Hardmeter capable of automatically identifying and planning welding seam and inlaid material detection point positions

Abstract

The invention relates to the technical field of image recognition, and discloses a durometer capable of automatically recognizing and planning detection point positions of a welding seam and a mosaic material, which comprises the following steps: S1, image acquisition and preprocessing: acquiring images of the welding seam and the mosaic material, S2, image feature information extraction, S3, detection point position calculation, and S4, detection point position calculation. The method comprises the following steps: S1, pre-processing an image, performing feature extraction on the pre-processed image by using a convolutional neural network in a deep learning algorithm, S3, automatically planning a detection point location, automatically planning the detection point location in a welding seam and mosaic material region according to identified feature information of the welding seam and the mosaic material, and S4, executing hardness detection. According to the method, feature information in the obtained image can be extracted through the established convolutional neural network algorithm, and edge curve recognition can be performed on the image, so that accurate contours and key feature points of the welding seam and the mosaic material are obtained, and compared with existing manual recognition, the accuracy and efficiency of image recognition are remarkably improved, and the recognition efficiency of the welding seam and the mosaic material is improved. And the labor cost in the identification process is also reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a hardness tester that can automatically identify and plan the detection points of welds and inlaid materials. Background Art

[0002] A hardness tester is an instrument used to measure the hardness of materials. Hardness, as an important mechanical property index of materials, reflects their ability to resist local deformation, especially plastic deformation, indentation, or scratching.

[0003] There are various types of hardness testers. Common ones include Brinell hardness testers, which determine the hardness value by pressing a hard alloy ball with a certain diameter into the surface of the material and based on the size of the indentation diameter. It is suitable for measuring relatively soft metals such as annealed and normalized metals; Rockwell hardness testers use a diamond cone with a vertex angle of 120° or a steel ball with a diameter of 1.588 mm as the indenter and calculate the hardness according to the indentation depth, and can quickly measure relatively hard materials such as quenched steel; Vickers hardness testers use a regular square pyramid diamond indenter and calculate the hardness based on the length of the indentation diagonal. It has a wide test range and high precision and can be used to measure thin parts, coatings, etc. In many fields such as mechanical manufacturing, metal processing, and aerospace, hardness testers play a key role in helping to ensure the quality of materials and workpieces.

[0004] During the actual use process, the surface information of the object to be measured will be recorded by the CCD vision imaging device on the inner surface of the hardness tester, and the positions of welds or inlaid materials will be marked through manual marking and finally the detection points will be planned. However, during the marking and planning process, due to the subjective influence of humans, there is a risk of deviation in determining the detection points, which affects the accuracy and efficiency of hardness detection, and the overall process takes a long time, resulting in a long detection time for the object to be measured. In view of this, we propose a hardness tester that can automatically identify and plan the detection points of welds and inlaid materials. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a hardness tester that can automatically identify and plan the detection points of welds and inlaid materials, and solves the problem that during the marking and planning process, due to the subjective influence of humans, there is a risk of deviation in determining the detection points, which affects the accuracy and efficiency of hardness detection.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A hardness tester that can automatically identify and plan the detection points of welds and inlaid materials, including the following steps:

[0007] S1: Image acquisition and preprocessing

[0008] Use a CCD vision detection device to collect images of welds and inlaid materials, obtain image data information including surface texture and shape, and at the same time preprocess the collected images;

[0009] S2: Image Feature Information Extraction

[0010] Use the convolutional neural network in the deep learning algorithm to extract features from the preprocessed image, and obtain the edge, contour, and texture feature information of the weld and the inlaid material;

[0011] S3: Automatic Detection Point Planning

[0012] According to the identified feature information of the weld and the inlaid material, combined with the detection rules and algorithms, automatically plan the detection points in the weld and inlaid material areas, and generate the detection point layout;

[0013] S4: Hardness Detection Execution

[0014] The hardness tester automatically moves to the corresponding position for hardness detection according to the generated detection point layout, and records the detection data.

[0015] Preferably, in the S1 image acquisition and preprocessing step, the CCD vision detection device acquires the high-resolution surface images of the weld and the inlaid material.

