Production scene personnel supervision and protection system based on AI

Abstract

The invention relates to the technical field of intelligent industrial management, in particular to an AI-based production scene personnel supervision and protection system, which comprises a data acquisition module, a communication module, a data processing module, a positioning module, an AI analysis module and a control module, compared with the prior art that a single signal source or a simple geometric algorithm is adopted for positioning, environmental interference is easily caused, and precision is insufficient, the scheme adopts a UWB module and multi-base-station layout, UWB pulse signals are transmitted and received, a complex positioning algorithm principle equation set is used for accurately converting the UWB pulse signals into actual coordinate values, and positioning accuracy is improved. According to the invention, high-precision plane positioning of a workshop crane hoisting article is realized, and the position of the hoisting article in an image is intelligently identified by combining feature information provided by the big data processing module on the basis of obtaining plane coordinates, so that the height coordinates of the hoisting article are accurately calculated, and the three-dimensional space precision and the intelligent level of the positioning system are comprehensively improved; and the accurate positioning requirement in a complex workshop environment is effectively met.

Description

Technical Field

[0001] The present invention relates to the field of intelligent industrial management technology, and in particular to an AI-based production scene personnel supervision and protection system. Background Art

[0002] In modern industrial production, workshop crane hoisting operations are an indispensable part, which undertakes key tasks such as material handling and assembly. However, with the improvement of production efficiency and the complexity of the working environment, ensuring the safety and accuracy of crane hoisting operations has become an urgent problem to be solved. Especially when there are people under the hoisted objects, if they cannot be detected and warned in a timely and accurate manner, serious safety accidents may occur.

[0003] Traditional personnel detection systems mostly use a single signal source or a simple geometric algorithm for positioning. Although these methods can achieve a preliminary judgment of the personnel's location to a certain extent, they have many limitations. For example, a single signal source is susceptible to environmental interference, such as metal structures in the workshop, electromagnetic wave interference, etc., which may cause signal distortion and thus affect the accuracy of positioning. Simple geometric algorithms are often based on idealized physical models and ignore complex factors in the actual environment, such as obstacles and multipath effects. Therefore, it is difficult to achieve ideal positioning results in practical applications.

[0004] In order to overcome these limitations, in recent years, with the rapid development of wireless positioning technology, especially the widespread application of ultra-wideband (UWB) technology, a new solution has been provided for personnel detection under crane hoisting objects in workshops. UWB technology has shown great potential in the field of positioning due to its high bandwidth, low power consumption, and strong anti-interference ability. By transmitting and receiving UWB pulse signals and extracting signal arrival time and strength data, accurate measurement of the target position can be achieved.

[0005] However, relying solely on UWB technology for plane positioning is still not enough to meet the precise positioning needs in complex workshop environments. Because in actual operations, the height information of hoisted objects is also crucial. In order to obtain this information, it is necessary to combine image processing technology and big data analysis methods. Through image processing technology, the position of hoisted objects in the image can be intelligently identified; and the big data analysis module can provide characteristic information of hoisted objects, such as size, shape, etc., based on historical data and real-time data, so as to further calculate its height coordinates. Summary of the invention

[0006] In order to overcome the problems raised in the above-mentioned background technology, the present invention proposes an AI-based production scene personnel supervision and protection system.

[0007] The technical solution of the present invention is: an AI-based production scene personnel supervision and protection system, including: The data acquisition module is used to conduct real-time and uninterrupted monitoring of the objects hoisted by the crane in the workshop and collect high-definition images of the objects hoisted by the crane in the workshop; Communication module, used to provide a data transmission network to realize data transmission between the data acquisition module, positioning module, data processing module, AI analysis module and control module; A data processing module, used for processing the high-definition images collected by the data collection module; The positioning module is used to determine the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module; The AI ​​analysis module is used to analyze whether there are people walking under the objects hoisted by the crane in the workshop based on the data results of the data processing module and the positioning module; The control module is used to control the workshop crane according to the data results of the AI ​​analysis module, and issue an early warning to notify the personnel under the objects hoisted by the workshop crane.

