

[0003] fold

[0004] As attached Figure 1 As shown in the figure, folding is the main reason for forged aluminum alloy wheels to become scrap. According to statistics, among all the scrapped forged aluminum alloy wheels, nearly 75% are due to folding. The reasons for folding may include large shape changes of forgings, complex movement trajectories of internal structures, failure to comply with operating procedures during forging, unreasonable design of preforms and pre-forging dies, and improper pressure application speed.

[0005] The internal filling of the forging is not sufficient

[0006] As attached Figure 2 As shown in the figure, improper operation or unreasonable process parameter setting during the forging process may lead to the risk of looseness and insufficient filling inside the forging. This will affect the mechanical properties and service life of the wheel hub. In order to prevent the occurrence of insufficient filling, effective measures such as increasing the forging ratio and casting time are needed.

[0007] crack

[0008] As attached Figure 3 As shown in the figure, cracks are very likely to appear on the surface and inside of forged aluminum alloy wheels. The generation of cracks may be related to the heating temperature of the material not meeting the requirements or the insulation time not being long enough, which will cause excessive deformation of the material. Not only can the folding and bending phenomena generated during the forging process not be eliminated, but more cracks may be generated on the surface or inside of the material. In addition, oxide inclusions inside the material are also an important cause of cracks.

[0009] In addition, forged wheels may have other defects, such as substandard surface roughness and poor dimensional accuracy. These defects are related to factors such as forging process, equipment conditions, and operator skill level. Therefore, in the production process of forged wheels, it is necessary to strictly control various process parameters and operating procedures to ensure that product quality meets relevant standards and requirements.

[0010] In traditional forging quality inspection, the main focus is on the surface defects of forged parts, such as flatness, and the quality inspection of the internal structure of forged wheels is rarely involved. Even for the quality inspection of the internal structure of forged wheels, it is also manually sampled and inspected. This inspection method has the following technical defects:

[0011] First, it requires manpower, long-term contact with forgings, and on-site testing, which increases the workload of workers and long-term contact with the forging environment, which affects their health. In addition, manual testing requires sampling and testing in sequence, and the testing methods and levels vary, depending on the professional level of the testers. The entry requirements are high and the difficulty is great. There is no unified testing standard and method;

[0012] Second, in the inspection of large quantities of forged wheels, even if it is a sampling inspection, the workload is relatively huge, it takes a long time, and the efficiency of manual inspection is low. Summary of the invention

[0013] In order to solve the above problems, the present application proposes a wheel hub intelligent recognition method, system and electronic equipment based on image recognition.

[0014] In one aspect, the present application proposes a wheel hub intelligent recognition method based on image recognition, comprising the following steps:

[0015] S1. Obtaining an ultrasonic scanning image of a forged wheel hub;

[0016] S2, performing image formatting processing on the ultrasonic scanning atlas, and performing image slicing processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas;

[0017] S3, traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deployed wheel hub forging intelligent recognition model;

[0018] S4, importing the wheel hub image feature into a pre-deployed wheel hub forging intelligent recognition model, and identifying and judging whether the wheel hub image feature in the current wheel hub slice image is a defect feature through the wheel hub forging intelligent recognition model:

[0019] If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image;

[0020] If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed;

[0021] S5. The backend server collects statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identifies the statistical results according to the preset wheel hub forging quality inspection rules, determines whether the forged wheel hub is qualified, and outputs the corresponding judgment result.

[0022] As an optional embodiment of the present application, optionally, S1, obtaining an ultrasonic scanning spectrum of the forged wheel hub, comprises:

[0023] The backend server generates a preset robot arm motion trajectory and ultrasonic scanning data, and sends them to the robot arm and ultrasonic scanning system respectively;

[0024] The robot arm moves to grab the forged wheel hub to be inspected according to a preset robot arm motion trajectory, and places it in the scanning space of the ultrasonic scanning system;

[0025] The ultrasonic scanning system starts scanning according to preset ultrasonic scanning data, performs ultrasonic scanning on the forged wheel hub according to a predetermined scanning direction, obtains the ultrasonic scanning graphs in different scanning directions, and uploads them to the backend server;

[0026] The backend server records and stores the ultrasonic scanning graphs in each scanning direction.

[0027] As an optional implementation of the present application, optionally, in step S2, image formatting is performed on the ultrasonic scanning atlas, including:

[0028] According to a preset image size format, the pixels and sizes of the ultrasonic scanning atlas in each scanning direction are processed into an image size format of a unified format;

[0029] According to the preset image saving format, the ultrasonic scanning atlas in each scanning direction is stored in the MySQL database on the backend server.

[0030] As an optional implementation scheme of the present application, optionally, in step S2, the ultrasonic scanning atlas is sliced ​​according to a preset image format area to obtain a slice image set consisting of hub slice images in different image format areas, including:

[0031] Constructing a blank image cutting template, on which a number of image format areas for cutting the hub slice image are reserved and marked with serial numbers;

[0032] The ultrasonic scanning atlas in each scanning direction is sequentially imported into the cutting blank template, and the ultrasonic scanning atlas is subjected to cutting processing using the cutting blank template. The hub slice image in the image format area is cut and extracted, and the ultrasonic scanning atlas is sliced ​​into a slice image set consisting of a plurality of the hub slice images.

[0033] As an optional implementation scheme of the present application, optionally, S3, traversing the slice image set, sequentially extracting the hub image features in the hub slice image and importing them into a pre-deployed hub forging intelligent recognition model, including:

[0034] Reading the slice image set in each scanning direction according to the scanning order;

[0035] Traversing the slice image set, sequentially reading each of the hub slice images in the slice image set, and inputting them into an image processing algorithm library one by one, performing image feature recognition on the hub slice images by several image processing algorithms in the image processing algorithm library, and identifying and extracting several hub image features in the hub slice images;

[0036] According to a preset quality inspection feature type, a hub image feature for defect feature detection is selected from a plurality of hub image features in the hub slice image;

[0037] The wheel hub image features used for defect feature detection are output, recorded by the background server and imported into the pre-deployed wheel hub forging intelligent recognition model for defect recognition.

