# US 10,769,774 B2 · Method and device for detecting a defect in a steel plate, as well as apparatus and server therefor > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,769,774 B2 - Document ID: US10769774B2 - Status: Granted - Kind: B2 - Application number: 16/165,518 - Filed: 2018-10-19 - Published: 2019-07-11 - Granted: 2020-09-08 - Earliest claimed: 2018-01-09 - Assignees: Beijing Baidu Netcom Science Technology Co., Ltd. - Applicants: Beijing Baidu Netcom Science and Technology Co., Ltd. - Inventors: Jiabing Leng; Minghao Liu; Yang Liang; Yawei Wen; Faen Zhang; Jiangliang Guo; Jin Tang; Shiming Yin - Web page: https://patentgrep.com/patent/US10769774B2 ## CPC - G06T 7/0004: {Industrial image inspection} - G06F 18/2413: based on distances to training or reference patterns - G06T 7/0008: {checking presence/absence} - G06T 7/001: {using an image reference approach} - G06V 10/454: {Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]} - G06V 10/764: using classification, e.g. of video objects - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 10/95: {structured as a network, e.g. client-server architectures} - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - G06T 2207/30136: Metal - G06V 2201/06: Recognition of objects for industrial automation - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract The application provides a method and a device for detecting a defect in a steel plate, as well as an apparatus and a server therefor. The method for detecting a defect in a steel plate comprises: receiving image data of the steel plate, and generating a defect detection request according to the image data; monitoring computing loads of a plurality of servers, and sending the image data and the defect detection request to a first server; receiving, from the first server, a detection result obtained by calculating the image data using the detection model; and operating according to the detection result, wherein the detection result comprises a selected one of a pass result and a defect result. With the proposed method, the position and the classification of the at least one defect can be obtained, so that the detection accuracy is improved. ## Claims ### Claim 1 A method for detecting a defect in a steel plate, comprising: receiving image data of the steel plate, and generating a defect detection request according to the image data; monitoring computing loads of a plurality of servers, and sending the image data and the defect detection request to a first server, wherein each of the plurality of servers is provided with a detection model, and the first server has a lowest load in the plurality of servers; receiving, from the first server, a detection result obtained by calculating the image data using the detection model; and operating according to the detection result, wherein the detection result comprises a selected one of a pass result and a defect result, wherein the pass result indicates that the steel plate is qualified, and the defect result indicates at least one defect in the steel plate and comprises a position and a classification of the at least one defect, and wherein calculating the image data using the detection model comprises: extracting a steel plate feature from the image data using a deep convolutional neural network; and obtaining the detection result according to the extracted steel plate feature. ### Claim 2 (depends on 1) The method according to claim 1, wherein the method further comprises: sending an updating instruction to the first server, wherein the updating instruction instructs the first server to update the detection model according to the image data and the detection result, and to send the updated detection model to other servers of the plurality of servers. ### Claim 3 (depends on 1) An apparatus, comprising: one or more processors; a memory configured to store one or more programs; and a communication interface configured to cause the processor and the memory to communicate with an external apparatus; wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method of claim 1. ### Claim 4 (depends on 1) A non-transitory computer readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method of claim 1. ### Claim 5 A method for detecting a defect in a steel plate, comprising: receiving image data of the steel plate and a defect detection request; and calculating the image data using a detection model according to the defect detection request to obtain and output a detection result, wherein the detection result comprises a selected one of a pass result and a defect result, wherein the pass result indicates that the steel plate is qualified, and the defect result indicates at least one defect in the steel plate and comprises a position and a classification of the at least one defect, and wherein calculating the image data using a detection model to obtain and output a detection result comprises: extracting a steel plate feature from the image data using a deep convolutional neural network; and obtaining the detection result according to the extracted steel plate feature. ### Claim 6 (depends on 5) The method according to claim 5, wherein calculating the image data using a detection model to obtain and output a detection result comprises: in response to determining that there is at least one defect in the steel plate according to the extracted steel plate feature, determining the position and the classification of the at least one defect using a preset classifier, and generating the defect result according to the position and the classification as the detection result; and in response to determining that the steel plate is qualified according to the extracted steel plate feature, generating the pass result as the detection result. ### Claim 7 (depends on 6) The method according to claim 6, wherein before extracting a steel plate feature from the image data using a deep convolutional neural network, the method further comprises: pre-processing the image data; and wherein extracting the steel plate feature from the image data using