Parts offset angle detection method

Abstract

The present application discloses a method for detecting the offset angle of a part, which uses a camera to collect a part image from directly above the part, performs edge detection on it to obtain a part contour image, extracts a part standard contour image from a database, and uses image processing technology based on deep learning to perform contour feature extraction and contour saliency processing on the part contour image and the part standard contour image, and then intelligently estimates the part offset angle based on the feature difference between the two. In this way, the accuracy and efficiency of part offset angle detection can be effectively improved, production costs can be reduced, and product quality can be improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a method for detecting a part offset angle. Background Art

[0002] In modern manufacturing, especially in high-end manufacturing fields such as automobiles and aviation, the precise installation and positioning of parts is one of the key factors to ensure product quality and production efficiency. However, in the actual production process, due to equipment wear, environmental factors, improper operation and other reasons, parts may shift during assembly, welding or installation, resulting in an increase in product failure rate and reduced production efficiency.

[0003] In order to ensure the precise installation of parts, traditional methods of detecting the offset angle of parts usually rely on manual inspection or mechanical measuring equipment. However, manual inspection methods are not only time-consuming and laborious, but also greatly affected by human factors, making it difficult to ensure the accuracy and consistency of the inspection results. Some existing inspection equipment, such as X-ray detectors, three-coordinate measuring machines, and optical projectors, are not only expensive, but also complex to operate and have high maintenance costs, making them difficult to be widely used in large-scale production lines.

[0004] Therefore, an optimized method for detecting part offset angle is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method for detecting the offset angle of a part, which uses a camera to collect a component image from directly above the component, and performs edge detection on it to obtain a component contour image, and at the same time extracts a component standard contour image from a database, and uses image processing technology based on deep learning to extract contour features and perform contour saliency processing on the component contour image and the component standard contour image, and then intelligently estimates the offset angle of the part based on the feature difference between the two. In this way, the accuracy and efficiency of the part offset angle detection can be effectively improved, the production cost can be reduced, and the product quality can be improved.

[0006] According to one aspect of the present application, a method for detecting a part offset angle is provided, which comprises:

[0007] Take an image of the component from directly above the component;

[0008] Performing edge detection on the component image to obtain a component contour image;

[0009] Extracting component standard outline images from a database;

[0010] Performing contour feature extraction and feature gradient enhancement on the component contour image and the component standard contour image to obtain a component contour significant detection feature map and a component contour significant standard reference feature map;

[0011] Based on the feature comparison between the component contour significant detection feature map and the component contour significant standard reference feature map, a part offset angle decoding estimate is generated.

[0012] Compared with the prior art, the present application provides a method for detecting the offset angle of a part, which uses a camera to collect a part image from directly above the part, performs edge detection on it to obtain a part contour image, extracts a part standard contour image from a database, and uses deep learning-based image processing technology to perform contour feature extraction and contour saliency processing on the part contour image and the part standard contour image, and then intelligently estimates the part offset angle based on the feature difference between the two. In this way, the accuracy and efficiency of part offset angle detection can be effectively improved, production costs can be reduced, and product quality can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0014] Figure 1 is a flow chart of a method for detecting a part offset angle according to an embodiment of the present application;

[0015] Figure 2 A data flow diagram of a method for detecting a part offset angle according to an embodiment of the present application;

[0016] Figure 3 This is a flowchart of sub-step S4 of the part offset angle detection method according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0018] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0019] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0020] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0022] In order to ensure the precise installation of parts, traditional methods of detecting the offset angle of parts usually rely on manual inspection or mechanical measuring equipment. However, manual inspection methods are not only time-consuming and laborious, but also greatly affected by human factors, making it difficult to ensure the accuracy and consistency of the inspection results. Some existing inspection equipment, such as X-ray detectors, three-coordinate measuring machines, and optical projectors, are not only expensive, but also complex to operate and have high maintenance costs, making them difficult to be widely used in large-scale production lines. Therefore, an optimized method for detecting the offset angle of parts is expected.