[0038] As an optional implementation scheme of the present application, optionally, in step S4, the method for generating the wheel hub forging intelligent recognition model includes:

[0039] Obtaining ultrasonic scanning images of several forged wheel hubs having defect features in different scanning directions;

[0040] Performing slice preprocessing on the ultrasonic scanning atlas to obtain a slice image set;

[0041] Performing feature engineering on each wheel hub slice image in the slice image set, extracting wheel hub defect image features with defect features in the wheel hub slice images, and obtaining a wheel hub defect image feature set;

[0042] According to a preset ratio, the wheel hub defect image feature set is divided into a training set and a verification set;

[0043] Importing the training set into a preset CNN model to perform feature learning and generate an initial wheel forging intelligent recognition model;

[0044] Using the verification set to perform defect recognition verification on the wheel hub forging intelligent recognition model;

[0045] After the defect recognition verification is passed, the wheel hub forging intelligent recognition model is randomly verified by using wheel hub image features of several normal forged wheel hubs;

[0046] If the abnormality recognition fails, the wheel hub forging intelligent recognition model is deployed on the background server.

[0047] As an optional implementation scheme of the present application, optionally, S5, the backend server performs statistics on the wheel hub slice images marked with the defect feature type in the slice image set, and identifies the statistical results according to the preset wheel hub forging quality inspection rules, determines whether the forged wheel hub is qualified, and outputs the corresponding judgment result, including:

[0048] Traversing the slice image set, and finding the hub slice image marked with the defect feature type;

[0049] According to the preset wheel hub forging quality inspection rules, it is determined whether the defect feature of the wheel hub slice image marked with the defect feature type belongs to the following defect types: crack, fold or insufficient filling:

[0050] If it is, the forged wheel hub is deemed unqualified;

[0051] If not, then count the number n of the wheel hub slice images marked with the defect feature type in the slice image set, if:

[0052] If n>3-5%N, the forged wheel hub is considered scrap;

[0053] 0.5-1.5%<n≤3-5%N, it is considered as defective product;

[0054] 0<n≤0.5-1.5%, it is considered as defective;

[0055] N is the total number of the hub slice images in the slice image set;

[0056] The judgment result of the forged wheel hub is output, and the selling price of the forged wheel hub is calculated according to different unit prices, and the judgment result and the selling price of the forged wheel hub are stored in the MySQL database on the background server.

[0057] On the other hand, the present application proposes a system for implementing the wheel hub intelligent recognition method based on image recognition, comprising:

[0058] (1) a robotic arm, used to grab the forged wheel hub to be inspected according to a preset robotic arm motion trajectory and place it in the scanning space of the ultrasonic scanning system;

[0059] (2) an ultrasonic scanning system, used to start scanning according to preset ultrasonic scanning data, perform ultrasonic scanning on the forged wheel hub according to a predetermined scanning direction, obtain the ultrasonic scanning graphs in different scanning directions, and upload them to a backend server;

[0060] (3) Backend server, on which are deployed:

[0061] A robotic arm control system, used to generate a preset robotic arm motion trajectory;

[0062] An ultrasonic control system, used for generating preset ultrasonic scanning data;

[0063] A slice module, used for performing image formatting processing on the ultrasonic scanning atlas, and performing slice processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas;

[0064] A feature engineering module, used for traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deploying a wheel hub forging intelligent recognition model;

[0065] The wheel forging intelligent recognition model is used to perform intelligent defect recognition on the imported wheel image features and determine whether the wheel image features in the current wheel slice image are defect features:

[0066] If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image;

[0067] If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed;

[0068] A wheel hub forging quality inspection module, used to collect statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identify the statistical results according to the preset wheel hub forging quality inspection rules, judge whether the forged wheel hub is qualified, and output the corresponding judgment result;

[0069] MySQL database, for data storage;

[0070] (4) a front end for inputting control and displaying the judgment result of the forged wheel hub;

[0071] The mechanical arm, ultrasonic scanning system and front end are respectively connected to the backend server for communication.

[0072] In another aspect, the present application further provides an electronic device, comprising:

[0073] processor;

[0074] a memory for storing processor-executable instructions;

[0075] Wherein, the processor is configured to implement the wheel hub intelligent identification method based on image recognition when executing the executable instructions.

[0076] Technical effects of the present invention:

[0077] Based on the implementation of the present invention, the present invention combines ultrasonic scanning technology to obtain ultrasonic scanning spectra of forged wheels and perform slicing operations. The wheel hub forging features are extracted from the ultrasonic spectra, and the wheel hub forging image features are identified using an AI model to determine whether there are defect features in the internal structure of the forged wheel hub. If there are defect features, the corresponding defect feature type is output and the forging quality is judged to determine whether the forged wheel hub is qualified based on the defect features. Forging quality inspection is carried out through automated ultrasonic spectrum scanning and intelligent defect feature recognition, which can replace manual high-efficiency forging inspection and save manpower; at the same time, according to unified wheel hub forging quality inspection means and inspection standards, the forged wheel hub is inspected for quality, and defects are identified through intelligent models, which can liberate inspection personnel, greatly improve the efficiency of wheel hub quality inspection, reduce the entry requirements and difficulty of quality inspection, and reduce the workload of wheel hub quality identification.

[0078] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0080] Figure 1 A schematic diagram showing defect characteristics of folding inside the hub is shown;

[0081] Figure 2 A schematic diagram showing defect characteristics of insufficient filling inside the hub;

[0082] Figure 3 The figure shows a schematic diagram of defect characteristics of cracks generated inside the wheel hub;

[0083] Figure 4 It is shown as a schematic diagram of the implementation process of the present invention;

[0084] Figure 5 It is shown as a schematic diagram of the application system composition of the present invention;

[0085] Figure 6 It is a schematic diagram showing the application of the electronic device of the present invention. DETAILED DESCRIPTION

[0086] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0087] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0088] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present disclosure.

