一种基于ResNet网络的工具识别方法

ResNet network-based tool identification method

文档序号：8684 发布日期：2021-09-17 浏览：34次中文

1. A tool identification method based on ResNet network is characterized in that: the method comprises the following steps:

a1: the worker trains and tests the ResNet network;

a2: connecting a plurality of cameras and a computer provided with a ResNet network by a worker through the Internet, wherein each camera is arranged on a tool placing rack;

a3: and D, sequentially photographing and identifying the two-dimensional code by the staff through the plurality of cameras in the step A2 and a computer provided with a ResNet network.

A4: and identifying the two-dimensional code information by the computer provided with the ResNet network, and keeping the record, wherein if the computer is abnormal, the computer provided with the ResNet network records the information and gives an alarm.

2. The ResNet network-based tool identification method of claim 1, wherein: the step A1 comprises the following steps:

a11: a worker divides a plurality of two-dimensional codes into a training set and a test set;

a12: the staff unifies the sizes of the two-dimensional code dataset pictures in the training set and the testing set;

a13: converting the two-dimensional code data set in the training set and the testing set with unified sizes in the step A12 into a TFRecord format by workers;

a14: after extracting the two-dimensional code data set in the training set and the test set which are converted into the TFRecord format in the step A13, the worker constructs a queue mode;

a15: the staff constructs a corresponding ResNet network according to the queue mode constructed in the step A14;

a16: training the ResNet network constructed in the step A15 by a worker through a training set;

a17: and D, testing the ResNet network trained in the step A16 by the staff through the test set, wherein if the test is successful, the ResNet network outputs a test result and then ends, and if the test is failed, the ResNet network directly ends.

3. The ResNet network-based tool identification method of claim 2, wherein: in step a17, if the ResNet network reacts correctly, the ResNet network outputs a test result and then ends, and if the ResNet network reacts incorrectly, the ResNet network ends directly.

4. The ResNet network-based tool identification method of claim 1, wherein: the step A3 includes the following steps:

a31: after the test application software is started, a computer displays a login interface, an operator can log in a software system through a password, and if the software is used for the first time, the operator can log in the system after adding user authority by an administrator; if the operator has the authority, the operator can directly log in, if the password is wrong, the computer can display that the login fails and needs to log in again after being confirmed again by the operator;

a32: after the operator successfully logs in the system, clicking a checking function button displayed on a computer to control cameras on a plurality of tool placement racks to shoot and capture the two-dimensional code;

a33: uploading the screenshots to a computer by a plurality of cameras, if uploading fails, checking whether the network configuration is intact by an operator, and uploading again for calculation after correction is finished, if uploading succeeds, starting to analyze the screenshots by software;

a34: and the test application software starts to recognize the uploaded tool pictures in batch based on the trained ResNet model algorithm, if the recognition fails, an error prompt is popped up, and if the recognition succeeds, all tool information is displayed in a computer interface list.

5. The ResNet network-based tool identification method according to claim 4, wherein: if the registration work in the step A31 is successful, the step A31 is carried out again, and if the registration work is failed, the registration information is popped up for viewing and carrying out the registration work again.

Background

At the present stage, aiming at an open type storehouse tool informatization management scene, the traditional image recognition algorithm based on Opencv is adopted to recognize tools more generally, but the accuracy and the recognition speed are not ideal, and the uniqueness is not solved, so a new method needs to be designed.

Disclosure of Invention

In view of this, the present method aims to provide a tool identification method based on the ResNet network, so as to improve the efficiency and performance of tool identification.

In order to achieve the purpose, the technical scheme of the method is realized as follows:

a tool identification method based on a ResNet network comprises the following steps:

a1: the worker trains and tests the ResNet network;

a2: connecting a plurality of cameras and a computer provided with a ResNet network by a worker through the Internet, wherein each camera is arranged on a tool placing rack;

a3: and D, sequentially photographing and identifying the two-dimensional code by the staff through the plurality of cameras in the step A2 and a computer provided with a ResNet network.

