Fire identification method, device and desktop laser processing equipment

Abstract

The present application provides a fire identification method, device and desktop laser processing equipment, wherein the method includes: triggering the desktop laser processing equipment to perform a panoramic camera image acquisition operation to obtain a panoramic image, wherein the panoramic image describes the status of laser processing performed within the coverage range of the panoramic camera; preprocessing the panoramic image to obtain a number of local images adapted to the segmentation of the panoramic image, wherein the local images are mapped to the local screen of the laser processing performed; performing flame prediction on the local images under a convolutional neural network to obtain the flame probability corresponding to each local image; triggering the alarm signal of the desktop laser processing equipment according to the flame probability distribution of the local image. The method proposed in the present application effectively reduces the complexity of the hardware system, avoids waste of resources, and saves money for users.

Description

Technical Field

[0001] The present application relates to the field of laser processing technology, and in particular to a fire identification method, device and desktop laser processing equipment. Background Art

[0002] Desktop laser processing equipment performs laser processing on processing materials in a cabin in which it is configured. During the laser processing, flames often appear in the cabin, resulting in fire. Therefore, it is necessary to realize fire identification in the cabin of the desktop laser processing equipment.

[0003] Existing desktop laser processing equipment is equipped with AI dedicated chips, which can accurately identify the occurrence of fire in the cabin through the execution of complex algorithms under the control of AI dedicated chips. This ensures accuracy and high reliability for fire identification, for example, its identification accuracy can reach 99.8%, but it also causes the AI ​​dedicated chip to be idle for a long time.

[0004] Therefore, it can be seen that fire identification is a waste of hardware resources for desktop laser processing equipment. Summary of the invention

[0005] One purpose of this application is to solve the technical problem that the AI ​​dedicated chip used for fire identification in desktop laser processing equipment is idle for a long time, wasting hardware resources.

[0006] According to one aspect of an embodiment of the present application, a fire identification method in a desktop laser processing device is disclosed, comprising:

[0007] Triggering the desktop laser processing device to execute a panoramic camera image acquisition operation to obtain a panoramic image, wherein the panoramic image describes the status of the laser processing performed within the coverage range of the panoramic camera;

[0008] Preprocessing the panoramic image to obtain a number of local images adapted to the segmentation of the panoramic image, wherein the local images are mapped to the local screen to be laser processed;

[0009] Perform flame prediction on several local images using convolutional neural networks to obtain the flame probability corresponding to each local image;

[0010] An alarm signal of the desktop laser processing equipment is triggered according to the flame probability distribution of the local image.

[0011] According to one aspect of an embodiment of the present application, a method according to claim 1 is disclosed, comprising:

[0012] The general processing chip triggers the panoramic camera to take pictures of the desktop laser processing equipment that has not generated a fire alarm according to the set detection interval;

[0013] The general processing chip is triggered by the image acquisition operation to obtain a panoramic image from the panoramic camera.

[0014] According to one aspect of an embodiment of the present application, a method according to claim 2 is disclosed, comprising:

[0015] For the laser processing process, the general processing chip initiates the next round of detection of fire identification according to the set detection interval;

[0016] In the initiated round of detection, the general processing chip triggers a picture-taking operation on the panoramic camera.

[0017] According to one aspect of an embodiment of the present application, a method according to claim 1 is disclosed, comprising:

[0018] Reducing the panoramic image to a small size image according to a specified resolution;

[0019] The small-size picture is divided to obtain a plurality of partial pictures whose resolution is reduced twice relative to the panoramic picture.

[0020] According to one aspect of an embodiment of the present application, a method according to claim 1 is disclosed, comprising:

[0021] Inputting the plurality of local images into a predefined convolutional neural network, performing convolution operations and pooling operations on the local images twice, and obtaining feature mapping data corresponding to the local images;

[0022] The linear combination of the feature map data is performed through the linear layer superimposed in the convolutional neural network to output the flame probability corresponding to the local image.

[0023] According to one aspect of an embodiment of the present application, a method according to claim 5 is disclosed, comprising:

[0024] The number of superimposed linear layers is three.

[0025] According to one aspect of an embodiment of the present application, a method according to claim 1 is disclosed, comprising:

[0026] Determining whether the number of local images with high fire probability exceeds a limit value according to the corresponding flame probability, wherein the limit value is adapted to the cabin space in which the desktop laser processing device performs the laser processing process;

[0027] If the number of the partial images exceeds a limit, an alarm signal of the desktop laser processing equipment is triggered.

[0028] According to one aspect of an embodiment of the present application, a method according to claim 7 is disclosed, comprising:

[0029] If the number of the partial images does not exceed the limit, the next round of detection of the fire identification is waited for.

[0030] According to one aspect of an embodiment of the present application, a fire identification device in a desktop laser processing device is disclosed, comprising:

[0031] A picture taking trigger module, used to trigger the desktop laser processing device to perform a picture taking operation of the panoramic camera to obtain a panoramic picture, wherein the panoramic picture describes the status of the laser processing performed within the coverage range of the panoramic camera;

[0032] A preprocessing module, used for preprocessing the panoramic image to obtain a plurality of local images adapted to the segmentation of the panoramic image, wherein the local images are mapped to the local picture to be laser processed;

[0033] The prediction module is used to perform flame prediction on several local images under the convolutional neural network to obtain the flame probability corresponding to each local image;

[0034] A trigger alarm module is used to trigger an alarm signal of the desktop laser processing equipment according to the flame probability distribution of the local image.