[0089] Example 1

[0090] like Figure 4 As shown, in one aspect, the present application proposes a wheel hub intelligent recognition method based on image recognition, comprising the following steps:

[0091] S1. Obtaining an ultrasonic scanning image of a forged wheel hub;

[0092] S2, performing image formatting processing on the ultrasonic scanning atlas, and performing image slicing processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas;

[0093] S3, traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deployed wheel hub forging intelligent recognition model;

[0094] S4, importing the wheel hub image feature into a pre-deployed wheel hub forging intelligent recognition model, and identifying and judging whether the wheel hub image feature in the current wheel hub slice image is a defect feature through the wheel hub forging intelligent recognition model:

[0095] If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image;

[0096] If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed;

[0097] S5. The backend server collects statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identifies the statistical results according to the preset wheel hub forging quality inspection rules, determines whether the forged wheel hub is qualified, and outputs the corresponding judgment result.

[0098] Wheel forging defect identification based on ultrasonic images is an efficient and non-destructive quality inspection method. This method uses the characteristics of ultrasonic waves that will produce reflection, refraction, transmission and scattering when encountering different medium interfaces when propagating in the material. By capturing and analyzing these sound wave signals, defects in the wheel forging process can be detected and identified.

[0099] During the inspection process, the ultrasonic probe is placed on the surface of the wheel hub and emits ultrasonic pulses into the inside of the wheel hub. When the sound wave encounters internal defects (such as cracks, pores, inclusions, etc.), it will be reflected and generate specific sound wave signals. These signals are then received by the probe and converted into electrical signals, which are finally displayed in the form of images after signal processing and analysis.

[0100] Ultrasonic images can directly display the defects inside the wheel hub, including the location, size, shape and nature of the defects. Inspectors can accurately evaluate the forging quality of the wheel hub based on the image information and take appropriate treatment measures.

[0101] In order to achieve high-quality defect identification, the method usually involves multiple steps and technical links. First, it is necessary to select appropriate ultrasonic probes and detection parameters to ensure that the sound waves can fully penetrate the wheel hub and accurately reflect the internal situation. Secondly, advanced signal processing technology is needed during the detection process to filter, enhance and extract features of the received sound wave signals to improve the signal-to-noise ratio and clarity of the image. Finally, it is also necessary to combine professional knowledge and experience to accurately interpret and judge the ultrasonic images to identify various defects inside the wheel hub.

[0102] Wheel forging defect recognition based on ultrasonic images is a scientific, reliable and efficient quality inspection method. Through the application of this method, comprehensive monitoring and effective control of wheel forging quality can be achieved to ensure the stability and reliability of product quality.

[0103] Combined with Figure 5 As shown in the application system shown, the present invention uses ultrasound to scan and obtain the ultrasonic map of the internal structure of the forged wheel hub and performs feature engineering on the image, extracts the image features therein, and uses the pre-trained and generated wheel hub forging intelligent recognition model to perform defect recognition on the extracted feature image, and judges whether there are defect image features learned and recognized by the model in the ultrasonic scanning image of the forged wheel hub. If so, it is determined to be an unqualified product, and subsequent quality inspection judgment is carried out.

[0104] In order to realize the automated quality inspection of forged wheel hubs, the present invention also utilizes a mechanical arm and an ultrasonic scanning system. The mechanical arm is controlled by the background to grab the forged wheel hub to the scanning environment, and the ultrasonic scanning system performs ultrasonic scanning on the forged wheel hub in different scanning directions, obtains the ultrasonic scanning spectrum of the forged wheel hub in different directions, and uploads it to the background server. The corresponding image processing, image feature engineering and image intelligent recognition are then performed on the background server, and subsequent forging defect judgment is performed, and the corresponding judgment results are output to the front end. Through the implementation of this system, the quality defect inspection of the forged wheel hub can be replaced by manual work, thereby improving the inspection efficiency, realizing unmanned intelligent inspection, and avoiding the health risks caused by the inspection personnel contacting the wheel hub for a long time; at the same time, according to the intelligent and automated inspection scheme of this system, objective and unified quality inspection can be carried out, the quality inspection requirements of the wheel hub forging defects can be reduced, and the difficulty of inspection can be reduced.

[0105] As an optional embodiment of the present application, optionally, S1, obtaining an ultrasonic scanning spectrum of the forged wheel hub, comprises:

[0106] The backend server generates a preset robot arm motion trajectory and ultrasonic scanning data, and sends them to the robot arm and ultrasonic scanning system respectively;

[0107] The robot arm moves to grab the forged wheel hub to be inspected according to a preset robot arm motion trajectory, and places it in the scanning space of the ultrasonic scanning system;

[0108] The ultrasonic scanning system starts scanning according to preset ultrasonic scanning data, performs ultrasonic scanning on the forged wheel hub according to a predetermined scanning direction, obtains the ultrasonic scanning graphs in different scanning directions, and uploads them to the backend server;

[0109] The backend server records and stores the ultrasonic scanning graphs in each scanning direction.

[0110] In the forging workshop, under the control of the background server, the robotic arm can be controlled to grab the forged wheel hub and place it in the scanning space of the ultrasonic scanning system. The ultrasonic scanning system set on one side can be used to perform ultrasonic scanning on the forged wheel hub in different directions to obtain ultrasonic scanning maps of the forged wheel hub in various directions, which are then transmitted back to the background server by the ultrasonic scanning system.

[0111] The backend server records and saves the atlas acquisition data of the current forged wheel hub.

[0112] In this solution, the background control software of the robotic arm and the ultrasonic scanning system are respectively deployed on the background server, which can perform corresponding operation control.

[0113] For the robotic arm and ultrasonic scanning system in this solution, the user can configure the corresponding intelligent device by himself, and complete the communication configuration with the background server according to the background communication conditions supported by it.

[0114] At the same time, the coordinates between the initial coordinates of the robot arm and the scanning space can be calibrated in advance and recorded on the background server. Subsequently, the background server can generate the robot arm's motion trajectory of grabbing the forged wheel hub and placing it in the scanning environment. It can be generated according to the spatial coordinates calibrated in the early stage. The specific understanding can be combined with the existing robot arm motion control, and this solution will not go into details.