[0016] Preferably, in the S2 image feature information extraction step, the convolutional neural network includes multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer performs convolutional operations through convolutional kernels of different sizes to extract different levels of features of the image. The pooling layer uses the maximum pooling or average pooling method to reduce the dimensionality of the feature map. The fully connected layer integrates the features after convolution and pooling and outputs the feature vector.

[0017] Preferably, in the S2 image feature extraction and recognition step, the transfer learning technique is adopted. The model parameters pre-trained on a large-scale image dataset are used to initialize the convolutional neural network, and the trained model is fine-tuned on the weld and inlaid material image dataset to simplify the adjustment steps.

[0018] Preferably, in the S2 image feature extraction and recognition step, a generative adversarial network is introduced to enhance the image recognition ability. The generator generates image data similar to the weld and the inlaid material, which is used to expand the training dataset and alleviate the data shortage. The discriminator confronts the generator and at the same time assists the convolutional neural network in feature representation during the training process.

[0019] Preferably, in the S3 automatic detection point planning step, the density-based spatial clustering algorithm is used to perform clustering analysis on the feature points in the weld and inlaid material images. According to the clustering results, the detection point distribution in different regions is determined, and the detection points are kept more concentrated in the key feature regions.

[0020] Preferably, for the S3 automatic detection point planning of the inlay material, according to its material information, it is matched with the edge blocks in the acquired image, and weighted determination is carried out according to the detection point planning rules assisted for different material regions respectively.

[0021] Preferably, the S3 automatic detection point planning uses finite element analysis and historical data to obtain the stress distribution of the weld. In the stress concentration areas, including the starting point and the corner of the weld, the detection points are planned at a smaller interval than other areas.

[0022] Preferably, during the execution of the S4 hardness detection, multiple groups of sensors are equipped inside the hardness tester, which can measure the hardness values of the weld and the inlay material. At the same time, the hardness tester also has the functions of data storage and transmission, and can upload the detection data to the computer or cloud server in real time.

[0023] Preferably, during the execution of the S4 hardness detection, the detection head of the hardness tester has the functions of automatic positioning and adjustment, and can move to each detection point according to the coordinate information of the detection point layout, and automatically adjust the detection angle and pressure.

[0024] The present invention provides a hardness tester that can automatically identify and plan the detection points of the weld and the inlay material, having the following beneficial effects:

[0025] 1. Through the established convolutional neural network algorithm, the present invention can extract the feature information in the acquired image and identify the edge curves of the image, so as to obtain the accurate contours and key feature points of the weld and the inlay material. Compared with the existing manual identification, the accuracy and efficiency of image recognition are significantly improved, and the labor cost in the recognition process is also reduced.

[0026] 2. Through the detection point planning established by edge block matching and weighted determination, the present invention can match the material information of the actual product and the edge block information, and carry out weighted determination according to the detection point planning rules corresponding to different materials, so as to determine the detection point distribution of each region finally, making the detection points more reasonable, avoiding the subjectivity and deviation of manual planning, and improving the accuracy and reliability of hardness detection.

[0027] 3. Through the detection point planning introducing finite element analysis and historical data stress distribution analysis, the present invention can carry out more refined detection point planning for the stress concentration areas of the weld, such as the starting point and the corner, to ensure that these key areas are fully detected, further improving the comprehensiveness and accuracy of hardness detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flowchart of the method for automatically identifying and planning the detection points of the weld and the inlay material;

[0029] Figure 2 Schematic diagram of the detection point planning process for edge block matching and weighted determination of the present invention;

[0030] Figure 3 Schematic diagram of the weld example of the present invention. Specific implementation manners

[0031] Next, in combination with the drawings in the specification of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] Embodiment:

[0033] Please refer to the attached Figure 1 - attached Figure 3 , the embodiment of the present invention provides a hardness tester for automatically identifying and planning the detection points of welds and inlaid materials, including the following steps:

[0034] S1: Image acquisition and preprocessing

[0035] Use a CCD vision detection device to collect images of welds and inlaid materials, obtain image data information including surface texture and shape, and at the same time preprocess the collected images;