[0008] Preferably, the data acquisition module includes two groups of high-definition cameras, which are respectively arranged at both ends of the workshop crane. The communication module includes four groups of UWB base stations and one group of UWB modules. The four groups of UWB base stations are respectively arranged at the four corners of the workshop. The UWB module is arranged above the workshop crane hook.

[0009] Preferably, when the data processing module processes the high-definition image collected by the data collection module, the data processing module includes the following steps: S11: image preprocessing, preprocessing the collected high-definition image, wherein the preprocessing methods used include filtering, enhancement, smoothing, contrast adjustment and brightness adjustment; S12: Image segmentation: using image segmentation algorithm to segment the image into multiple regions with similar attributes; S13: Feature extraction, extracting required feature information from each segmented region image, wherein the required feature information includes color features, texture features and shape features.

[0010] Preferably, when filtering an image, the principle formula is: ; in, is the pixel value of the filtered image at coordinate (x, y), is the standard deviation of the Gaussian function, k is the size of the filter, i and j are loop variables used to traverse all elements of the filter, The input image is at coordinates The pixel value at is an element in the filter.

[0011] Preferably, when using an image segmentation algorithm to segment an image into a plurality of regions with similar attributes, the following steps are included: S21: Initialize the cluster center and select the initial cluster center evenly on the image; S22: Search field, defining a search field for each cluster center, wherein the size of the search field is related to the scaling ratio of the high-definition camera; S23: Clustering, in the search domain, the color distance and spatial distance between each pixel and the cluster center are calculated, and the color distance and spatial distance are integrated into a comprehensive distance metric by a weighted sum method; S24: updating the cluster center, and updating the cluster center of each search area according to the clustering result, wherein the updated cluster center selects the average color value of all pixels in the search area as the new cluster center; S25: Iterate and repeat steps S22-S24 until the cluster center no longer changes significantly.

[0012] Preferably, when the positioning module determines the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module, the principle steps are as follows: S31: signal transmission and reception, the UWB module transmits a UWB pulse signal, and after the UWB base station receives the pulse signal sent by the UWB module, it extracts the required data from the received pulse signal, wherein the required data includes the signal arrival time and signal strength; S32: Data processing and analysis: converting the extracted data into actual coordinate values ​​through a positioning algorithm, wherein the principle equation group of the positioning algorithm used is: ; ; ; in, is the coordinate of the uwb module, that is, the plane coordinate of the crane hoisting items in the workshop. , , and are the coordinates of four groups of UWB base stations, c is the speed of light, , , and They are the time when the pulse signals sent by the uwb module arrive at the four groups of uwb base stations; S33: Height coordinate analysis. After calculating the plane coordinates of the objects hoisted by the crane in the workshop, the position of the objects hoisted by the crane in the workshop is identified in the image according to the feature information obtained by the big data processing module, and then the height coordinates of the objects hoisted by the crane are identified.

[0013] Preferably, after calculating the plane coordinates of the lifting object of the workshop crane, the position of the lifting object of the workshop crane in the image is identified according to the feature information obtained by the big data processing module, and then the height coordinates of the lifting object of the workshop crane are identified, including the following steps: S41: extracting the feature information obtained by the data processing module, and identifying the image information of the objects hoisted by the crane in the workshop through AI; S42: Identify the bottom contour of the lifting object of the workshop crane, and calculate the height of the lifting object according to the projection position of the bottom contour of the lifting object of the workshop crane in the image, wherein the principle formula is: , ; Where h is the height of the hoisted object, is the height of the HD camera, is the coordinate difference between the coordinates of the object hoisted by the workshop crane and one of the sets of high-definition cameras calculated in step S32, is the coordinate difference between the coordinates of the lifting object of the workshop crane and another set of high-definition cameras calculated in step S32, d is half the width of the lifting object, and S is the distance from the projection position of the bottom contour of the lifting object of the workshop crane in the image to the edge of the image; S43: At the same time, according to Calculate the width of the objects hoisted by the workshop crane, where Z is the distance between the two sets of high-definition cameras.