[0115] For the ultrasonic scanning system, the forged wheel hub can be scanned multiple times according to different scanning sequences and scanning directions. For example, during the scanning process, the background can control the robotic arm to rotate the forged wheel hub to the corresponding scanning direction according to the preset scanning direction, and let the ultrasonic scanning system scan to complete the collection of ultrasonic scanning maps in different scanning directions. The background can record the scanning direction of each forged wheel hub, bind its corresponding ultrasonic scanning map with the corresponding scanning direction and store it in the database, so as to facilitate the subsequent identification of the ultrasonic scanning map of the forged wheel hub in each scanning direction and its forging defect identification results according to the scanning information, for example, according to the scanning direction, so as to facilitate the spot check of the forging quality of the forged wheel hub from multiple scanning directions, avoid the lack of comprehensiveness of the spot check in a single scanning direction, avoid the risk of incomplete quality inspection and failure to inspect in place, and improve the comprehensiveness of the forging quality inspection.

[0116] The ultrasonic probe is the core component of the detection system, responsible for emitting ultrasonic waves and receiving the reflected signals. According to the shape, size, material properties and detection depth requirements of the metal to be detected, the appropriate probe type and frequency need to be selected. For example, for small-sized or thin-walled metal materials, a high-frequency probe may be required to improve resolution; for large-sized or thick-walled metal materials, a low-frequency probe may be required to ensure sufficient penetration.

[0117] Before testing, it is also necessary to check and confirm the main technical indicators of the ultrasonic instrument, such as the incident point of the oblique probe, the slope K value or angle, etc. At the same time, the instrument time baseline should be adjusted according to the size of the workpiece to be tested, and the distance-amplitude (DAC) curve should be drawn. The DAC curve is drawn from the measured data on the comparison test block and is used to set the detection sensitivity and evaluate the severity of the defect.

[0118] Next, the specific operation of ultrasonic testing is carried out. Before testing, the detection surface should be trimmed or polished to remove welding spatter, oil stains and other impurities to ensure that the surface roughness meets the requirements. Then, apply an appropriate amount of coupling agent on the detection surface so that the ultrasonic wave can be smoothly transmitted into the inside of the material being tested. The coupling agent should have good sound permeability and appropriate fluidity, and at the same time not cause damage to the material and the human body.

[0119] During the inspection process, the ultrasonic probe is placed close to the metal surface to be inspected and scanned according to the predetermined scanning method. The scanning speed should be controlled within a certain range (such as not exceeding 150mm / s) to ensure the accuracy and reliability of the inspection. At the same time, the two adjacent probe movement areas should maintain a certain proportion of probe width overlap to avoid missed inspections.

[0120] By transmitting and receiving ultrasound, image information of the internal structure of the metal can be obtained. This image information reflects the structure and defects inside the metal. Next, the received echo signal needs to be analyzed and processed to identify and evaluate the defects in the metal structure. Using signal processing technology, the location, size, shape and other information of the defect can be extracted and further quantitatively evaluated.

[0121] As an optional implementation of the present application, optionally, in step S2, image formatting is performed on the ultrasonic scanning atlas, including:

[0122] According to a preset image size format, the pixels and sizes of the ultrasonic scanning atlas in each scanning direction are processed into an image size format of a unified format;

[0123] According to the preset image saving format, the ultrasonic scanning atlas in each scanning direction is stored in the MySQL database on the backend server.

[0124] Image size format, that is, the scanned image needs to maintain a consistent size in all directions, such as pixels, so that the image can be cut in a unified size format to avoid different sizes of cuts, which is inconvenient for image recognition.

[0125] Image saving format, image scanning in various directions may result in different formats due to the change of scanning sequence of the ultrasonic scanning system. Therefore, if the image format is inconsistent, the image saving format is converted to output a unified image format.

[0126] In addition, the image is preprocessed, including denoising, contrast enhancement, grayscale and other steps, to improve the image quality and reduce the complexity of subsequent processing.

[0127] As an optional implementation scheme of the present application, optionally, in step S2, the ultrasonic scanning atlas is sliced ​​according to a preset image format area to obtain a slice image set consisting of hub slice images in different image format areas, including:

[0128] Constructing a blank image cutting template, on which a number of image format areas for cutting the hub slice image are reserved and marked with serial numbers;

[0129] The ultrasonic scanning atlas in each scanning direction is sequentially imported into the cutting blank template, and the ultrasonic scanning atlas is subjected to cutting processing using the cutting blank template. The hub slice image in the image format area is cut and extracted, and the ultrasonic scanning atlas is sliced ​​into a slice image set consisting of a plurality of the hub slice images.

[0130] In order to facilitate the image cutting operation, the present invention deploys a blank image cutting template designed by C++ programming in the background, which has an image format area set in a matrix grid (such as 8*8, or more, not less than 8*8) and has a serial number. The specific programming design is completed by the user.

[0131] Because the ultrasonic scan map is formatted, its size is consistent with the blank template for cutting. After the ultrasonic scan map is imported into the blank template for cutting, it will overlap with the blank template for cutting. At this time, the image format area on the blank template for cutting will be filled with the local image of the ultrasonic scan map of the corresponding part. At this time, the local image in the image format area is taken to realize the cutting operation of each part.

[0132] That is, one ultrasonic scanning image (or one scanning direction) will be cut into several local images, namely, wheel hub slice images, by the "blank cutting template". Because the image format area of ​​the "blank cutting template" carries a serial number, each wheel hub slice image obtained by slicing will also be marked with the serial number of the corresponding image format area. In this way, the slice image set composed of wheel hub slice images with different serial numbers, namely, the image slices on the ultrasonic scanning image, will subsequently identify the features on each image slice to determine whether there are forging defect features.

[0133] The "blank template for cutting" is used to cut the ultrasonic scanning atlas, which can quickly obtain forging image slices of different parts, and identify the slice images of each forging part on the atlas, so as to carry out refined defect identification and quality inspection, and avoid the large recognition error caused by the whole atlas recognition in defect identification. When the whole atlas is used for image recognition, due to pixel problems, the system cannot perform detailed recognition and extraction of atlas features, and can only perform rough recognition, and the recognition accuracy is insufficient.

[0134] This solution uses a blank template to slice the atlas and slice the scanned image. The pixels of each slice are greatly reduced, which allows the system to perform high-precision image recognition, identify defects in local parts of each screenshot, and determine whether there are forging defects, thereby improving the accuracy and efficiency of defect recognition.