[0023] In the technical solution of the present application, a method for detecting the offset angle of a part is proposed. Figure 1 Flow chart of a method for detecting part offset angle according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a method for detecting a part offset angle according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the part offset angle detection method according to the embodiment of the present application includes the steps of: S1, taking a component image from directly above the component; S2, performing edge detection on the component image to obtain a component contour image; S3, extracting a component standard contour image from a database; S4, performing contour feature extraction and feature gradient enhancement on the component contour image and the component standard contour image to obtain a component contour significant detection feature map and a component contour significant standard reference feature map; S5, generating a part offset angle decoding estimate based on feature comparison between the component contour significant detection feature map and the component contour significant standard reference feature map.

[0024] In particular, the S1, captures the component image from directly above the component. It should be understood that capturing from directly above can ensure that the camera viewing angle is consistent with the vertical direction of the component, thereby minimizing the error introduced by the tilt or offset of the viewing angle, helping to more accurately reflect the actual shape and contour of the component, and making the offset angle detection of the component more accurate.

[0025] In particular, the S2 performs edge detection on the component image to obtain a component contour image. In a specific example of the present application, the component image is processed using the Canny edge detection operator to obtain the component contour image. It should be understood that in the detection of part offset angles, the contour information of the component is the basis for estimating the offset angle. In view of the fact that the component image may be affected by factors such as ambient light and component surface material, the contour of the component may not appear clear enough. To this end, the present application further uses the Canny edge detection operator to process the component image. Those of ordinary skill in the art should know that the Canny edge detection operator has good edge detection performance and noise resistance. Through steps such as Gaussian filtering, gradient calculation, non-maximum suppression and hysteresis threshold, it can effectively detect edge information in the image, while suppressing noise interference, so as to obtain a component contour image with clear contour information, thereby providing accurate contour data for subsequent offset angle detection.

[0026] Accordingly, in a possible implementation, the component image can be processed using the Canny edge detection operator to obtain the component contour image through the following steps, for example: converting the component image into a grayscale image; smoothing the image using a Gaussian filter to remove noise; using the Sobel operator to calculate the horizontal and vertical gradients of the image; for each pixel, only retaining the gradient value in the direction with the largest gradient amplitude; using two thresholds to threshold the gradient amplitude. Pixels below the lower threshold are set to zero, and pixels above the higher threshold are set to the maximum value. Pixels between the two thresholds are set to zero unless they are adjacent to pixels above the higher threshold; using connected component analysis to connect edge pixels into contours to obtain the component contour image.

[0027] In particular, the step S3 extracts a component standard contour image from a database. It should be understood that the standard contour image is a component contour image obtained under ideal conditions, which has a high degree of accuracy and clarity, and can be used as a reference standard for offset angle detection to achieve fast and accurate offset angle detection.

[0028] In particular, the S4 performs contour feature extraction and feature gradient enhancement on the component contour image and the component standard contour image to obtain a component contour salient detection feature map and a component contour salient standard reference feature map. In particular, in a specific example of the present application, Figure 3 As shown, the S4 includes: S41, respectively extracting the image features of the component contour image and the component standard contour image to obtain a component contour detection feature map and a component contour standard reference feature map; S42, inputting the component contour detection feature map and the component contour standard reference feature map into a feature distribution gradient mask salient device to obtain the component contour salient detection feature map and the component contour salient standard reference feature map.