[0036] S2: Image feature information extraction

[0037] Use a convolutional neural network in the deep learning algorithm to extract features from the preprocessed images, and obtain the edge, contour, and texture feature information of welds and inlaid materials;

[0038] S3: Automatic detection point planning

[0039] According to the identified feature information of welds and inlaid materials, combined with detection rules and algorithms, automatically plan detection points in the areas of welds and inlaid materials to generate a detection point layout;

[0040] S4: Hardness detection execution

[0041] The hardness tester automatically moves to the corresponding position for hardness detection according to the generated detection point layout and records the detection data.

[0042] In the step S1 of image acquisition and preprocessing, the CCD vision detection device collects high-resolution surface images of welds and inlaid materials.

[0043] In the S2 image feature information extraction step, the convolutional neural network includes multiple convolutional layers, pooling layers and fully connected layers. The convolutional layer performs convolution operations through convolution kernels of different sizes to extract different levels of image features. The pooling layer uses maximum pooling or average pooling to reduce the dimension of the feature map. The fully connected layer integrates the features after convolution and pooling and outputs a feature vector. The convolutional neural network algorithm established here is as follows:

[0044] Convolutional Layer:

[0045] Function: The convolution layer is the core component of CNN. It extracts local features of the image by sliding the convolution kernel on the input image or feature map. Convolution kernels of different sizes can capture features of different scales. Small convolution kernels are suitable for extracting detailed features, while large convolution kernels are better at capturing overall structural features.

[0046] Algorithm formula: Let the input feature map be X and the size be H in ×W in ×C in (Height H in , Width W in 、Number of channels C in ), the size of the convolution kernel K is h×w×C in ×C out (height h, width w, number of input channels C in , Output channel number C out ), with a bias of b and a size of C out The size of the output feature map Y is H out ×W out ×C out , then the formula for the convolution operation is:

[0047]

[0048] where p = 0, 1, ..., H out -1,q=0,1,…,W out -1,r=0,1,…,C out -1;

[0049] generally,

[0050]

[0051] Here paddin is the padding size, which is used to control the size of the output feature map, and stride is the step size, which determines the stride of the convolution kernel sliding on the input feature map;

[0052] Pooling layer:

[0053] Function: The pooling layer is mainly used to reduce the dimensionality of the feature map, reduce the amount of data, and at the same time retain important feature information. It reduces the spatial dimensions (height and width) of the feature map by performing pooling operations on local regions, thereby accelerating the calculation speed and preventing overfitting. Common pooling methods include max pooling and average pooling;

[0054] Max pooling algorithm formula: Let the input feature map be X, with size H in ×W in ×C, the pooling window size is h×w, and the stride is s. The size of the output feature map Y is H out ×W out ×C, where:

[0055]

[0056] Then the max pooling operation formula is:

[0057]

[0058] where p = 0, 1, …, H out -1, q = 0, 1, …, W out -1, r = 0, 1, …, C - 1.

[0059] Average pooling algorithm formula: Replace the operation of taking the maximum value with the operation of calculating the average value, that is:

[0060]

[0061] Fully connected layer:

[0062] Function: The fully connected layer integrates the features extracted by the convolutional layer and the pooling layer, converts the multi-dimensional feature vector into a one-dimensional vector, and performs a linear transformation through the weight matrix and the bias vector, and finally outputs the result for classification or regression. In an image classification task, the output of the fully connected layer usually passes through a Softmax function to convert it into the probability distribution of each category;

[0063] Algorithm formula: Let the dimension of the input vector x be d in , the dimension of the weight matrix W is d out ×d in , the dimension of the bias vector b is d out , then the output y of the fully connected layer is:

[0064] y = Wx + b

[0065] where the dimension of y is d out

[0066] Example

[0067] Scenario setting: There is an image containing weld seams and inlaid materials. The weld seams present curves of different shapes, and the inlaid materials also have different contours. Different regions need to be divided by identifying these edge curves;