[0135] In the present invention, after the image cutting is completed, for each image cutting sequence, the administrator can determine whether the slice contains a forged wheel hub. Because there may be blank areas during the image scanning process, if there are blank images in the image cutting, the administrator will delete them and re-compose the slice data set to avoid image recognition of blank slices during subsequent image recognition, which will take up image recognition time.

[0136] As an optional implementation scheme of the present application, optionally, S3, traversing the slice image set, sequentially extracting the hub image features in the hub slice image and importing them into a pre-deployed hub forging intelligent recognition model, including:

[0137] Reading the slice image set in each scanning direction according to the scanning order;

[0138] Traversing the slice image set, sequentially reading each of the hub slice images in the slice image set, and inputting them into an image processing algorithm library one by one, performing image feature recognition on the hub slice images by several image processing algorithms in the image processing algorithm library, and identifying and extracting several hub image features in the hub slice images;

[0139] According to a preset quality inspection feature type, a hub image feature for defect feature detection is selected from a plurality of hub image features in the hub slice image;

[0140] The wheel hub image features used for defect feature detection are output, recorded by the background server and imported into the pre-deployed wheel hub forging intelligent recognition model for defect recognition.

[0141] Quality inspection feature type, that is, the type of wheel hub defects that should be excluded in this forging, such as folding, cracks, etc., is specifically configured by the administrator.

[0142] This solution slices the slice data set of the ultrasonic scanning atlas in different scanning directions and intelligently recognizes the slice images one by one. The pre-deployed wheel forging intelligent recognition model performs image recognition on each slice on each atlas in turn to identify defects in the forging image of the forged wheel hub in the scanning direction. This can perform refined forging quality inspection on the forging structure of each part and comprehensively inspect the defects of the forged section.

[0143] This method can not only conduct a comprehensive defect inspection on the forging section, but also reduce the detection pressure (after all, the detection of the entire map requires more computing power and occupies more threads in the background), and the detection accuracy can be improved through slice detection.

[0144] For the image features on the slice image, this solution uses an image processing algorithm to identify and extract the wheel image features. In this solution, because the defect image features generated by forging are of different types, if the contour image features of this type are to be detected in a targeted manner, it is necessary to match the corresponding image processing algorithm in a targeted manner. In order to avoid this problem, the present invention is provided with an image processing algorithm library in the background, in which several image processing algorithms are configured, which are respectively used to identify and extract the defect features such as cracks and flatness of the wheel forging. Therefore, in order to avoid the matching algorithm affecting the detection efficiency, this solution inputs each slice image into the image processing algorithm library, and allows each image processing algorithm in the algorithm library to perform divergent image recognition processing on the input wheel slice image, and randomly allows each algorithm to extract the image features on the slice image (or call the algorithm in sequence to extract, and stop the recognition if the algorithm extracts the defect feature), so as to identify and extract the wheel image features on the slice. It can save the time of configuring the image processing algorithm without affecting the detection efficiency, and allow the system to match the corresponding algorithm by itself, which can avoid manual intervention and save manual operation time. After extracting the wheel image features on the car plate through image processing algorithms, the wheel forging defect image features can be output.

[0145] Image processing algorithms are used to extract key features of the wheel hub. These features may include the number, length, angle, spacing of the spokes, as well as defects such as flatness, cracks, and bubbles on the surface. These features can be extracted through edge detection, morphological processing, texture analysis, and other technologies.

[0146] Image processing algorithms that extract image features are key technologies in the field of image processing. They use different mathematical and computational methods to extract useful information from images for subsequent image analysis, understanding, and recognition. The following are some common image feature extraction algorithms:

[0147] 1. Local Binary Patterns (LBP)

[0148] LBP is an operator used to describe the local texture features of an image, and has grayscale invariance. Its basic idea is to define a window (usually 3x3) around the target pixel, take the target pixel as the threshold, compare the grayscale values ​​of adjacent pixels with the target pixel, and generate a binary number based on the comparison result to obtain the LBP value of the pixel. The LBP algorithm calculates the LBP value of each pixel in the image and counts its histogram to obtain the LBP texture feature vector of the image.

[0149] 2. Histogram of Oriented Gradients (HOG)

[0150] The HOG algorithm constructs features by calculating and counting the gradient direction histogram of the local area of ​​the image. First, the image is grayed and gamma corrected to standardize the color space. Then, the gradient (including size and direction) of each pixel of the image is calculated to capture the contour information. Next, the image is divided into small areas (cels), and the gradient histogram of each cel is counted to form the descriptor of the cel. Finally, multiple cels are combined into a block, and the descriptors of all cels in the block are connected in series to obtain the HOG feature descriptor of the block. The HOG feature descriptor of the entire image is the series connection of the HOG feature descriptors of all cels.

[0151] 3. Scale Invariant Feature Transform (SIFT)

[0152] The SI FT algorithm obtains features by finding the feature points in an image and their descriptions of size and direction, and matches the image feature points. The SI FT algorithm has scale invariance and rotation invariance. It first constructs the scale space of the image and detects the extreme points in the scale space as feature points through the Difference of Gaussian (DoG). Then, low-contrast and unstable edge response points are removed to enhance the stability of the matching. Next, a directional parameter is specified for each key point to make it rotation invariant. Finally, a window is taken with the key point as the center, and the gradient directional histogram of the pixels in the window is calculated to form the descriptor of the key point.

[0153] 4. Other Algorithms

[0154] In addition to the above algorithms, there are many other image feature extraction algorithms, such as SURF (Speeded-Up Robust Features), ORB (Oriented FAST and Rotated BRIEF), HAAR, etc. These algorithms have their own characteristics and are suitable for different application scenarios and requirements.

[0155] The specific configuration algorithm can be configured by the administrator.

[0156] In summary, image feature extraction algorithms are important technologies in image processing. They use different methods to extract useful information from images, providing a basis for subsequent image analysis, understanding, and recognition. In practical applications, appropriate algorithms can be selected for image feature extraction according to specific needs.