[0029] Specifically, the S41 extracts the image features of the component contour image and the component standard contour image respectively to obtain a component contour detection feature map and a component contour standard reference feature map. In a specific example of the present application, the component contour image and the component standard contour image are input into a contour feature extractor based on a hole convolutional neural network model to obtain the component contour detection feature map and the component contour standard reference feature map. In order to ensure that the component contour image and the component standard contour image have the same representation form and dimension in the high-dimensional feature space, so as to facilitate subsequent feature comparison and difference analysis, the present application further inputs the component contour image and the component standard contour image into a common contour feature extractor based on a hole convolutional neural network model for processing, and uses the large receptive field of the hole convolutional neural network to capture the rich component contour information in the image, and further mines the key detail features in the image, such as edges, corners, curves, etc., so as to obtain a component contour detection feature map and a component contour standard reference feature map, which provide a richer information basis for the subsequent offset angle estimation.

[0030] It is worth noting that the dilated convolutional neural network is a special convolutional neural network (CNN) that expands the receptive field by using a dilated convolution operation. The receptive field refers to the area covered by the convolution kernel on the input feature map. The dilated convolution is a modified convolution operation that introduces "holes" in the convolution kernel. The hole refers to the spacing between elements in the convolution kernel. The dilated convolutional neural network has the following advantages: Expanding the receptive field: dilated convolution allows the network to expand the receptive field without increasing the number of parameters. This is very useful for tasks such as semantic segmentation and object detection, which require a large range of contextual information; Maintaining spatial resolution: Unlike pooling layers, dilated convolutions do not reduce the spatial resolution of feature maps. This is very important for tasks that require precise positioning, such as boundary detection.

[0031] Specifically, in S42, the component contour detection feature map and the component contour standard reference feature map are input into the feature distribution gradient mask salient device to obtain the component contour salient detection feature map and the component contour salient standard reference feature map. In order to further improve the contrast between the component contour detection feature and the component contour standard reference feature, thereby improving the accuracy of offset angle detection, the present application further introduces a feature distribution gradient mask salient device to process the component contour detection feature map and the component contour standard reference feature map, and enhances the feature expression of the edge contour position in the feature map by analyzing the feature gradient information. Specifically, the feature distribution gradient mask salient device first quantifies the local change intensity of the pixel at each position by calculating the gradient amplitude value of each position in the input feature map. Then, the gradient amplitude local description operator is calculated by measuring the significance of the gradient amplitude value of each position relative to the gradient amplitude values ​​of other positions in the local neighborhood, and the feature importance of each position is evaluated. Then, the gradient amplitude local description operator of each position is nonlinearly activated using the GELU function to generate a gated mask to guide the enhancement of the contour information in the feature map. Finally, the feature map is weighted element-by-element based on the generated gated mask to focus on the contour-salient area in the feature map and remove background noise to obtain a clearer and more discriminative component contour-salient detection feature map and a component contour-salient standard reference feature map.

[0032] In a specific example of the present application, the component contour detection feature map and the component contour standard reference feature map are input into a feature distribution gradient mask salient device to obtain the component contour salient detection feature map and the component contour salient standard reference feature map, including: calculating the multi-directional gradient value distribution of each position in the component contour detection feature map, and determining the gradient amplitude value of each position in the component contour detection feature map based on the multi-directional gradient value distribution of each position to obtain a component contour detection feature gradient amplitude distribution map; calculating the gradient amplitude local description operator of each position in the component contour detection feature gradient amplitude distribution map to obtain a component contour detection feature gradient amplitude local significant distribution map; inputting the component contour detection feature gradient amplitude local significant distribution map into a gated masker based on a GELU function to obtain a gradient amplitude local significant gated mask map; calculating the position point multiplication between the gradient amplitude local significant gated mask map and the component contour detection feature map to obtain the component contour detection salient feature map.

[0033] Among them, the process of calculating the gradient amplitude local description operator of each position in the component contour detection feature gradient amplitude distribution map to obtain the component contour detection feature gradient amplitude local significant distribution map includes: determining the scale of the local neighborhood, calculating the mean of the difference between the gradient amplitude value of a predetermined position in the component contour detection feature gradient amplitude distribution map and the gradient amplitude values ​​of other positions in the local neighborhood to obtain the gradient amplitude local description operator corresponding to the predetermined position.