[0035] According to one aspect of an embodiment of the present application, a desktop laser processing device is disclosed, comprising:

[0036] A housing, a cover plate, a panoramic camera, a laser emitting device, a track device for driving the laser emitting device to slide, a memory, and a processor. The housing and the cover plate are enclosed together to form a cabin. The laser emitting device, the panoramic camera, and the track device are arranged inside the cabin. The memory is arranged inside the housing and is used to store computer-readable instructions.

[0037] The processor connected to the memory by electrical signals reads the computer readable instructions stored in the memory to execute the method as described above.

[0038] In the embodiment of the present application, under the control of a general-purpose processing chip, the laser processing process of the desktop laser processing equipment will be triggered to execute the image-taking operation of the panoramic camera to obtain a panoramic picture, so as to know the status of laser processing within the coverage range of the panoramic camera through the panoramic picture, and then the panoramic picture is preprocessed to obtain several local pictures adapted to the segmentation of the panoramic picture. Under the action of the local pictures, flame prediction is performed under the convolutional neural network on the local picture where laser processing is performed to obtain the corresponding flame probability, and then an alarm signal is triggered according to the flame probability distribution to achieve fire identification. The whole process is realized under the control of the general-purpose processing chip, and there is no need to add an additional AI dedicated chip, thereby avoiding the long-term idleness of the AI ​​dedicated chip and causing no waste of hardware resources.

[0039] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.

[0040] It should be understood that the foregoing general description and the following detailed description are exemplary only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above and other objects, features and advantages of the present application will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0042] Figure 1 A system architecture diagram of a fire identification method applied in a desktop laser processing device according to an embodiment of the present application is shown.

[0043] Figure 2 A schematic diagram of a desktop laser processing device according to an embodiment of the present application is shown.

[0044] Figure 3 A flow chart of a fire identification method in a desktop laser processing device according to an embodiment of the present application is shown.

[0045] Figure 4 A flow chart of a fire identification method in a desktop laser processing device according to another embodiment of the present application is shown.

[0046] Figure 5 Shown according to Figure 4 A flowchart of the steps of triggering a panoramic camera to take images for a desktop laser processing device that has not generated a fire alarm according to a set detection interval provided by the general processing chip of the corresponding embodiment.

[0047] Figure 6 A flow chart of a fire identification method in a desktop laser processing device according to another embodiment of the present application is shown.

[0048] Figure 7 A general architecture diagram of a convolutional neural network according to an embodiment of the present application is shown.

[0049] Figure 8 A flow chart of a fire identification method in a desktop laser processing device according to another embodiment of the present application is shown.

[0050] Fig. 9 A flowchart of the steps of triggering an alarm signal of a desktop laser processing device according to the flame probability distribution of a local image according to an embodiment of the present application is shown.

[0051] Fig.10 The present invention shows a chipset structure diagram on a main control board of a desktop laser processing device in the prior art.

[0052] Fig.11 A chipset structure diagram in a device main control board according to an embodiment of the present application is shown.

[0053] Fig.12 A flow chart of a fire identification algorithm for desktop laser processing equipment according to another embodiment of the present application is shown.

[0054] Fig.13 A hardware structure diagram of a desktop laser processing device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present application will be more comprehensive and complete and the concepts of the example embodiments will be fully conveyed to those skilled in the art. The accompanying drawings are only schematic illustrations of the present application and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and thus their repeated description will be omitted.

[0056] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present application. However, those skilled in the art will appreciate that the technical solution of the present application may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring the present application and making the various aspects of the present application obscure.

[0057] Some of the blocks shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0058] See also Figure 1 , Figure 1 The system architecture diagram of the fire identification method applied in a desktop laser processing device according to an embodiment of the present application is shown. The system architecture may include: at least one external device 11 and a desktop laser processing device 12 facing each external device 11, and at least one production jig may be used by the desktop laser processing device to form a function that can be customized for fire identification for users.

[0059] It should be understood that Figure 1The number of external devices 11 in the embodiment is only for illustration. Any number of external devices 11 may be provided according to implementation requirements.

[0060] Some technical solutions of the embodiments of the present application can be based on Figure 1 The system architecture shown or its variant architecture is used for specific implementation.

[0061] See also Figure 2 , Figure 2 A schematic diagram of a desktop laser processing device according to an embodiment of the present application is shown. A panoramic camera 210 and a general processing chip 220 are configured in the cabin of the desktop laser processing device. The panoramic camera 210 is integrated directly above the cabin, and the panoramic camera 210 is used to capture the laser processing status performed in the cabin in full-frame within the coverage range of the panoramic camera 210; the general processing chip 220 is not designed only to perform a specific operation, so the general processing chip 220 has the ability to process various control instructions to obtain calculation results and determine whether there is a fire in the cabin.

[0062] In other words, the execution process of determining whether a fire has occurred in the cabin is implemented under the control of the general-purpose processing chip 220, and no longer requires an additional AI-specific chip as set in the existing implementation.

[0063] To this end, in order to enable the general processing chip 220 to perform fire identification, an algorithm suitable for the general processing chip 220 is customized, that is, the fire identification method of the embodiment of the present application, to ensure that the amount of calculation is small and the operating capacity of the general processing chip 220 can support it.