[0014] Preferably, the AI ​​analysis module includes the following steps when analyzing whether there is a person walking under the objects hoisted by the crane in the workshop according to the data results of the data processing module and the positioning module: S51: Divide the dangerous area according to the specific coordinates of the objects hoisted by the crane in the workshop, the height of the objects hoisted by the crane in the workshop, and the width of the objects hoisted by the crane in the workshop, which are identified by the positioning module; S52: Image recognition, using a recognition algorithm to perform target recognition on each segmented area and identify the area containing people; S53: Position determination and tracking: calculating the specific position coordinates of the person through the distance between the person's feet and the edge of the image, and tracking the person's image; S54: Risk assessment is performed based on the location information of personnel, information obtained by image tracking, the moving speed of the objects hoisted by the crane in the workshop, the moving direction of the objects hoisted by the crane in the workshop, and the designated dangerous areas.

[0015] Preferably, the control module controls the workshop crane according to the data results of the AI ​​analysis module, issues an early warning, and notifies personnel under the workshop crane hoisting objects, including the following steps: S61: When a collision risk is detected, a prompt message is first issued to notify the personnel to avoid the collision, and the vehicles in the workshop are controlled to slow down; S62: If the personnel are not away from the crane hoisting objects in the workshop, the crane hoisting objects in the workshop is controlled to stop moving; S63: When no personnel enter under the workshop crane hoisting objects, control the workshop crane hoisting objects to slowly descend to 30 cm from the ground.

[0016] Preferably, when using a recognition algorithm to perform target recognition on each segmented area and identifying an area containing a person, the following steps are included: S71: Data preparation, collecting a large amount of image data containing people, and annotating the data, wherein which areas in the image contain images of people are annotated; S72: Feature extraction: extract features from the annotated image through a convolutional neural network model and automatically learn useful features in the image; S73: Model training, inputting the extracted features and annotation information into the convolutional neural network model at the same time, and performing iterative training; S74: Model application, using the trained model to identify and predict the extracted image features, and output the prediction result, where the prediction result is the area containing the person image.

[0017] Beneficial effects of the present invention: 1. Compared with the existing technology that uses a single signal source or simple geometric algorithm for positioning, which is susceptible to environmental interference and has insufficient accuracy, this solution uses UWB modules and multi-base station layout, transmits and receives UWB pulse signals, extracts signal arrival time and intensity data, and uses complex positioning algorithm principle equations to accurately convert them into actual coordinate values, thereby achieving high-precision plane positioning of workshop crane hoisting objects. Furthermore, based on the acquisition of plane coordinates, combined with the feature information provided by the big data processing module, the position of the hoisting objects in the image is intelligently identified, thereby accurately calculating its height coordinates, comprehensively improving the three-dimensional spatial accuracy and intelligence level of the positioning system, and effectively meeting the precise positioning needs in complex workshop environments; 2. Compared with the existing technology that uses a single sensor or manual measurement method to identify the position and height of the hoisted objects in the workshop, which has the disadvantages of low precision, low efficiency and susceptibility to human errors, this solution uses a big data processing module combined with AI recognition technology to first extract feature information to identify the image position of the hoisted objects, and then calculates the height through a specific principle formula combined with the height and position coordinate difference of the high-definition camera and the projection position of the bottom contour of the hoisted objects in the image, and at the same time calculates the width according to the camera spacing. This solution has the advantages of high degree of automation, high measurement accuracy, significantly improved efficiency and reduced human errors; 3. Compared with the existing technology that uses simple physical fences or manual monitoring to prevent the risk of people walking under the objects hoisted by cranes in the workshop, there are disadvantages such as limited monitoring range, delayed response and easy missed reports. This solution uses an AI analysis module. First, it divides the dangerous area according to the data of the positioning module, and then uses the recognition algorithm to perform image recognition, accurately locate and track the position of personnel, and finally integrates information from multiple parties to conduct risk assessment. This solution has the advantages of a wide monitoring range, rapid response, high warning accuracy and can effectively reduce the risk of safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 What is shown is a schematic diagram of the structure of an AI-based production scene personnel supervision and protection system of the present invention; Figure 2 What is shown is a schematic diagram of the workflow of a positioning module in an AI-based production scene personnel supervision and protection system of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0020] Embodiment 1 See also Figure 1 The present invention provides an embodiment: an AI-based production scene personnel supervision and protection system, comprising: The data acquisition module is used to conduct real-time and uninterrupted monitoring of the objects hoisted by the crane in the workshop and collect high-definition images of the objects hoisted by the crane in the workshop; Communication module, used to provide a data transmission network to realize data transmission between the data acquisition module, positioning module, data processing module, AI analysis module and control module; A data processing module, used for processing the high-definition images collected by the data collection module; The positioning module is used to determine the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module; The AI ​​analysis module is used to analyze whether there are people walking under the objects hoisted by the crane in the workshop based on the data results of the data processing module and the positioning module; The control module is used to control the workshop crane according to the data results of the AI ​​analysis module, and issue an early warning to notify the personnel under the objects hoisted by the workshop crane.