[0034] In summary, in the above embodiment, the component contour detection feature map and the component contour standard reference feature map are input into the feature distribution gradient mask salient device to obtain the component contour salient detection feature map and the component contour salient standard reference feature map, including: processing the component contour detection feature map with the following feature distribution gradient mask formula to obtain the component contour detection salient feature map, wherein the feature distribution gradient mask formula is: ;in, , , , , and Respectively represent the grayscale values ​​of the corresponding positions of the component contour detection feature map, Indicates the component contour detection feature map The gradient value in the horizontal direction of the pixel point. Indicates the component contour detection feature map The gradient value in the ordinate direction of the pixel point. The component contour detection feature map is shown in The channel direction gradient value of the position pixel point, The first The gradient magnitude value at the position, Represents the characteristic gradient amplitude distribution map detected by the component contour The set of gradient magnitude values ​​in the local neighborhood centered at position, represents the number of gradient magnitude values ​​in the set of gradient magnitude values ​​in the local neighborhood, , and Respectively represent the offset in the horizontal axis direction, the vertical axis direction and the channel direction, The first The local description operator of the gradient magnitude of the position, express function, The first part of the gradient magnitude local significant gated mask map is represented by The local description operator of the gradient magnitude of the position, represents the local significant gated mask map of the gradient magnitude, It means point multiplication by position. The figure represents a salient feature map for detecting the component contour.

[0035] It is worth mentioning that in other specific examples of the present application, contour feature extraction and feature gradient enhancement can be performed on the component contour image and the component standard contour image in other ways to obtain a component contour salient detection feature map and a component contour salient standard reference feature map, for example: input the component contour image and the component standard contour image; input the component contour image and the component standard contour image; use a contour feature extractor based on a hole convolutional neural network to extract features of the component contour image and the component standard contour image to obtain a component contour detection feature map and a component contour standard reference feature map; calculate the gradient of the component contour detection feature map and the component contour standard reference feature map; use gradient enhancement technology to enhance the saliency information in the feature map; combine the enhanced gradient with the component contour detection feature map and the component contour standard reference feature map to obtain a component contour salient detection feature map and a component contour salient standard reference feature map.

[0036] In particular, the S5 generates a decoded estimated value of the part offset angle based on the feature comparison between the significant detection feature map of the part contour and the significant standard reference feature map of the part contour. In a specific example of the present application, first, the difference feature between the significant detection feature map of the part contour and the significant standard reference feature map of the part contour is extracted to obtain a differential representation feature map of the part offset angle. Specifically, the differential feature map between the significant detection feature map of the part contour and the significant standard reference feature map of the part contour is calculated to obtain the differential representation feature map of the part offset angle. In order to intuitively display and quantify the offset of the part contour compared to the standard contour image, the present application further adopts a differential calculation method to calculate the feature difference between the significant detection feature map of the part contour and the significant standard reference feature map of the part contour position by position to obtain a differential representation feature map of the part offset angle, thereby providing intuitive quantitative data support for the estimation of the offset angle of the part. Further, the differential representation feature map of the part offset angle is input into a part offset angle estimation module based on a decoder to obtain the decoded estimated value of the part offset angle. It should be understood that the decoder is mainly used to decode the encoded feature back to the original space or a specific form. In the technical solution of the present application, the decoder-based part offset angle estimation module can fully utilize the part contour offset information therein by performing multi-layer nonlinear transformation on the part offset angle differential representation feature map, map it to the numerical space of the part offset angle, thereby obtaining a decoded estimate of the part offset angle, so as to achieve accurate estimation of the part offset angle.