[0064] See also Figure 3 , Figure 3 A flow chart of a method for fire identification in a desktop laser processing device according to an embodiment of the present application is shown, including:

[0065] Step S310, triggering the desktop laser processing device to execute the image acquisition operation of the panoramic camera to obtain a panoramic image, which describes the status of the laser processing performed within the coverage range of the panoramic camera;

[0066] Step S320, pre-processing the panoramic image to obtain a number of local images adapted for the segmentation of the panoramic image, and mapping the local images to the local screen to be laser processed;

[0067] Step S330, performing flame prediction on a number of partial images using a convolutional neural network to obtain a flame probability corresponding to each partial image;

[0068] Step S340 , triggering an alarm signal of the desktop laser processing equipment according to the flame probability distribution of the local image.

[0069] These steps are described in detail below.

[0070] In step S310, the fire identification process under the control of the general processing chip 220 will first trigger the panoramic camera's image acquisition operation, which is used to trigger the panoramic camera to take pictures of its field of view coverage to obtain the current status of the desktop laser processing equipment performing laser processing.

[0071] The panoramic camera is integrated above the cabin and can take pictures of the processing materials placed in the cabin and capture the images of the processing materials being laser processed. It is understandable that during the current laser processing process, if flames are generated, they will inevitably be presented on the panoramic picture taken.

[0072] In other words, the panoramic picture describes through graphic content whether there is a fire in the current laser processing. As the laser processing proceeds, fires may occur from time to time due to materials or various factors, but the conditions of fire are often different. For example, the conditions of fire include the existence of sporadic flames and open flames with larger fires. The two have different effects on the current laser processing and desktop laser processing equipment. Sporadic flames often exist in the laser processing process and generally go out quickly. Therefore, they do not affect the current laser processing, nor do they cause damage to the desktop laser processing equipment. Therefore, there is no need to respond to the occurrence of sporadic flames and identify them as fires, let alone initiate an alarm. For open flames with larger fires, it is necessary to sense and initiate an alarm before they are close to getting out of control.

[0073] Therefore, it should be understood that the fire identification under the control of the general processing chip 220 is aimed at large open flames, and the occurrence of fire is identified before it is close to being out of control, thereby triggering an alarm signal. This kind of fire condition is not normal for the operation of desktop laser processing equipment, nor does it occur frequently. Therefore, for the fire identification performed under the control of the general processing chip 220, there is no need for real-time detection, nor is there a need to use complex algorithms and consume huge amounts of computing power to accurately identify any flames that exist. The low amount of computing power to achieve fire identification makes it possible to deploy it on the general processing chip 220, thereby achieving accurate and reliable fire identification on the general processing chip 220.

[0074] Under the control of the general processing chip 220, a round of detection is performed to complete the current round of fire identification, and the panoramic camera is first triggered to take pictures. The general processing chip 220 will trigger a new round of detection according to the set detection interval, ensuring that a large fire can be accurately detected while reducing the occupied computing resources, thereby ensuring that the fire identification does not require the support of an AI dedicated chip.

[0075] Exemplarily, the desktop laser processing equipment integrates a panoramic camera and a close-up camera. The close-up camera is arranged on the laser emitting device, and can take local high-definition images / pictures to update the panoramic images / pictures taken by the panoramic camera for precise processing, positioning, etc. The shell of the desktop laser processing equipment and the cover plate connected to the shell together enclose an internal space, that is, the cabin of the desktop laser processing equipment. The cover plate is used to open and close the cabin. The cabin is provided with a laser emitting device, a metal tool, and a track device for driving the laser emitting device and / or the metal tool to slide, wherein the track device can be an X-axis guide rail, and the X-axis guide rail can be a linear guide rail, or a guide rail in which the optical axis and the roller slide together, etc. It only needs to be able to drive the laser emitting device to move and process on the X-axis and the Y-axis. A Z-axis moving track can also be arranged in the laser emitting device for moving and focusing in the Z-axis direction before and / or during processing. The bottom of the cabin is used to place the workpiece to be processed. The user can open the cover plate to put the workpiece in and then close it, or open the cover plate to take out the workpiece.

[0076] In one embodiment, the laser light source of the desktop laser processing equipment can be generated by a laser emitting device. In another embodiment, the laser light source can be generated by other components such as a carbon dioxide laser tube, etc., and enter the laser emitting device through a reflector, etc., and finally emit after passing through the laser emitting device to process the workpiece.

[0077] During the operation of the desktop laser processing equipment, a panoramic camera is used to obtain a panoramic picture, so that the laser processing status in the cabin can be viewed in its entirety, and it can be known whether there is a flame in any area, and even the distribution of flames in the entire area.

[0078] See also Figure 4 , Figure 4 A flow chart of a method for identifying fire in a desktop laser processing device according to another embodiment of the present application is shown. The present application embodiment provides a step S310 of triggering the desktop laser processing device to perform a panoramic camera image acquisition operation to obtain a panoramic image, including:

[0079] Step S311, the general processing chip triggers the panoramic camera to take pictures for the desktop laser processing equipment that has not generated a fire alarm according to the set detection interval;

[0080] Step S312, triggering the general processing chip to obtain a panoramic picture from the panoramic camera through the image acquisition operation.

[0081] The following describes these two steps in detail.

[0082] In step S311, according to the set detection interval, the general processing chip 220 triggers a new round of detection to identify the fire. With the cooperation of the panoramic camera 210, the general processing chip 220 controls and executes a round of detection, which is the whole process of fire identification. The acquisition of images from the panoramic camera, the preprocessing of the panoramic images, and the flame prediction for the local images are all performed in one round of detection.

[0083] Exemplarily, the set detection interval can be dynamically configured according to the hardware status of the desktop laser processing equipment and the status of the laser processing performed. For example, the detection interval can be 30 seconds.