[0068] Data preparation: 200 similar images were collected. Each image was manually labeled to mark the regions enclosed by the edge curves of the weld seams and inlaid materials, and these images were divided into a training set (160 images) and a test set (40 images);

[0069] CNN model construction:

[0070] Convolutional layer 1: Use a 3x3 convolutional kernel, with a stride of 1 and padding of 1. The number of input channels is 3 (RGB image), and the number of output channels is 32. This layer is used to initially extract the edge features of the image;

[0071] Pooling layer 1: Adopt max pooling, with a pooling window size of 2x2 and a stride of 2 to reduce the dimension of the output of convolutional layer 1;

[0072] Convolutional layer 2: Use a 5x5 convolutional kernel, with a stride of 1 and padding of 2. The number of input channels is 32, and the number of output channels is 64. This layer aims to capture more complex edge curve features;

[0073] Pooling layer 2: Again, adopt max pooling, with a pooling window size of 2x2 and a stride of 2;

[0074] Fully connected layer 1: Flatten the feature map output by pooling layer 2 into a one-dimensional vector and input it into fully connected layer 1, which has 128 neurons;

[0075] Fully connected layer 2: Fully connected layer 2 has 2 neurons, and the two output values are respectively used to represent the probability that each pixel point in the image belongs to the edge curve (such as the weld edge or inlaid material edge) and the probability of belonging to the background;

[0076] Training process: Use the cross-entropy loss function to measure the difference between the model prediction value and the true label, adopt the Adam optimizer, and set the learning rate to 0.001; during the training process, input the images in the training set into the model in turn, calculate the prediction results through forward propagation, then calculate the gradient according to the loss function, and update the model parameters (convolutional kernel weights, fully connected layer weights, and biases) through backpropagation, and continuously iterate and train until the loss function converges;

[0077] Prediction and region division: Assume there is a new test image. After preprocessing, it is input into the trained model. The image extracts features through the convolutional layer and pooling layer in turn, and then makes predictions through the fully connected layer. The model outputs the probabilities that each pixel point belongs to the edge curve and the background. Set a threshold, and the pixel points with probabilities greater than the threshold are considered to belong to the edge curve;

[0078] Edge curve drawing and region division: Based on the predicted edge curve pixel points, use image processing algorithms (such as contour tracking algorithms) to draw continuous edge curves. These edge curves divide the image into different regions, such as the weld region, the inlay material region, and the background region; for example, for the weld region, the region surrounded by the identified weld edge curve is the weld region; for the inlay material, its location region is also determined according to its edge curve. In this way, the task of dividing the image region by identifying the edge curves on the image is completed.

[0079] In the S2 image feature extraction and recognition step, transfer learning technology is adopted. The model parameters pre-trained on a large-scale image dataset are used to initialize the convolutional neural network, and the trained model is fine-tuned on the weld and inlay material image datasets to simplify the adjustment steps.

[0080] In the S2 image feature extraction and recognition step, a generative adversarial network is introduced to enhance the image recognition ability. The generator generates image data similar to the weld and inlay materials to expand the training dataset and alleviate the data shortage; the discriminator competes with the generator and at the same time assists the convolutional neural network in feature representation during the training process.

[0081] In the S3 automatic detection point planning step, the density-based spatial clustering algorithm is used to perform clustering analysis on the feature points in the weld and inlay material images. According to the clustering results, the distribution of detection points in different regions is determined, and the detection points are kept more concentrated in the key feature regions. The detection point planning process based on the matching and weighted determination of material information and edge blocks is established as follows:

[0082] First, encode the material information of the inlay material, and then obtain the edge blocks in the image through the edge detection algorithm. Match the material information with the edge blocks, and perform weighted determination according to the detection point planning rules corresponding to different materials to determine the distribution of detection points in each region.

[0083] Algorithm steps:

[0084] Material information encoding: For example, encode aluminum alloy as [1,0,0], copper alloy as [0,1,0], and other materials as [0,0,1], etc.;

[0085] Edge detection: Use the convolutional neural algorithm established in S2 to obtain the edge information in the image and get the edge blocks;

[0086] Matching and weighted determination: For each edge block, calculate its similarity with different material encodings. Assume that the similarity calculation function is sim(m,e), where m is the material encoding vector and e is the feature vector extracted from the edge block.