[0037] In a preferred example, inputting the part offset angle differential representation feature map into a decoder-based part offset angle estimation module to obtain a part offset angle decoded estimation value comprises the following steps:

[0038] All the eigenvalues ​​of the feature graph represented by the difference in the offset angle of the part are subjected to the eigenvalue-based Clustering of distances and arranging cluster features as cluster vectors representing part offset angle differences;

[0039] Determine a clustering ratio value of the number of eigenvalues ​​of the clustering vector represented by the part offset angle difference and the number of eigenvalues ​​of the characteristic graph represented by the part offset angle difference;

[0040] Dividing the second norm of the part offset angle difference representation cluster vector by the second norm of the part offset angle difference representation feature vector obtained after the part offset angle difference representation feature graph is expanded to obtain a part offset angle difference representation conflict representation value;

[0041] Divide a first power value of a norm of the cluster vector of the part offset angle difference representation with the cluster ratio value as an exponent by a second power value of a norm of the feature vector of the part offset angle difference representation with the cluster ratio value as an exponent to obtain a part offset angle difference representation adversarial representation value;

[0042] For each eigenvalue of the part offset angle difference representation clustering vector, multiply it by the inverse of the difference between the part offset angle difference representation conflict representation value and the part offset angle difference representation confrontation representation value to obtain an optimized eigenvalue of the part offset angle difference representation clustering vector;

[0043] For each feature value outside the cluster in the part offset angle difference representation feature map, multiply it by the reciprocal of the sum of the part offset angle difference representation conflict representation value and the part offset angle difference representation confrontation representation value to obtain an optimized out-of-class feature value of the part offset angle difference representation feature map;

[0044] Combining the optimized eigenvalues ​​of the part offset angle difference representation clustering vector and the optimized out-of-class eigenvalues ​​of the part offset angle difference representation feature graph into an optimized part offset angle difference representation feature graph; and

[0045] The optimized part offset angle differential representation feature map is input into a decoder-based part offset angle estimation module to obtain a part offset angle decoding estimation value.

[0046] It is expressed as: ;in, is the feature vector representing the part offset angle difference, is the number of eigenvalues ​​of the eigenvector representing the part offset angle difference, is the cluster vector representing the part offset angle difference, is the number of eigenvalues ​​of the clustering vector representing the difference in part offset angles, The difference in the offset angle of the part represents the cluster feature set corresponding to the cluster vector, and Represents the binary norm and the uninorm of the vector respectively. Power.

[0047] That is, considering that the component contour salient detection feature map and the component contour salient standard reference feature map respectively represent the image semantic features of the component contour image and the component standard contour image based on the enhancement of the feature distribution gradient saliency, when further calculating the differential feature map therebetween to perform aggregation mapping to the image semantic differential representation domain, the image semantic aggregation alignment conflict will be caused due to the difference in image semantic feature enhancement, thereby causing the loss of aggregation key information and affecting the expression effect of the part offset angle differential representation feature map.

[0048] Therefore, in order to avoid the loss of key suffix semantic information relative to the original feature set as a whole due to aggregation conflict when the part offset angle differential representation feature map is based on aggregated features, the clustering ratio of the number of eigenvalues ​​of the part offset angle differential representation cluster vector relative to the number of eigenvalues ​​of the part offset angle differential representation feature map is used as a decision function to perform an adversarial decision on the set absolute representation of the part offset angle differential representation cluster vector and the one-norm of the part offset angle differential representation feature vector, and positive and negative interactions are performed with the clustering intrinsic conflict representation of the two-norm of the part offset angle differential representation cluster vector and the part offset angle differential representation feature vector, respectively, to construct a solid alignment guardrail between the optimized part offset angle differential representation feature map and the original feature set as a whole based on aggregated features, thereby achieving the mitigation of harmful information loss intention of the optimized part offset angle differential representation feature map based on the aggregation risk transferability, improving the expression effect of the optimized part offset angle differential representation feature map, and thus improving the accuracy of the part offset angle decoding estimate obtained by inputting the part offset angle differential representation feature map into the decoder-based part offset angle estimation module.