[0084] The normal operation of the desktop laser processing device means that the desktop laser processing device has no fire alarm and is continuously performing laser processing. In the normal operation of the desktop laser processing device, after a round of detection is completed, the general processing chip 220 waits for the set detection interval and then triggers the next round of new detection.

[0085] As a result, the fire identification implemented by the general processing chip 220 of the desktop laser processing equipment will be performed at intervals rather than in real time, thereby ensuring that the general processing chip 220 can support and adapt to the fire conditions of the laser processing equipment.

[0086] See also Figure 5 , Figure 5 Shown according to Figure 4 A flowchart of the steps of triggering a panoramic camera to take images for a desktop laser processing device that has not generated a fire alarm according to a set detection interval provided by the general processing chip of the corresponding embodiment.

[0087] The embodiment of the present application provides a step S311 in which a general processing chip triggers a panoramic camera to take an image for a desktop laser processing device that has not generated a fire alarm according to a set detection interval, including:

[0088] Step S3111, for the processing process of the desktop laser processing equipment, the general processing chip initiates the next round of detection of fire identification according to the set detection interval;

[0089] Step S3112: In the initiated round of detection, the general processing chip triggers a picture-taking operation on the panoramic camera.

[0090] Through the execution of step S3111 and step S3112, the fire identification of the desktop laser processing equipment can perform the next round of detection according to the set detection interval, and when one round of detection is completed, the general processing chip 220 no longer performs fire identification, but processes other control instructions, thereby being able to take into account both fire identification and the normal operation of the desktop laser processing equipment.

[0091] In step S312, after the panoramic camera 210 is triggered by the image acquisition operation initiated by the general processing chip 220, it performs a photo-taking action to generate a panoramic image. At this time, the general processing chip 220 obtains the panoramic image from the panoramic camera.

[0092] In step S320, after obtaining the panoramic image, the general processing chip 220 pre-processes the panoramic image and provides input data for the convolutional neural network.

[0093] The preprocessing process performed includes, but is not limited to, segmenting the panoramic image to obtain a number of local images, so as to obtain the local picture of the laser processing through the local images.

[0094] As mentioned above, the occurrence of fire is often a sporadic flame in the desktop laser processing equipment cabin, which is only distributed in one or a few local pictures; on the other hand, it is a larger open flame, which is distributed in most local pictures. Therefore, the local pictures obtained can be used to know whether a fire occurs during the laser processing process.

[0095] Therefore, through the preprocessing, a visual description of the fire will be provided for the fire identification. On the other hand, it will also provide data with reduced data volume and can accurately characterize the fire for the real realization of fire identification, that is, the prediction of the convolutional neural network, that is, the local picture. In addition, under the action of the local picture obtained by preprocessing, the deployment and prediction of the convolutional neural network under the general processing chip 220 will reduce the amount of computation, so that the general processing chip 220 can support the fire identification function of the desktop laser processing equipment.

[0096] Exemplarily, the preprocessing of the panoramic image may include reducing the panoramic image and performing a segmentation process thereafter. This not only ensures that the tiny flames occurring inside the desktop laser processing equipment are not missed through the description of the panoramic image, but also reduces the amount of data and calculation while ensuring the comprehensiveness of the described image, thereby enabling the normal use of the general-purpose processing chip 220.

[0097] Specifically, the preprocessing performed obtains several local images through image reduction and segmentation operations. The obtained local images are reduced twice relative to the panoramic image, which greatly enhances the usability of the general processing chip 220 with the newly added fire recognition function in the desktop laser processing equipment. The several local images mapped to the panoramic image will also make up for the accuracy that may be reduced as a result, and the reliability of the fire recognition function in the desktop laser processing equipment is improved.

[0098] See also Figure 6 , Figure 6A flow chart of a method for fire identification in a desktop laser processing device according to another embodiment of the present application is shown. The present application embodiment provides a step S320 of preprocessing a panoramic image to obtain a plurality of local images adapted for panoramic image segmentation, including:

[0099] Step S321, reducing the panoramic image according to the specified resolution to obtain a small-size image;

[0100] Step S322 , dividing the small-size picture into a plurality of partial pictures whose division ratio is reduced twice relative to the panoramic picture.

[0101] The following describes these two steps in detail.

[0102] For the convolutional neural network, the execution of step S321 and step S322 both reduce the data dimension, and initiate parallel execution through several local images obtained by segmentation, thereby reducing the amount of calculation while improving the calculation efficiency, further reducing the calculation burden of the general processing chip 220.

[0103] For example, a 4656*3496 (16 million pixels) panoramic image is reduced to a small-size image of 400*300, and then the small-size image is segmented to obtain 300 local images with a resolution of 20*20.

[0104] In step S330, all the obtained partial images are used as inputs of the convolutional neural network, and flame prediction is performed on each partial image to obtain the flame probability corresponding to each partial image.

[0105] See also Figure 7 , Figure 7 A general architecture diagram of a convolutional neural network according to an embodiment of the present application is shown. The architecture may include: an Input Layer, two Convolution Layers, two Pool Layers, a Dropout Layer, three Linear Layers, and an OutLayer.

[0106] Exemplarily, as mentioned above, the obtained several local images are used as input, and convolution operations and pooling operations are performed through a convolutional neural network to obtain feature extraction of the local images and reduce the amount of data for the extracted features.

[0107] On this basis, other network layers provided by the convolutional neural network, such as the Dropout layer, act on the output of the pooling operation to ensure the reliability of the convolutional neural network, and then the flame probability can be output for each local image under the action of the linear layer.