Further, the step a1 includes the following steps:

a11: a worker divides a plurality of two-dimensional codes into a training set and a test set;

a12: the staff unifies the sizes of the two-dimensional code dataset pictures in the training set and the testing set;

a13: converting the two-dimensional code data set in the training set and the testing set with unified sizes in the step A12 into a TFRecord format by workers;

a14: after extracting the two-dimensional code data set in the training set and the test set which are converted into the TFRecord format in the step A13, the worker constructs a queue mode;

a15: the staff constructs a corresponding ResNet network according to the queue mode constructed in the step A14;

a16: training the ResNet network constructed in the step A15 by a worker through a training set;

Further, in step a17, if the ResNet network reacts correctly, the ResNet network outputs the test result and then ends, and if the ResNet network reacts incorrectly, the ResNet network ends directly.

Further, the step a3 includes the following steps:

a31: after the test application software (namely the ResNet network after the training test in the step A17) is started, a computer displays a login interface, an operator can login the software system through a password, and if the software is used for the first time, after an administrator adds user authority, the operator can login the system; if the operator has the authority, the operator can directly log in, if the password is wrong, the computer can display that the login fails and needs to log in again after being confirmed again by the operator;

Further, if the registration work in the step a31 is successful, the step a31 is performed again, and if the registration work fails, the registration information is popped up for viewing and performing the registration work again.

Compared with the prior art, the tool identification method based on the ResNet network has the following beneficial effects:

(1) the ResNet network-based tool identification method is applied to an intelligent management system of the warehouse tool, facilitates the promotion of automation and informatization construction of the warehouse tool, and improves the tool management efficiency and accuracy.

(2) According to the ResNet network-based tool identification method, all operations in the step A3 can be completed through one-key operation, the computer is provided with a ResNet network model for completing training test work, two-dimensional code identification of all tools can be completed quickly, uniqueness identification can be achieved, and tool management efficiency is improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the method and, together with the description, serve to explain the method and are not intended to limit the method. In the drawings:

FIG. 1 is a flow chart of step A1 according to an embodiment of the present method;

FIG. 2 is a flowchart of step A3 according to an embodiment of the present method;

fig. 3 is a schematic diagram of the connection of a computer and a camera.

Detailed Description

It should be noted that the embodiments and features of the embodiments of the method may be combined with each other without conflict.

In the description of the present method, it is to be understood that the terms "central," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like are used in an orientation or positional relationship indicated in the drawings for convenience in describing the method and for simplicity in description, and are not intended to indicate or imply that the referenced device or element must have a particular orientation, be constructed in a particular orientation, and be operated in a particular manner, and thus should not be considered limiting with respect to the present method. Furthermore, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first," "second," etc. may explicitly or implicitly include one or more of that feature. In the description of the present method, unless otherwise indicated, "a plurality" means two or more.

In the description of the present method, it is to be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the present method can be understood by those of ordinary skill in the art from the specific context.

The noun explains:

ResNet (residual network): the residual network is a convolutional neural network proposed by 4 scholars from Microsoft Research, and wins image classification and object Recognition in the 2015 ImageNet Large Scale Visual Recognition Competition (ILSVRC). The residual network is characterized by easy optimization and can improve accuracy by adding considerable depth. The inner residual block uses jump connection, and the problem of gradient disappearance caused by depth increase in a deep neural network is relieved.

TFRecord: the TFRecord uses a Protocol Buffer binary data coding scheme, only occupies one memory block, only needs a mode of loading a binary file at one time, is simple and quick, is particularly friendly to large training data, and can divide the data into a plurality of TFRecord files to improve the processing efficiency when the training data volume is large.

The method will be described in detail below with reference to the embodiments and with reference to the drawings.

As shown in fig. 1-3, a ResNet network-based tool identification method includes the following steps:

a1: the worker trains and tests the ResNet network;

a2: connecting a plurality of cameras and a computer provided with a ResNet network by a worker through the Internet, wherein each camera is arranged on a tool placing rack;

a3: and D, sequentially photographing and identifying the two-dimensional code by the staff through the plurality of cameras in the step A2 and a computer provided with a ResNet network.