[0087] According to the similarity results, weighting is performed according to the detection point planning rules for different materials. For example, for aluminum alloy materials, the planning rule may be to set n 1 detection points per unit area, and for copper alloy materials, it is to set n 2 detection points per unit area. Let the similarity be s 1 (similarity between aluminum alloy and the edge block), s 2 (similarity between copper alloy and the edge block). Then the detection point density d of this edge block is:

[0088] d = s 1 × n 1 + s 2 × n 2

[0089] Suppose there is an inlay material area. Through material analysis, it may be aluminum alloy or copper alloy. After edge detection, an edge block is obtained, and the feature vector e = [0.8, 0.2, 0] is extracted from this edge block.

[0090] Aluminum alloy code m 1 = [1, 0, 0], copper alloy code m 2 = [0, 1, 0].

[0091] Calculate the similarity s 1 = sim(m 1 , e). Suppose the cosine similarity formula is used Then

[0092]

[0093] Suppose the detection point planning rule for aluminum alloy is to set 5 detection points per 10 mm 2 (n 1 = 0.5 per mm 2 ), and for copper alloy, it is to set 3 detection points per 10 mm 2 (n 2 = 0.3 per mm 2 );

[0094] Then the detection point density d of this edge block is d = 0.97×0.5 + 0.24×0.3 = 0.485 + 0.072 = 0.557 per mm 2 . According to this density, detection points are planned in the area where this edge block is located.

[0095] For the automatic planning of the S3 detection points, for the inlaid material, according to its material information, it is matched with the edge blocks in the acquired image, and the weighted determination is carried out according to the detection point planning rules assisted by different material regions respectively.

[0096] The automatic planning of the S3 detection points uses finite element analysis and historical data to obtain the stress distribution of the weld. In the stress concentration areas, including the starting point and the corner of the weld, the detection points are planned at a smaller spacing than other areas. The detection point planning in the weld stress concentration area based on finite element analysis and historical data is as follows:

[0097] Finite element analysis: Establish a finite element model of the weld, apply the corresponding loads and boundary conditions, solve the stress distribution. Assume that the stress distribution function σ(x,y) is obtained through finite element analysis, representing the stress value of the weld at the coordinate (x,y);

[0098] Historical data correction: According to the positions and stress values of the stress concentration areas in the historical detection data, correct the finite element analysis results. For example, if the historical data shows that at a certain specific starting point position of the weld, the actual stress is k times higher than the finite element analysis result, then the corrected stress distribution function is σ 修正 (x,y) = σ(x,y) × f(x,y), where f(x,y) is the correction coefficient obtained from the historical data;

[0099] Detection point planning: Set the stress concentration threshold σ 阈值，对于 σ 修正 (x,y) ≥ σ 阈值 The area is defined as the stress concentration area. In the stress concentration area, set the detection point spacing as d 1 ; in other areas, the detection point spacing is d 2 , and d 1 < d 2 ;

[0100] Example:

[0101] Suppose the weld is a rectangular shape with a length of L = 100mm and a width of W = 50mm. The simplified example obtained through finite element analysis is:

[0102] The stress distribution function σ(x,y) = 0.01x + 0.02y;

[0103] According to the historical data, near the starting point (0,0) of the weld, the actual stress is 1.5 times higher than the finite element analysis result, and the correction coefficient is 1 at other positions. Then the corrected stress distribution function is:

[0104]

[0105] Set the stress concentration threshold σ阈值 = 1;

[0106] For the region where x < 10 and y < 10, σ 修正 (x, y) = 1.5×(0.01x + 0.02y), when x = 5, y = 5;

[0107] σ 修正 (5, 5) = 1.5×(0.01×5 + 0.02×5) = 1.5×0.15 = 0.225 < 1;

[0108] The stress concentration condition is not satisfied. Assume that near the weld corner (0, 50), the corrected stress satisfies σ 修正 (x, y) ≥ σ 阈值 , which is defined as the stress concentration region. Let d 1 = 5mm, d 2 = 10mm. In the stress concentration region, a detection point is planned every 5mm; in other regions, a detection point is planned every 10mm.