[0108] Specifically, the convolutional layer and the pooling layer are combined into a main structure of the convolutional neural network, which performs convolution and pooling operations on the input local image twice. In other words, the convolutional neural network contains two convolutional layers and two pooling layers, and the convolutional layer is used as the starting network layer, and the convolutional layers and pooling layers are set alternately.

[0109] For a convolutional neural network with convolutional layers and pooling layers, a Dropout layer is connected to the output of the second pooling layer, and three linear layers are stacked on the Dropout layer to build a complete convolutional neural network.

[0110] At this time, several segmented local images are input into the first pooling layer. After passing through several network layers, the corresponding flame probability can be output by the linear layer. And so on, the flame prediction of all local images is completed.

[0111] See also Figure 8 , Figure 8 A flow chart of a method for fire identification in a desktop laser processing device according to another embodiment of the present application is shown. The present application embodiment provides a step S330 of performing flame prediction on a plurality of local images under a convolutional neural network, thereby obtaining a flame probability corresponding to each local image, including:

[0112] Step S331, inputting a number of local images into a predefined convolutional neural network, performing convolution operations and pooling operations on the local images twice, and obtaining feature mapping data corresponding to the local images;

[0113] Step S332, performing linear combination of feature mapping data through the superimposed linear layer in the convolutional neural network to output the flame probability corresponding to the local image.

[0114] The following describes these two steps in detail.

[0115] In step S331, several local images will be input into a predefined convolutional neural network and undergo convolution and pooling operations.

[0116] Exemplarily, as mentioned above, the obtained several local images will enter the predefined convolutional neural network. The alternating distribution of convolutional layers and pooling layers in the convolutional neural network makes it possible to perform convolution and pooling operations twice on the local images, and finally output feature mapping data corresponding to the several local images from the pooling layer.

[0117] As mentioned above, the predefined convolutional neural network is used to extract features from several local images, reduce the pixels or values ​​mapped by the features, and then use the feature mapping data obtained in this way to predict the flame probability of each local image. The predefined convolutional neural network includes two convolutional layers and two pooling layers that are alternately distributed to achieve feature extraction and mapping for secondary execution.

[0118] Specifically, for the convolution operation and pooling operation performed twice, it can be understood that a convolution operation is performed by a convolution layer, and correspondingly, a pooling operation can also be performed by a pooling layer. Based on this, the convolution operation and pooling operation performed twice refer to the data corresponding to the local image entering the convolution layer twice and entering the pooling layer twice. The convolution layer and the pooling layer overlap each other, and the convolution layer is used as the starting point to receive the input local image. The convolution layer is used to extract the features of the local image, and the pooling layer is used to reduce the dimension and retain the features of the local image at the same time to reduce the operation of the general processing chip. Feature mapping data is extracted from several local images by performing convolution operations and pooling operations. The extracted features are data mapped to specific feature dimensions. Feature mapping data is data obtained after reducing the mapped pixels or values ​​of the features. In other words, feature mapping data is a mapping of the features extracted from the local image, which will be used to characterize the picture content of the corresponding local image. For example, the feature mapping data can indicate whether there is a flame in the local image in terms of content.

[0119] For each local image, feature extraction and feature mapping are implemented under the action of the predefined convolutional neural network, and the final data is reduced through feature mapping. Under the control of the secondary execution pooling operation, the availability of the general processing chip 220 for many control processes and fire identification processes is guaranteed.

[0120] In other words, for the execution of fire prediction on the general-purpose processing chip 220, feature extraction and mapping of the extracted features are repeatedly performed, which reduces the amount of computation for the calculation of the flame probability corresponding to each local image while providing accurate features. This enables the general-purpose processing chip 220 that is the only one configured for the desktop laser processing equipment to support the operation of the desktop laser processing equipment, and there is no need to configure other dedicated chips, thereby making full use of the hardware resources of the desktop laser processing equipment and reducing the cost of the desktop laser processing equipment.

[0121] In step S332, the feature map data obtained after the convolution operation and the pooling operation is used as input, and then the flame probability corresponding to each local image is predicted through the superimposed linear layer in the convolutional neural network.

[0122] Exemplarily, as mentioned above, the obtained local images will enter the predefined convolutional neural network. After the convolution operation and pooling operation, the obtained feature mapping data will enter the three linear layers in sequence.

[0123] The flame probability is the probability of flame occurrence obtained by predicting the feature mapping data of each local image through the linear layer. The flame probability can be used to clarify the possibility of the existence of flames in the corresponding local image. For example, the larger the flame displayed in the local image, the greater the flame probability corresponding to the local image. The flame probability can then be used to identify local images containing flames.

[0124] On the other hand, the predefined convolutional neural network is trained by the local images corresponding to a large number of panoramic images. For the large number of panoramic images taken during the laser processing process, there are panoramic images with flames corresponding to large fires, panoramic images without flames, and panoramic images with a small amount of flames that quickly go out by themselves. The local images are obtained from these panoramic images and annotated to obtain the training data of the convolutional neural network.

[0125] Furthermore, the labeling process performed on the partial images refers to marking the flame probability corresponding to the partial images of the burning part as 1, and marking the flame probability corresponding to the partial images of the non-burning part as 0.

[0126] In step S340, the flame probability distribution will characterize the possibility of flames existing in various places in the panoramic image, and indicate whether there is flame in the cabin and the distribution of the existing flames through the flame probability distribution, thereby determining whether an alarm signal is triggered for the desktop laser processing equipment.