A4: the computer with the ResNet network recognizes the two-dimensional code information and retains records, if the computer with the ResNet network is abnormal, the computer with the ResNet network records the information and gives an alarm, and the system is applied to intelligent management of the warehouse tool, is beneficial to promoting automation and information construction of the warehouse tool and improves tool management efficiency and accuracy.

As shown in fig. 1, the step a1 includes the following steps:

a11: a worker divides a plurality of two-dimensional codes into a training set and a testing set, wherein the training set is used for subsequent training work, and the testing set is used for subsequent testing work;

a12: the staff unifies the sizes of the two-dimensional data set pictures in the training set and the testing set (the ResNet network used by the invention is the ResNet50 network, so the size of the input two-dimensional code is required to be 224 x 3, which is convenient for the division of the training set and the testing set), and is beneficial to the subsequent format conversion;

a13: converting the two-dimensional code data sets in the training set and the testing set with unified sizes in the step A12 into a TFRecord format by workers (because a Tensorflow framework with highest heat and best performance in the market is introduced for network training, the framework provides a proprietary TFrecord data format, the training set and the testing set are respectively packaged, and 1 TFrecord file is generated for every 1000 pictures, so that the network training is facilitated);

a14: after extracting the two-dimensional code data set in the training set and the test set converted into the TFRecord format in the step a13, a worker constructs a queue mode, so that a suitable ResNet network can be conveniently constructed later;

a15: the staff builds a corresponding ResNet network according to the queue mode built in the step A14, the training test process is carried out based on a tensoflow framework, and the programming language is python;

a16: a worker trains the ResNet network constructed in the step A15 through a training set, so that the ResNet network is learned to prepare for subsequent test work, and a random sequence queue mode is constructed to reduce waiting time for reading data in order to improve the machine learning efficiency;

a17: the worker tests the ResNet network trained in the step A16 through the test set, if the test is successful, the ResNet network outputs the test result and then ends, if the test is failed, the ResNet network directly ends, the trained network model predicts the picture of the test set, the accuracy rate of the model is displayed, and the quality of the trained model can be clearly verified;

in the step a17, if the ResNet network makes a correct response, the ResNet network outputs a test result and then ends, and if the ResNet network makes an incorrect response, the ResNet network ends directly, and the training reaches a cutoff condition (that is, an optimal solution is found or a specified training step number is reached), so that the accuracy of model prediction is displayed.

As shown in fig. 2, the step a3 includes the following steps:

a32: after the operator successfully logs in the system, the operator clicks the checking function button displayed on the computer to control the cameras on the tool placement racks to take pictures and capture the two-dimensional codes, so that the informatization and networking degrees of tool management can be improved, and the working efficiency is improved;

a33: uploading screenshots (the screenshots are clear screenshots) to a computer by a plurality of cameras, if the uploading fails, checking whether the network configuration is intact by an operator, and uploading again for calculation after the correction is finished, if the uploading is successful, starting to analyze the screenshots by software;

a34: the test application software starts to recognize uploaded tool pictures in batch based on a trained ResNet model algorithm, if recognition fails, an error prompt pops up, if recognition succeeds, all tool information is displayed in a computer interface list, so that a keyboard point tool is realized, the tool has uniqueness, one-key operation can be performed, the computer is provided with a ResNet network model for finishing training and testing work, two-dimensional code recognition of all tools can be finished quickly, uniqueness recognition can be realized, and tool management efficiency is improved;

if the registration work in the step A31 is successful, the step A31 is carried out again, if the registration work is failed, registration information is popped up for checking and carrying out the registration work again, and the flexibility of logging in the system is improved.

While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not to be limited to the disclosed embodiment, but on the contrary, is intended to cover various modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

完整详细技术资料下载

上一篇：石墨接头机器人自动装卡簧、装栓机

下一篇：一种施工现场安全监管方法及系统

ResNet network-based tool identification method

相关技术

网友询问留言