[0021] Among them, the data acquisition module includes two groups of high-definition cameras, which are respectively set at the two ends of the workshop crane. The communication module includes four groups of UWB base stations and one group of UWB modules. The four groups of UWB base stations are respectively set at the four corners of the workshop, and the UWB module is set above the workshop crane hook.

[0022] As described above, in the workshop, the system first uses the high-definition camera of the data acquisition module to monitor the area below the hoisted objects in real time and continuously, and collects high-definition images. These images are then transmitted to the data processing module through the communication module for preprocessing, such as denoising and contrast enhancement, to improve image quality. At the same time, the positioning module uses the image data provided by the data acquisition module and other relevant information (such as the location and speed of the vehicle) transmitted by the communication module to accurately determine the location of the hoisted objects through image recognition technology and geometric algorithms. Then, the AI ​​analysis module receives the preprocessed image output by the data processing module and the location information of the hoisted objects provided by the positioning module. The AI ​​analysis module uses deep learning algorithms (such as convolutional neural networks) to extract features and identify targets from the image to determine whether there are people walking under the hoisted objects. In order to improve the accuracy and efficiency of recognition, the AI ​​analysis module may use a pre-trained model and fine-tune it in actual applications. Once the AI ​​analysis module detects that there are people walking under the hoisted objects, the control module will immediately receive this information and control the workshop driving according to the preset safety strategy. For example, the control module can issue a command to suspend the driving operation, or adjust the driving path to avoid collision with personnel. At the same time, the control module will also send a signal to the early warning system in the workshop through the communication module, triggering an audible and visual alarm to remind personnel under the hoisted objects to avoid it. During the specific operation process, the system may set a series of parameters to ensure the accuracy and timeliness of the detection. For example, parameters such as the resolution, frame rate, and monitoring range of the high-definition camera need to be set according to the actual situation of the workshop. The deep learning algorithm of the AI ​​analysis module also needs to be selected and adjusted according to the specific detection task. In addition, parameters such as the system's response time and the triggering conditions of the early warning mechanism also need to be carefully designed and optimized.

[0023] Embodiment 2 See also Figure 2In the AI-based production scene personnel supervision and protection system described in the first embodiment, when the data processing module processes the high-definition image collected by the data collection module, the following steps are included: S11: image preprocessing, preprocessing the collected high-definition image, wherein the preprocessing methods used include filtering, enhancement, smoothing, contrast adjustment and brightness adjustment; S12: Image segmentation: using image segmentation algorithm to segment the image into multiple regions with similar attributes; S13: Feature extraction, extracting required feature information from each segmented region image, wherein the required feature information includes color features, texture features and shape features.