[0049] In summary, the part offset angle detection method according to the embodiment of the present application is explained, which uses a camera to collect a part image from directly above the part, and performs edge detection on it to obtain a part contour image, and extracts a part standard contour image from a database, and uses deep learning-based image processing technology to perform contour feature extraction and contour saliency processing on the part contour image and the part standard contour image, and then intelligently estimates the part offset angle based on the feature difference between the two. In this way, the accuracy and efficiency of part offset angle detection can be effectively improved, production costs can be reduced, and product quality can be improved.

[0050] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0051] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0052] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. A method for detecting a part offset angle, characterized in that: include: Take an image of the component from directly above the component; Performing edge detection on the component image to obtain a component contour image; Extracting component standard outline images from a database; Performing contour feature extraction and feature gradient enhancement on the component contour image and the component standard contour image to obtain a component contour significant detection feature map and a component contour significant standard reference feature map; Generate a part offset angle decoding estimate based on feature comparison between the component contour significant detection feature map and the component contour significant standard reference feature map; The process of extracting contour features and enhancing feature gradients on the component contour image and the component standard contour image to obtain a component contour salient detection feature map and a component contour salient standard reference feature map comprises: Respectively extracting image features of the component contour image and the component standard contour image to obtain a component contour detection feature map and a component contour standard reference feature map; Calculating the multi-directional gradient value distribution of each position in the component contour detection feature map, and determining the gradient amplitude value of each position in the component contour detection feature map based on the multi-directional gradient value distribution of each position to obtain a component contour detection feature gradient amplitude distribution map; Calculating the local description operator of the gradient amplitude at each position in the component contour detection feature gradient amplitude distribution map to obtain a component contour detection feature gradient amplitude local significant distribution map; Inputting the local significant distribution map of the feature gradient amplitude of the component contour detection into a gated masker based on the GELU function to obtain a local significant gated mask map of the gradient amplitude; The point-by-point multiplication between the gradient magnitude local salient gated mask map and the component contour detection feature map is calculated to obtain the component contour detection salient feature map.

2. The part offset angle detection method according to claim 1, characterized in that: Performing edge detection on the component image to obtain a component contour image includes: The component image is processed using a Canny edge detection operator to obtain the component contour image.

3. The method for detecting part offset angle according to claim 2, characterized in that: Extracting image features of the component contour image and the component standard contour image respectively to obtain a component contour detection feature map and a component contour standard reference feature map, including: The component contour image and the component standard contour image are input into a contour feature extractor based on a hole convolutional neural network model to obtain the component contour detection feature map and the component contour standard reference feature map.

4. The method for detecting the part offset angle according to claim 3, characterized in that: Calculating the local description operator of the gradient amplitude at each position in the component contour detection feature gradient amplitude distribution map to obtain the component contour detection feature gradient amplitude local significant distribution map, including: The scale of the local neighborhood is determined, and the average of the difference between the gradient amplitude value of a predetermined position in the component contour detection feature gradient amplitude distribution map and the gradient amplitude values ​​of other positions in the local neighborhood is calculated to obtain a gradient amplitude local description operator corresponding to the predetermined position.

5. The part offset angle detection method according to claim 4, characterized in that: Based on the feature comparison between the component contour significant detection feature map and the component contour significant standard reference feature map, a part offset angle decoding estimate is generated, including: Extracting the difference features between the component contour significant detection feature map and the component contour significant standard reference feature map to obtain a part offset angle differential representation feature map; The part offset angle differential representation feature map is input into a decoder-based part offset angle estimation module to obtain the part offset angle decoding estimation value.

6. The part offset angle detection method according to claim 5, characterized in that: Extracting the difference features between the component contour significant detection feature map and the component contour significant standard reference feature map to obtain a part offset angle differential representation feature map, including: The differential feature map between the component contour significant detection feature map and the component contour significant standard reference feature map is calculated to obtain the part offset angle differential representation feature map.

Images