[0157] As an optional implementation scheme of the present application, optionally, in step S4, the method for generating the wheel hub forging intelligent recognition model includes:

[0158] Obtaining ultrasonic scanning images of several forged wheel hubs having defect features in different scanning directions;

[0159] Performing slice preprocessing on the ultrasonic scanning atlas to obtain a slice image set;

[0160] Performing feature engineering on each wheel hub slice image in the slice image set, extracting wheel hub defect image features with defect features in the wheel hub slice images, and obtaining a wheel hub defect image feature set;

[0161] According to a preset ratio, the wheel hub defect image feature set is divided into a training set and a verification set;

[0162] Importing the training set into a preset CNN model to perform feature learning and generate an initial wheel forging intelligent recognition model;

[0163] Using the verification set to perform defect recognition verification on the wheel hub forging intelligent recognition model;

[0164] After the defect recognition verification is passed, the wheel hub forging intelligent recognition model is randomly verified by using wheel hub image features of several normal forged wheel hubs;

[0165] If the abnormality recognition fails, the wheel hub forging intelligent recognition model is deployed on the background server.

[0166] This model is based on traditional machine learning algorithms, such as support vector machines (SVM), random forests, etc., or deep learning algorithms, such as convolutional neural networks (CNN). These models require a large number of training samples to learn the difference between normal features and defect features of the wheel hub.

[0167] After the model training is completed, the wheel image to be inspected can be input into the model. The model will automatically extract image features and compare them with the learned features to determine whether the wheel has defects and the type and degree of the defects.

[0168] Specifically, this time we collected ultrasonic scanning images of forged wheels with defect features and processed them by slicing, and performed feature engineering to construct a wheel defect image feature set, which was used to train and generate a wheel forging intelligent recognition model that could identify different forging defect image features.

[0169] This embodiment can take the CNN model as an example to perform intelligent training and application of the wheel forging intelligent recognition model. The general steps can be:

[0170] The training of the wheel forging intelligent recognition model based on CNN (convolutional neural network) aims to identify forging defects in forged wheels through deep learning technology. This process usually involves multiple key steps, including data preparation, model design, training and optimization, evaluation and verification, etc. The following is a detailed overview of the training steps:

[0171] 1. Data Preparation

[0172] 1. Data collection: First, a large amount of forged wheel image data needs to be collected, which should contain samples of various forging defects and normal wheels (negative verification). These images can be collected by industrial cameras or professional equipment.

[0173] 2. Data labeling: Finely label the collected images and mark key information such as the location and type of forging defects. This step is the basis of supervised learning and is crucial for model training.

[0174] 3. Data preprocessing: Preprocess the labeled images, including operations such as image cropping, scaling, and normalization, to improve the training efficiency and recognition accuracy of the model. In addition, data enhancement techniques such as rotation, flipping, and noise addition can also be used to increase the diversity and generalization ability of training data.

[0175] 2. Model Design

[0176] 1. Network architecture design: Based on the principle of CNN, a network architecture suitable for forged wheel defect recognition is designed. This usually includes multiple convolutional layers, pooling layers, activation functions, and fully connected layers. Among them, the convolutional layer is used to extract features from the image, the pooling layer is used to reduce the dimension of the feature map and reduce the amount of calculation, the activation function is used to introduce nonlinear factors, and the fully connected layer is used to map the feature map to the final classification result.

[0177] 2. Loss function and optimization algorithm selection: Select appropriate loss function and optimization algorithm according to task requirements. For multi-classification tasks, commonly used loss functions include cross entropy loss, etc.; optimization algorithms can be SGD (stochastic gradient descent), Adam, etc.

[0178] 3. Training and Optimization

[0179] 1. Model training: Use the preprocessed training data to train the model. During the training process, the weights and bias parameters of the model are updated through the back propagation algorithm to minimize the loss function.

[0180] 2. Hyperparameter tuning: Optimize the performance of the model by adjusting the model's hyperparameters (such as learning rate, batch size, convolution kernel size, etc.). This usually requires multiple experiments and verifications to find the best hyperparameter combination.

[0181] 3. Overfitting and underfitting: During the training process, we need to pay attention to the overfitting and underfitting problems of the model. Overfitting is usually manifested as the model performing well on the training set but deteriorating on the test set; underfitting is manifested as the model performing poorly on both the training set and the test set. For these problems, we can take corresponding measures such as increasing the amount of data, introducing regularization terms, and adjusting the complexity of the model.

[0182] IV. Evaluation and Verification

[0183] 1. Model evaluation: Use an independent validation set (positive validation) to evaluate the trained model. Evaluation indicators can include accuracy, recall, F1 score, etc. The model is validated using images of normal forging and defective forging to improve the accuracy of the model government sample validation.

[0184] 2. Model validation: In addition to using the test set for validation, you can also use methods such as cross-validation to further verify the generalization ability of the model.

[0185] Through the above steps, a CNN-based wheel forging intelligent recognition model can be trained, which can automatically identify forging defects in forged wheels and give corresponding classification results.

[0186] As an optional implementation scheme of the present application, optionally, S5, the backend server performs statistics on the wheel hub slice images marked with the defect feature type in the slice image set, and identifies the statistical results according to the preset wheel hub forging quality inspection rules, determines whether the forged wheel hub is qualified, and outputs the corresponding judgment result, including:

[0187] Traversing the slice image set, and finding the hub slice image marked with the defect feature type;

[0188] According to the preset wheel hub forging quality inspection rules, it is determined whether the defect feature of the wheel hub slice image marked with the defect feature type belongs to the following defect types: crack, fold or insufficient filling:

[0189] If it is, the forged wheel hub is deemed unqualified;

[0190] If not, then count the number n of the wheel hub slice images marked with the defect feature type in the slice image set, if:

[0191] If n>3-5%N, the forged wheel hub is considered scrap;

[0192] 0.5-1.5%<n≤3-5%N, it is considered as defective product;

[0193] 0<n≤0.5-1.5%, it is considered as defective;

[0194] If n=0, it is considered as qualified product;

[0195] N is the total number of the hub slice images in the slice image set;

[0196] The judgment result of the forged wheel hub is output, and the selling price of the forged wheel hub is calculated according to different unit prices, and the judgment result and the selling price of the forged wheel hub are stored in the MySQL database on the background server.