[0024] Preferably, when filtering an image, the principle formula is: ; in, is the pixel value of the filtered image at coordinate (x, y), is the standard deviation of the Gaussian function, k is the size of the filter, i and j are loop variables used to traverse all elements of the filter, The input image is at coordinates The pixel value at is an element in the filter.

[0025] As described above, the present invention significantly improves the accuracy and reliability of the personnel detection system under the hoisted objects in the workshop through a meticulous data processing process. In the image preprocessing stage, the image noise is effectively reduced, the image details are enhanced, and a high-quality data foundation is provided for subsequent analysis through filtering, enhancement, smoothing, and fine adjustment of contrast and brightness. The image segmentation step uses advanced algorithms to divide the image into multiple regions with clear features, which is convenient for subsequent feature extraction. The feature extraction link accurately captures key information such as color, texture, and shape, which is crucial for the AI ​​analysis module to accurately determine whether there are people under the hoisted objects. In particular, the Gaussian filtering principle adopted in the filtering process ensures the retention of image details and effective suppression of noise through precise mathematical models, further improving the detection performance of the system. In summary, the optimized combination of this series of data processing steps provides a solid technical support for the efficient operation and accurate detection of the system.

[0026] Preferably, when using an image segmentation algorithm to segment an image into a plurality of regions with similar attributes, the following steps are included: S21: Initialize the cluster center and select the initial cluster center evenly on the image; S22: Search field, defining a search field for each cluster center, wherein the size of the search field is related to the scaling ratio of the high-definition camera; S23: Clustering, in the search domain, the color distance and spatial distance between each pixel and the cluster center are calculated, and the color distance and spatial distance are integrated into a comprehensive distance metric by a weighted sum method; S24: updating the cluster center, and updating the cluster center of each search area according to the clustering result, wherein the updated cluster center selects the average color value of all pixels in the search area as the new cluster center; S25: Iterate and repeat steps S22-S24 until the cluster center no longer changes significantly.

[0027] As described above, efficient and accurate segmentation of high-definition images is achieved through a sophisticated image segmentation algorithm. By uniformly initializing the cluster centers and combining the definition of the search field related to the zoom ratio of the high-definition camera, the adaptability of the algorithm to different scenes and resolutions is effectively improved. In the clustering process, the color distance and spatial distance are comprehensively considered to ensure the accuracy and consistency of the segmentation results. The method of iteratively updating the cluster center further improves the stability of the algorithm and the segmentation accuracy until the cluster center is stable, thereby obtaining multiple regions with similar attributes, which provides a reliable basis for subsequent feature extraction and object recognition.

[0028] Preferably, when the positioning module determines the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module, the principle steps are as follows: S31: signal transmission and reception, the UWB module transmits a UWB pulse signal, and after the UWB base station receives the pulse signal sent by the UWB module, it extracts the required data from the received pulse signal, wherein the required data includes the signal arrival time and signal strength; S32: Data processing and analysis: converting the extracted data into actual coordinate values ​​through a positioning algorithm, wherein the principle equation group of the positioning algorithm used is: ; ; ; in, is the coordinate of the uwb module, that is, the plane coordinate of the crane hoisting items in the workshop. , , and are the coordinates of four groups of UWB base stations, c is the speed of light, , , and They are the time when the pulse signals sent by the uwb module arrive at the four groups of uwb base stations; S33: Height coordinate analysis. After calculating the plane coordinates of the objects hoisted by the crane in the workshop, the position of the objects hoisted by the crane in the workshop is identified in the image according to the feature information obtained by the big data processing module, and then the height coordinates of the objects hoisted by the crane are identified.