[0127] It should be understood that as flame prediction is performed on all partial images, a flame probability distribution is formed from the flame probabilities of the obtained partial images, and then the flame probability distribution is used as a basis to determine whether an alarm signal is triggered in the current round of detection.

[0128] Specifically, the number of local images corresponding to the high flame probability is obtained from the flame distribution formed by each local image, and then whether a fire occurs in the desktop laser processing equipment is determined based on the numerical value of the number of local images with high flame probability. It should be understood that the high flame probability is relative, and a threshold value can be configured. The flame probability that is numerically higher than the probability threshold can be determined as a high flame probability, and vice versa. It should also be understood that the limit value of the number of local images with high flame probability is also relative and can also be freely configured. If the number of local images with high flame probability exceeds the limit value, it can be determined that a fire occurs, and vice versa, no fire occurs.

[0129] It should be understood that the limit value is pre-configured by the desktop laser processing equipment. Different machines must be configured with different limit values ​​due to their different machine hardware environments. Exemplarily, the limit value is adapted to the cabin space of the desktop laser processing equipment. Specifically, the adaptation of the limit value to the cabin space means that the limit value is positively correlated with the size of the cabin space. For example, if the cabin space is larger, the corresponding limit value is configured as a larger value. Correspondingly, if the cabin space is smaller, the corresponding limit value is configured as a smaller value. The numerical value of the configured limit value will control the response accuracy of the desktop laser processing equipment to the existing flame.

[0130] For a specific cabin space, relative to the adapted value, if the configured limit is a larger value, the alarm signal will be triggered when the fire is out of control, causing the desktop laser processing equipment to be burned; if the configured limit is a smaller value, the desktop laser processing equipment will issue an alarm for the tiny flames that exist. The existence of tiny flames in the laser processing process is extremely normal and does not require intervention. The configuration of a smaller limit will affect the normal operation of the desktop laser processing equipment and may even prevent it from working properly.

[0131] In other words, when the cabin space is large, if the limit value is too small, it is likely that the alarm signal will be triggered by a spark generated by the laser processing equipment during processing, and this spark will automatically extinguish, which is likely to cause a false alarm; similarly, when the cabin space is small, if the limit value is too large, it is likely that the alarm signal will be triggered only when the flame is too large, causing damage to the laser processing equipment. Therefore, this limit value must be adapted to the cabin space where the desktop laser processing equipment performs the laser processing process.

[0132] For example, if the configured probability threshold is 50%, then the flame probability corresponding to a number of local images is greater than 50%, which is a high flame probability, otherwise it is a low flame probability. If the configured high flame probability local image number limit is 6, then among the local images, if the number of high flame probability local images is greater than 6, it can be determined that a fire has occurred, and the alarm signal of the desktop laser processing equipment can be triggered. Otherwise, no fire has occurred, and the alarm signal of the desktop laser processing equipment cannot be triggered, and it will continue to wait for the next round of fire identification.

[0133] See also Fig. 9 , Fig. 9 A flowchart of the steps of triggering an alarm signal of a desktop laser processing device according to the flame probability distribution of a local image according to an embodiment of the present application is shown.

[0134] The step S340 of triggering an alarm signal of a desktop laser processing device according to the flame probability distribution of a local image in the embodiment of the present application includes:

[0135] Step S341, judging whether the number of local images with high fire probability exceeds a limit value according to the corresponding flame probability, and the limit value is adapted to the cabin space in which the desktop laser processing equipment performs the laser processing process. If the judgment is yes, step S342 is executed, and if the judgment is no, step S343 is executed;

[0136] Step S342, triggering an alarm signal of the desktop laser processing equipment;

[0137] Step S343, waiting for the next round of fire identification detection.

[0138] These steps are described in detail below.

[0139] In step S341, as mentioned above, the local image with high fire probability refers to the local image corresponding to the high flame probability. The high fire probability is relative and is adapted to the configured probability threshold.

[0140] The flame probabilities calculated for each of the local images are used to determine which local images have flame probabilities corresponding to high fire probabilities according to the configured probability threshold, and then the number of local images with high fire probabilities is obtained by counting.

[0141] The number of local images with high fire probability indicates the size of the area with high fire probability in the flame probability distribution of the local image. It should be understood that the larger the area covered by the area with high fire probability, the larger the fire in the desktop laser processing equipment is and the higher the possibility of being close to being out of control; correspondingly, the smaller the area covered by the area with high fire probability, the laser processing process currently performed by the desktop laser processing equipment only has flames that will soon go out.

[0142] Therefore, the number of local images with high fire probability will indicate whether the fire occurring during the laser processing currently performed by the desktop laser processing device requires intervention. Specifically, whether to trigger the alarm signal of the desktop laser processing device is determined based on whether the number of local images with high fire probability obtained by statistics exceeds the limit.

[0143] The alarm signal of the desktop laser processing equipment is used to initiate a fire alarm of the desktop laser processing equipment so as to intervene in the current fire.

[0144] It should be understood that in the implementation of existing desktop laser processing equipment, the equipment main control board is the main part of its complex electronic system, and the chipset is an important component of the equipment main control board. In the prior art, the main chips used in the equipment main control board of desktop laser processing equipment include general processing chips 220 and various special chips. Among them, the general processing chip 220 is used to execute ordinary control instructions or calculation operations of the desktop laser processing equipment to obtain control instructions and calculation results. Various special chips, such as AI special chips, can be used specifically to cooperate with the camera to determine whether there is a fire in the desktop laser processing equipment.