[0109] During the execution of the S4 hardness detection, multiple groups of sensors are equipped inside the hardness tester, which can measure the hardness values at the weld and the inlaid material. At the same time, the hardness tester also has the functions of data storage and transmission, and can upload the detection data to the computer or cloud server in real time.

[0110] During the execution of the S4 hardness detection, the detection head of the hardness tester has the functions of automatic positioning and adjustment, and can move to each detection point according to the coordinate information of the detection point layout, and automatically adjust the detection angle and pressure.

[0111] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A hardness tester that automatically identifies and plans the detection points of welds and inlaid materials, Characterized in that, It includes the following steps: S1: Image acquisition and preprocessing Use a CCD vision detection device to collect images of welds and inlaid materials, obtain image data information including surface texture and shape, and at the same time preprocess the collected images; S2: Image feature information extraction Use a convolutional neural network in the deep learning algorithm to extract features from the preprocessed images, and obtain the edge, contour, and texture feature information of welds and inlaid materials; S3: Automatic planning of detection points According to the identified feature information of welds and inlaid materials, combined with detection rules and algorithms, automatically plan detection points in the weld and inlaid material areas, and generate a detection point layout; S4: Hardness detection execution The hardness tester automatically moves to the corresponding position for hardness detection according to the generated detection point layout and records the detection data.

2. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, Characterized in that, In the S1 image acquisition and preprocessing step, the CCD vision detection device collects high-resolution surface images of welds and inlaid materials.

3. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, Characterized in that, In the S2 image feature information extraction step, the convolutional neural network includes multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers perform convolutional operations through convolutional kernels of different sizes to extract different levels of features of the images. The pooling layers use max pooling or average pooling methods to reduce the dimensionality of the feature maps. The fully connected layers integrate the features after convolution and pooling and output feature vectors.

4. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, Characterized in that, In the S2 image feature extraction and recognition step, transfer learning technology is adopted. The model parameters pre-trained on a large-scale image dataset are used to initialize the convolutional neural network, and the trained model is fine-tuned on the weld and inlaid material image dataset to simplify the adjustment steps.

5. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, Characterized in that, In the S2 image feature extraction and recognition step, a generative adversarial network is introduced to enhance the image recognition ability. The generator generates image data similar to welds and inlaid materials to expand the training dataset and alleviate the data shortage; the discriminator competes with the generator and at the same time assists the convolutional neural network in feature representation during the training process.

6. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, Characterized in that, In the S3 automatic planning of detection points step, a density-based spatial clustering algorithm is used to perform clustering analysis on the feature points in the weld and inlaid material images, and the detection point distribution in different regions is determined according to the clustering results, so that the detection points are more concentrated in the key feature regions.

7. The hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1, It is characterized in that for the inlaid material in the automatic planning of the S3 detection points, according to its material information, it is matched with the edge blocks in the acquired image, and weighted determination is carried out according to the detection point planning rules assisted by different material regions respectively.

8. A hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1 It is characterized in that in the automatic planning of the S3 detection points, finite element analysis and historical data are used to obtain the stress distribution of the weld. In the stress concentration areas, including the starting point and corner of the weld, the detection points are planned at a smaller interval than other areas.

9. A hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1 It is characterized in that during the execution of the S4 hardness detection, multiple groups of sensors are equipped inside the hardness tester, which can measure the hardness values at the welds and inlaid materials. At the same time, the hardness tester also has the functions of data storage and transmission, and can upload the detection data to a computer or a cloud server in real time.

10. A hardness tester for automatically identifying and planning the detection points of welds and inlaid materials according to claim 1 It is characterized in that during the execution of the S4 hardness detection, the detection head of the hardness tester has the functions of automatic positioning and adjustment, and can move to each detection point according to the coordinate information of the detection point layout, and automatically adjust the detection angle and pressure.

Images