[0029] As mentioned above, compared with the existing technology that uses a single signal source or simple geometric algorithm for positioning, which is susceptible to environmental interference and lacks accuracy, this solution uses UWB (ultra-wideband) modules and multi-base station layouts, transmits and receives UWB pulse signals, extracts signal arrival time and intensity data, and uses complex positioning algorithm principle equations to accurately convert them into actual coordinate values, thus achieving high-precision plane positioning of workshop crane hoisting objects. Furthermore, based on the acquisition of plane coordinates, combined with the feature information provided by the big data processing module, the position of the hoisting objects in the image is intelligently identified, so as to accurately calculate its height coordinates, comprehensively improving the three-dimensional spatial accuracy and intelligence level of the positioning system, and effectively meeting the precise positioning needs in complex workshop environments.

[0030] Preferably, after calculating the plane coordinates of the lifting object of the workshop crane, the position of the lifting object of the workshop crane in the image is identified according to the feature information obtained by the big data processing module, and then the height coordinates of the lifting object of the workshop crane are identified, including the following steps: S41: extracting the feature information obtained by the data processing module, and identifying the image information of the objects hoisted by the crane in the workshop through AI; S42: Identify the bottom contour of the lifting object of the workshop crane, and calculate the height of the lifting object according to the projection position of the bottom contour of the lifting object of the workshop crane in the image, wherein the principle formula is: , ; Where h is the height of the hoisted object, is the height of the HD camera, is the coordinate difference between the coordinates of the object hoisted by the workshop crane and one of the sets of high-definition cameras calculated in step S32, is the coordinate difference between the coordinates of the lifting object of the workshop crane and another set of high-definition cameras calculated in step S32, d is half the width of the lifting object, and S is the distance from the projection position of the bottom contour of the lifting object of the workshop crane in the image to the edge of the image; S43: At the same time, according to Calculate the width of the objects hoisted by the workshop crane, where Z is the distance between the two sets of high-definition cameras.

[0031] As mentioned above, compared with the existing technology that uses a single sensor or manual measurement method to identify the position and height of the hoisted objects in the workshop, which has the disadvantages of low precision, low efficiency and susceptibility to human errors, this solution uses a big data processing module combined with AI recognition technology to first extract feature information to identify the image position of the hoisted objects, and then calculate the height through a specific principle formula combined with the height and position coordinate difference of the high-definition camera and the projection position of the bottom contour of the hoisted objects in the image, and at the same time calculate the width based on the camera spacing. This solution has the advantages of high degree of automation, high measurement accuracy, significantly improved efficiency and reduced human errors.

[0032] Preferably, the AI ​​analysis module includes the following steps when analyzing whether there is a person walking under the objects hoisted by the crane in the workshop according to the data results of the data processing module and the positioning module: S51: Divide the dangerous area according to the specific coordinates of the objects hoisted by the crane in the workshop, the height of the objects hoisted by the crane in the workshop, and the width of the objects hoisted by the crane in the workshop, which are identified by the positioning module; S52: Image recognition, using a recognition algorithm to perform target recognition on each segmented area and identify the area containing people; S53: Position determination and tracking: calculating the specific position coordinates of the person through the distance between the person's feet and the edge of the image, and tracking the person's image; S54: Risk assessment is performed based on the location information of personnel, information obtained by image tracking, the moving speed of the objects hoisted by the crane in the workshop, the moving direction of the objects hoisted by the crane in the workshop, and the designated dangerous areas.

[0033] As mentioned above, compared with the existing technology that uses simple physical fences or manual monitoring to prevent the risk of people walking under the crane hoisting objects in the workshop, there are shortcomings such as limited monitoring range, delayed response and easy missed reports. This solution uses an AI analysis module to first divide the dangerous area according to the data of the positioning module, and then use the recognition algorithm to perform image recognition, accurately locate and track the position of personnel, and finally integrate multiple information for risk assessment. This solution has the advantages of wide monitoring range, rapid response, high warning accuracy and can effectively reduce the risk of safety accidents.