[0145] The price of AI dedicated chip is high, because there is a gap in the AI ​​dedicated chip to monitor whether there is a fire in the desktop laser processing equipment, and it is not detected all the time, which causes it to be idle for a long time. Therefore, this application decides to use the general processing chip 220 to replace the AI ​​dedicated chip to perform fire identification operations on the desktop laser processing equipment. The structure of the chipset in the main control board of the equipment is as follows Fig.10 shown.

[0146] See also Fig.10 , Fig.10 The present invention shows a chipset structure diagram on a main control board of a desktop laser processing device in the prior art.

[0147] In order to complete the function of fire identification in the cabin, the current laser processing equipment will use a special chip to achieve this purpose, such as the AI ​​dedicated chip commonly used in the prior art. In this way, when the laser processing equipment needs to achieve purpose A, it may need to add a chip A, and when it needs to achieve purpose B, it may need to add a chip B. Doing so greatly increases the complexity of the hardware system of the laser processing equipment. Therefore, this application chooses to use a general-purpose processing chip for general control / calculation, and also for completing the fire identification task in the laser processing equipment cabin, such as Fig.11 shown.

[0148] See also Fig.11 , Fig.11 The chipset structure diagram in the device main control board according to an embodiment of the present application is shown. The device main control board of the present application embodiment is only equipped with a general processing chip 220, which is used to complete the general control instructions or calculation operations of the desktop laser processing equipment.

[0149] It should be understood that, in addition to achieving the above functions, the general processing chip in the main control board of the device in the embodiment of the present application can also cooperate with the camera to complete the entire process of the method described above.

[0150] Specifically, the general processing chip 220 can control the panoramic camera 210 to perform shooting operations, analyze and process the image information captured by the panoramic camera 210, and make a judgment on whether a fire occurs in the cabin of the desktop processing equipment.

[0151] See also Fig.12 , Fig.12 The flowchart of the desktop laser processing equipment fire identification algorithm according to another embodiment of the present application is shown. Before completing the laser processing equipment fire identification, it is necessary to perform offline training on several local images obtained to prove that there is no flame in the desktop laser processing equipment. At this time, the training data obtained through offline training can be used as a judgment standard. The training data is obtained based on the actual situation when offline. It should be understood that the offline training can be performed only once, and training can be performed again when the subsequent data is updated. Including:

[0152] Step S410, scaling down the panoramic image in equal proportion, dividing the scaled-down image into a number of local images with a resolution of 20×20, and mapping the local images to the local screen to be laser processed;

[0153] Step S420, judging whether flames are generated on a plurality of local images obtained by segmentation according to the training data obtained in advance;

[0154] Step S430, performing flame prediction on a number of partial images using a convolutional neural network to obtain a flame probability corresponding to each partial image;

[0155] Step S440 , determining whether there are 4 or more partial images corresponding to flame probabilities greater than 0.5, thereby determining whether to trigger an alarm signal of the desktop laser processing equipment.

[0156] These steps are described in detail below.

[0157] In step S410 , after obtaining the panoramic picture, the general processing chip 220 scales down the panoramic picture in equal proportion, wherein the scaling down means that the aspect ratio of the panoramic picture remains unchanged.

[0158] For example, if the aspect ratio of the panoramic image is 3:2, then the aspect ratio of the reduced image after reduction is also 3:2; if the aspect ratio of the panoramic image is 1:1, then the aspect ratio of the reduced image after reduction is also 1:1.

[0159] After obtaining the reduced image, the reduced image is segmented into several local images with a resolution of 20X20.

[0160] It should be understood that the several local pictures obtained after segmentation are of equal size.

[0161] For example, if the resolution of one of the segmented partial images is 20X20, then the resolution of all other partial images is also 20X20.

[0162] In step S420, the partial images obtained after segmentation are judged whether flames are generated according to the training data obtained in advance. The data generated by each partial image is compared with the training data obtained by offline training. If the data of the two are the same, the partial image does not generate flames. If the data of the two are different, the partial image may generate flames.

[0163] In step S430, flame prediction is performed on several local images using a convolutional neural network to obtain the flame probability corresponding to each local image.

[0164] The flame probability is between 0 and 1, and the specific value is determined by the size of the small flame in the local image.

[0165] In step S440, the limit value of the number of local images can be obtained based on empirical data. The original image is reduced and segmented to obtain local images with a resolution of 20X20. At this time, the limit value of the number of local images is 4.

[0166] If the flame probability corresponding to 4 or more local images is greater than 0.5, the alarm signal of the desktop laser processing equipment can be triggered; if the number of local images with the corresponding flame probability greater than 0.5 is less than 4, the alarm signal of the desktop laser processing equipment cannot be triggered, and the next round of detection will be carried out after waiting for a few seconds, and the desktop laser processing equipment will be triggered again through the general processing chip to execute the panoramic camera's image acquisition operation.

[0167] The desktop laser processing equipment fire identification method according to the embodiment of the present application can be Figure 1 The desktop laser processing equipment 12 is used to achieve this. Fig.13 The desktop laser processing device 12 according to the embodiment of the present application is described. Fig.13 The desktop laser processing device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0168] like Fig.13 As shown, the desktop laser processing device 12 is in the form of a general-purpose computing device. The components of the desktop laser processing device 12 may include, but are not limited to: at least one processing unit 810, at least one storage unit 820, and a bus 830 connecting different system components (including the storage unit 820 and the processing unit 810).