[0197] After slicing each atlas, traversing and testing, the background will mark the sliced ​​images with detected defect features. Therefore, the sliced ​​data set of the atlas can be traversed later to find the wheel slice images marked with defect feature types, and identify the defect feature types on the slices to determine whether the current forged wheel has the preset defect type. If there is a preset defect type, such as cracks, folds, or insufficient filling, the forged wheel with such forging defects is identified as an unqualified product; if there are no forging defects such as cracks and folds, the defective products can be identified based on the statistics of the number of defective slices.

[0198] According to the quality inspection rules of the above wheel segments, the corresponding defective products are identified. Specifically, during the identification, the number of all images marked with defective feature types in all directions of the forged wheel can be counted and compared with the total number of all slices. According to the above judgment method between nN, it is judged whether the current forged wheel is scrap, defective or substandard. These forged wheels can be recycled or used in informal road places such as teaching, so scrap, defective or substandard products can be reused. Defective products are judged based on the statistical results, and billed according to the corresponding unit price (set by yourself), and billing can be quickly performed according to the quality inspection results. For qualified products, they are directly output.

[0199] Therefore, it is possible to conduct self-inspection and judgment of the forged wheel hub according to the defect characteristics and output the corresponding judgment results and selling price, thereby improving quality inspection efficiency.

[0200] Obviously, those skilled in the art should understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned control embodiments. Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned control embodiments. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated as: HDD) or a solid-state drive (Solid-State Drive, SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0201] Example 2

[0202] Based on the implementation principle of Example 1, on the other hand, the present application proposes a system for implementing the wheel hub intelligent identification method based on image recognition, comprising:

[0203] (1) a robotic arm, used to grab the forged wheel hub to be inspected according to a preset robotic arm motion trajectory and place it in the scanning space of the ultrasonic scanning system;

[0204] (2) an ultrasonic scanning system, which is used to start scanning according to preset ultrasonic scanning data, perform ultrasonic scanning on the forged wheel hub in a predetermined scanning direction, obtain the ultrasonic scanning graphs in different scanning directions, and upload them to a backend server;

[0205] (3) Backend server, on which are deployed:

[0206] A robotic arm control system, used to generate a preset robotic arm motion trajectory;

[0207] An ultrasonic control system, used for generating preset ultrasonic scanning data;

[0208] A slice module, used for performing image formatting processing on the ultrasonic scanning atlas, and performing slice processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas;

[0209] A feature engineering module, used for traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deploying a wheel hub forging intelligent recognition model;

[0210] The wheel forging intelligent recognition model is used to perform intelligent defect recognition on the imported wheel image features and determine whether the wheel image features in the current wheel slice image are defect features:

[0211] If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image;

[0212] If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed;

[0213] A wheel hub forging quality inspection module, used to collect statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identify the statistical results according to the preset wheel hub forging quality inspection rules, judge whether the forged wheel hub is qualified, and output the corresponding judgment result;

[0214] MySQL database, for data storage;

[0215] (4) a front end for inputting control and displaying the judgment result of the forged wheel hub;

[0216] The mechanical arm, ultrasonic scanning system and front end are respectively connected to the backend server for communication.

[0217] The functions and interactions of the above-mentioned subjects should be understood and implemented in conjunction with Example 1, and will not be described in detail in this example.

[0218] The modules or steps of the present invention described above can be implemented by a general-purpose computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0219] Example 3

[0220] like Figure 6As shown, further, in another aspect, the present application also proposes an electronic device, including:

[0221] processor;

[0222] a memory for storing processor-executable instructions;

[0223] Wherein, the processor is configured to implement the wheel hub intelligent recognition method based on image recognition described in Example 2 when executing the executable instructions.

[0224] The electronic device of the embodiment of the present disclosure includes a processor and a memory for storing instructions executable by the processor. The processor is configured to implement the wheel hub intelligent recognition method based on image recognition described in the above embodiment 2 when executing the executable instructions.

[0225] Here, it should be noted that the number of processors can be one or more. At the same time, the electronic device of the embodiment of the present disclosure may also include an input system and an output system. Among them, the processor, memory, input system and output system may be connected through a bus or in other ways, which are not specifically limited here.

[0226] The memory, as a computer-readable storage medium, can be used to store software programs, computer executable programs and various modules, such as the program or module corresponding to the wheel hub intelligent recognition method based on image recognition in the embodiment of the present disclosure. The processor executes various functional applications and data processing of the electronic device by running the software programs or modules stored in the memory.

[0227] The input system can be used to receive input numbers or signals. The signal can be a key signal related to user settings and function control of the device / terminal / server. The output system can include display devices such as display screens.

[0228] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A wheel hub intelligent recognition method based on image recognition, It is characterized in that The steps include: S1. Obtaining an ultrasonic scanning image of a forged wheel hub; S2, performing image formatting processing on the ultrasonic scanning atlas, and performing image slicing processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas; S3, traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deployed wheel hub forging intelligent recognition model; S4, importing the wheel hub image feature into a pre-deployed wheel hub forging intelligent recognition model, and identifying and judging whether the wheel hub image feature in the current wheel hub slice image is a defect feature through the wheel hub forging intelligent recognition model: If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image; If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed; S5. The backend server collects statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identifies the statistical results according to the preset wheel hub forging quality inspection rules, determines whether the forged wheel hub is qualified, and outputs the corresponding judgment result.

2. According to the method for intelligent wheel hub recognition based on image recognition according to claim 1, It is characterized in that S1. Obtain the ultrasonic scanning map of the forged wheel hub, including: The backend server generates a preset robot arm motion trajectory and ultrasonic scanning data, and sends them to the robot arm and ultrasonic scanning system respectively; The robot arm moves to grab the forged wheel hub to be inspected according to a preset robot arm motion trajectory, and places it in the scanning space of the ultrasonic scanning system; The ultrasonic scanning system starts scanning according to preset ultrasonic scanning data, performs ultrasonic scanning on the forged wheel hub according to a predetermined scanning direction, obtains the ultrasonic scanning graphs in different scanning directions, and uploads them to the backend server; The backend server records and stores the ultrasonic scanning graphs in each scanning direction.