the deep convolutional neural network comprises: extracting the steel plate feature from the pre-processed image data using the deep convolutional neural network. ### Claim 8 (depends on 6) The method according to claim 6, wherein the steel plate feature is obtained by training through images of a plurality of steel plates, and is updated according to a predetermined time interval or according to an updating instruction, wherein the updating instruction instructs to update the detection model and the steel plate feature according to the image data and the detection result, and the plurality of steel plates are of the same size and formed in the same process. ### Claim 9 (depends on 6) The method according to claim 6, wherein determining that there is at least one defect in the steel plate according to the extracted steel plate feature comprises: comparing the extracted steel plate feature with a steel plate feature of a qualified steel plate, and determining at least one defect in the steel plate according to a comparison result. ### Claim 10 (depends on 5) A server, comprising: one or more processors; a memory configured to store one or more programs; and a communication interface configured to cause the processor and the memory to communicate with an external apparatus; wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method of claim 5. ### Claim 11 (depends on 5) A non-transitory computer readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method of claim 5. ### Claim 12 A device for detecting a defect in a steel plate, comprising: a data receiving module configured for receiving image data of the steel plate, and generating a defect detection request according to the image data; a data sending module configured for monitoring computing loads of a plurality of servers, and sending the image data and the defect detection request to a first server, wherein each of the plurality of servers is provided with a detection model, and the first server has a lowest load in the plurality of servers; a result receiving module configured for receiving, from the first server, a detection result obtained by calculating the image data using the detection model; and an operating module configured for operating according to the detection result, wherein the detection result comprises a selected one of a pass result and a defect result, wherein the pass result indicates that the steel plate is qualified, and the defect result indicates at least one defect in the steel plate and comprises a position and a classification of the at least one defect, and wherein calculating the image data using the detection model comprises: extracting a steel plate feature from the image data using a deep convolutional neural network; and obtaining the detection result according to the extracted steel plate feature. ### Claim 13 (depends on 12) The device according to claim 12, wherein the device further comprises: an instruction sending module configured for sending an updating instruction to the first server, wherein the updating instruction instructs the first server to update the detection model according to the image data and the detection result, and to send the updated detection model to other servers of the plurality of servers. ### Claim 14 A device for detecting a defect in a steel plate, comprising: a receiving module configured for receiving image data of the steel plate and a defect detection request; and a calculating module configured for calculating the image data using a detection model according to the defect detection request to obtain and output a detection result, wherein the detection result comprises a selected one of a pass result and a defect result, wherein the pass result indicates that the steel plate is qualified, and the defect result indicates at least one defect in the steel plate and comprises a position and a classification of the at least one defect, and wherein the calculating module comprises: a feature extracting unit configured for extracting a steel plate feature from the image data using a deep convolutional neural network; and a determining unit configured to obtain the detection result according to the extracted steel plate feature. ### Claim 15 (depends on 14) The device according to claim 14, wherein the determining unit comprises: a first determining unit configured to, in response to determining that there is at least one defect in the steel plate according to the extracted steel plate feature, determine the position and the classification of the at least one defect using a preset classifier, and generating the defect result according to the position and the classification as the detection result; and a second determining unit configured to, in response to determining that the steel plate is qualified according to the extracted steel plate feature, generate the pass result as the detection result. ### Claim 16 (depends on 15) The device according to claim 15, wherein the calculating module further comprises: a pre-processing unit configured for pre-processing the image data; and wherein: the feature extracting unit is further configured for extracting the steel plate feature from the pre-processed image data using the deep convolutional neural network. ### Claim 17 (depends on 15) The device according to claim 15, wherein the steel plate feature is obtained by training through images of a plurality of steel plates, and is updated according to a predetermined time interval or according to an updating instruction, and wherein the updating instruction instructs to update the detection model and the steel plate feature according to the image data and the detection result, and the plurality of steel plates are of the same size and formed in the same process. ### Claim 18 (depends on 15) The device according to claim 15, wherein the first determining unit is further configured for comparing the extracted steel plate feature with a steel plate feature of a qualified steel plate, and determining at least one defect in the steel plate according to a comparison result.