[0169] The storage unit stores program codes, which can be executed by the processing unit 810, so that the processing unit 810 performs the steps according to various exemplary embodiments of the present invention described in the description of the exemplary method described above in this specification. For example, the processing unit 810 can perform the following steps: Figure 2 The steps shown in .

[0170] The storage unit 820 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 8201 and / or a cache storage unit 8202 , and may further include a read-only storage unit (ROM) 8203 .

[0171] The storage unit 820 may also include a program / utility 8204 having a set (at least one) of program modules 8205, such program modules 8205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0172] Bus 830 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0173] The desktop laser processing device 12 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the desktop laser processing device 12, and / or any device that enables the desktop laser processing device 12 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. In addition, the desktop laser processing device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 860. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the desktop laser processing device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0174] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0175] In an exemplary embodiment of the present application, a computer program medium is further provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor of a computer, the computer is enabled to execute the method described in the above method embodiment.

[0176] According to one embodiment of the present application, a program product for implementing the method in the above method embodiment is also provided, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited to this. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0177] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0178] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0179] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0180] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0181] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0182] In addition, although the steps of the method in the present application are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0183] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0184] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the appended claims.

Claims

1. A method for identifying fire in a desktop laser processing device, characterized in that: The desktop laser processing equipment is equipped with a panoramic camera and a general processing chip in the cabin, and the panoramic camera is integrated above the cabin. Under the control of the general processing chip, the method includes: Triggering the desktop laser processing device to execute a panoramic camera image acquisition operation to obtain a panoramic image, wherein the panoramic image describes the status of the laser processing performed within the coverage range of the panoramic camera; Preprocessing the panoramic image to obtain a number of local images adapted to the segmentation of the panoramic image, wherein the local images are mapped to the local screen to be laser processed; Perform flame prediction on several local images using convolutional neural networks to obtain the flame probability corresponding to each local image; Determining whether the number of local images with high fire probability exceeds a limit value according to the corresponding flame probability, wherein the high fire probability is adapted to a configured probability threshold value, and the limit value is positively correlated with the size of the cabin space in which the desktop laser processing device performs the laser processing process; If the number of the partial images exceeds a limit, an alarm signal of the desktop laser processing equipment is triggered.

2. The method according to claim 1, characterized in that The triggering of the desktop laser processing device to execute the image acquisition operation of the panoramic camera to obtain the panoramic image includes: The general processing chip triggers the panoramic camera to take pictures of the desktop laser processing equipment that has not generated a fire alarm according to the set detection interval; The general processing chip is triggered by the image acquisition operation to obtain a panoramic image from the panoramic camera.

3. The method according to claim 2, characterized in that The general processing chip triggers the panoramic camera to take pictures of the desktop laser processing equipment that has not caused a fire alarm according to the set detection interval, including: For the laser processing process, the general processing chip initiates the next round of detection of fire identification according to the set detection interval; In the initiated round of detection, the general processing chip triggers a picture-taking operation on the panoramic camera.

4. The method according to claim 1, characterized in that: The preprocessing of the panoramic picture to obtain a plurality of partial pictures adapted to the segmentation of the panoramic picture includes: Reducing the panoramic image to a small size image according to a specified resolution; The small-size picture is divided to obtain a plurality of partial pictures whose resolution is reduced twice relative to the panoramic picture.

5. The method according to claim 1, characterized in that The flame prediction is performed on a plurality of local images using a convolutional neural network to obtain the flame probability corresponding to each local image, including: Inputting the plurality of local images into a predefined convolutional neural network, performing convolution operations and pooling operations on the local images twice, and obtaining feature mapping data corresponding to the local images; The linear combination of the feature map data is performed through the linear layer superimposed in the convolutional neural network to output the flame probability corresponding to the local image.

6. The method according to claim 5, characterized in that The stacked linear layers are three layers.

7. The method according to claim 1, characterized in that The method further comprises: If the number of the partial images does not exceed the limit, the next round of detection of the fire identification is waited for.

8. A fire identification device in a desktop laser processing equipment, characterized in that: The desktop laser processing equipment is equipped with a panoramic camera and a general processing chip in the cabin, the panoramic camera is integrated above the cabin, and the device facing the general processing chip includes: A picture taking trigger module, used to trigger the desktop laser processing device to perform a picture taking operation of the panoramic camera to obtain a panoramic picture, wherein the panoramic picture describes the status of the laser processing performed within the coverage range of the panoramic camera; A preprocessing module, used for preprocessing the panoramic image to obtain a plurality of local images adapted to the segmentation of the panoramic image, wherein the local images are mapped to the local picture to be laser processed; A prediction module is used to perform flame prediction on several local images using a convolutional neural network to obtain the flame probability corresponding to each local image; An alarm module is triggered to determine whether the number of local images with high fire probability exceeds a limit value according to the corresponding flame probability, wherein the high fire probability is adapted to a configured probability threshold value, and the limit value is positively correlated with the size of the cabin space in which the desktop laser processing equipment performs the laser processing process; if the number of local images exceeds the limit value, an alarm signal of the desktop laser processing equipment is triggered.

9. A desktop laser processing device, characterized in that: It includes a shell, a cover plate, a panoramic camera, a laser emitting device, a track device for driving the laser emitting device to slide, a memory, and a processor. The shell and the cover plate are enclosed together to form a cabin. The laser emitting device, the panoramic camera, and the track device are arranged inside the cabin. The memory is arranged inside the shell and is used to store computer-readable instructions. The processor connected to the memory by electrical signals reads the computer readable instructions stored in the memory to execute the method according to any one of claims 1 to 7.

Images