3. According to the method for intelligent wheel hub recognition based on image recognition according to claim 1, It is characterized in that In step S2, the ultrasonic scanning atlas is subjected to image formatting processing, including: According to a preset image size format, the pixels and sizes of the ultrasonic scanning atlas in each scanning direction are processed into an image size format of a unified format; According to the preset image saving format, the ultrasonic scanning atlas in each scanning direction is stored in the MySQL database on the backend server.

4. According to the method for intelligent wheel hub recognition based on image recognition according to claim 1, It is characterized in that In step S2, the ultrasonic scanning atlas is sliced ​​according to the preset image format area to obtain a slice image set consisting of wheel hub slice images in different image format areas, including: Constructing a blank image cutting template, on which a number of image format areas for cutting the hub slice image are reserved and marked with serial numbers; The ultrasonic scanning atlas in each scanning direction is sequentially imported into the cutting blank template, and the ultrasonic scanning atlas is subjected to cutting processing using the cutting blank template. The hub slice image in the image format area is cut and extracted, and the ultrasonic scanning atlas is sliced ​​into a slice image set consisting of a plurality of the hub slice images.

5. According to the method for intelligent wheel hub recognition based on image recognition according to claim 1, It is characterized in that S3, traversing the slice image set, extracting wheel hub image features in the wheel hub slice image in sequence and importing them into a pre-deployed wheel hub forging intelligent recognition model, including: Reading the slice image set in each scanning direction according to the scanning order; Traversing the slice image set, sequentially reading each of the hub slice images in the slice image set, and inputting them into an image processing algorithm library one by one, performing image feature recognition on the hub slice images by several image processing algorithms in the image processing algorithm library, and identifying and extracting several hub image features in the hub slice images; According to a preset quality inspection feature type, a hub image feature for defect feature detection is selected from a plurality of hub image features in the hub slice image; The wheel hub image features used for defect feature detection are output, recorded by the background server and imported into the pre-deployed wheel hub forging intelligent recognition model for defect recognition.

6. According to the method for intelligent wheel hub recognition based on image recognition according to claim 1, It is characterized in that In step S4, the method for generating the wheel hub forging intelligent recognition model includes: Obtaining ultrasonic scanning images of several forged wheel hubs having defect features in different scanning directions; Performing slice preprocessing on the ultrasonic scanning atlas to obtain a slice image set; Performing feature engineering on each wheel hub slice image in the slice image set, extracting wheel hub defect image features with defect features in the wheel hub slice images, and obtaining a wheel hub defect image feature set; According to a preset ratio, the wheel hub defect image feature set is divided into a training set and a verification set; Importing the training set into a preset CNN model to perform feature learning and generate an initial wheel forging intelligent recognition model; Using the verification set to perform defect recognition verification on the wheel hub forging intelligent recognition model; After the defect recognition verification is passed, the wheel hub forging intelligent recognition model is randomly verified by using wheel hub image features of several normal forged wheel hubs; If the abnormality recognition fails, the wheel hub forging intelligent recognition model is deployed on the background server.

7. According to the method for intelligent wheel hub recognition based on image recognition according to claim 5, It is characterized in that S5, the backend server collects statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identifies the statistical results according to the preset wheel hub forging quality inspection rules, judges whether the forged wheel hub is qualified, and outputs the corresponding judgment result, including: Traversing the slice image set, and finding the hub slice image marked with the defect feature type; According to the preset wheel hub forging quality inspection rules, it is determined whether the defect feature of the wheel hub slice image marked with the defect feature type belongs to the following defect types: crack, fold or insufficient filling: If it is, the forged wheel hub is deemed unqualified; If not, then count the number n of the wheel hub slice images marked with the defect feature type in the slice image set, if: If n>3-5%N, the forged wheel hub is considered scrap; 0.5-1.5%<n≤3-5%N, it is considered as defective product; 0<n≤0.5-1.5%, it is considered as defective; N is the total number of the hub slice images in the slice image set; The judgment result of the forged wheel hub is output, and the selling price of the forged wheel hub is calculated according to different unit prices, and the judgment result and the selling price of the forged wheel hub are stored in the MySQL database on the background server.

8. A system for implementing the wheel hub intelligent identification method based on image recognition as described in any one of claims 1 to 7, It is characterized in that include: (1) a robotic arm, used to grab the forged wheel hub to be inspected according to a preset robotic arm motion trajectory and place it in the scanning space of the ultrasonic scanning system; (2) an ultrasonic scanning system, which is used to start scanning according to preset ultrasonic scanning data, perform ultrasonic scanning on the forged wheel hub in a predetermined scanning direction, obtain the ultrasonic scanning graphs in different scanning directions, and upload them to a backend server; (3) Backend server, on which are deployed: A robotic arm control system, used to generate a preset robotic arm motion trajectory; An ultrasonic control system, used for generating preset ultrasonic scanning data; A slice module, used for performing image formatting processing on the ultrasonic scanning atlas, and performing slice processing on the ultrasonic scanning atlas according to a preset image format area, to obtain a slice image set consisting of wheel hub slice images in different image format areas; A feature engineering module, used for traversing the slice image set, sequentially extracting wheel hub image features in the wheel hub slice image and pre-deploying a wheel hub forging intelligent recognition model; The wheel forging intelligent recognition model is used to perform intelligent defect recognition on the imported wheel image features and determine whether the wheel image features in the current wheel slice image are defect features: If yes, the corresponding defect feature type is output, and the backend server marks the defect feature type on the corresponding wheel hub slice image; If not, identifying the wheel hub image features in the next wheel hub slice image until the identification is completed; A wheel hub forging quality inspection module, used to collect statistics on the wheel hub slice images marked with the defect feature type in the slice image set, identify the statistical results according to the preset wheel hub forging quality inspection rules, judge whether the forged wheel hub is qualified, and output the corresponding judgment result; MySQL database, for data storage; (4) a front end for inputting control and displaying the judgment result of the forged wheel hub; The mechanical arm, ultrasonic scanning system and front end are respectively connected to the backend server for communication.

9. Electronic equipment, It is characterized in that include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the wheel hub intelligent recognition method based on image recognition as described in any one of claims 1-7 when executing the executable instructions.

Images