[0034] Preferably, the control module controls the workshop crane according to the data results of the AI ​​analysis module, issues an early warning, and notifies personnel under the workshop crane hoisting objects, including the following steps: S61: When a collision risk is detected, a prompt message is first issued to notify the personnel to avoid the collision, and the vehicles in the workshop are controlled to slow down; S62: If the personnel are not away from the crane hoisting objects in the workshop, the crane hoisting objects in the workshop is controlled to stop moving; S63: When no personnel enter under the workshop crane hoisting objects, control the workshop crane hoisting objects to slowly descend to 30 cm from the ground.

[0035] Compared with the existing technology, when a person is found under the hoisted objects in the workshop, only simple parking measures may be taken, and there is a lack of early warning and gradual response mechanism. When the AI ​​analysis module finds the risk of collision with people through the control module, this solution first sends a prompt message and slows down. If the person is not far away, it stops moving and controls the hoisted objects to slowly descend to a safe height while ensuring safety. This gradually upgraded response strategy not only improves safety, but also effectively reduces the shaking of hoisted objects or the fright of people caused by sudden stopping, and has higher safety and operational flexibility.

[0036] Preferably, when using a recognition algorithm to perform target recognition on each segmented area and identifying an area containing a person, the following steps are included: S71: Data preparation, collecting a large amount of image data containing people, and annotating the data, wherein which areas in the image contain images of people are annotated; S72: Feature extraction: extract features from the annotated image through a convolutional neural network model and automatically learn useful features in the image; S73: Model training, inputting the extracted features and annotation information into the convolutional neural network model at the same time, and performing iterative training; S74: Model application, using the trained model to identify and predict the extracted image features, and output the prediction result, where the prediction result is the area containing the person image.

[0037] As mentioned above, compared with traditional manual annotation or simple image processing algorithms for target recognition, this technical solution uses a large amount of annotated personnel image data for deep learning, and uses a convolutional neural network model to automatically extract and learn useful features in the image. After iterative training, it can efficiently and accurately identify areas containing personnel images. This method not only significantly improves the accuracy and efficiency of recognition, but also reduces the cost and subjective errors of manual annotation, providing more reliable data support for subsequent monitoring and early warning systems.

[0038] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of those skilled in the art without departing from the spirit of the present invention.

Claims

1. An AI-based production scene personnel supervision and protection system; characterized by: Included are: The data acquisition module is used to conduct real-time and uninterrupted monitoring of the objects hoisted by the crane in the workshop and collect high-definition images of the objects hoisted by the crane in the workshop; Communication module, used to provide a data transmission network to realize data transmission between the data acquisition module, positioning module, data processing module, AI analysis module and control module; A data processing module, used for processing the high-definition images collected by the data collection module; The positioning module is used to determine the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module; The AI ​​analysis module is used to analyze whether there are people walking under the objects hoisted by the crane in the workshop based on the data results of the data processing module and the positioning module; The control module is used to control the workshop crane according to the data results of the AI ​​analysis module, and issue an early warning to notify the personnel under the objects hoisted by the workshop crane.

2. According to claim 1, the AI-based production scene personnel supervision and protection system is characterized by: The data acquisition module includes two groups of high-definition cameras, which are respectively set at both ends of the workshop crane. The communication module includes four groups of UWB base stations and one group of UWB modules. The four groups of UWB base stations are respectively set at the four corners of the workshop, and the UWB module is set above the workshop crane hook.

3. According to claim 2, the AI-based production scene personnel supervision and protection system is characterized by: When the data processing module processes the high-definition images collected by the data acquisition module, the following steps are included: S11: image preprocessing, preprocessing the collected high-definition image, wherein the preprocessing methods used include filtering, enhancement, smoothing, contrast adjustment and brightness adjustment; S12: Image segmentation: using image segmentation algorithm to segment the image into multiple regions with similar attributes; S13: Feature extraction, extracting required feature information from each segmented region image, wherein the required feature information includes color features, texture features and shape features.

4. According to claim 3, the AI-based production scene personnel supervision and protection system is characterized by: When filtering an image, the principle formula is: ； in, is the pixel value of the filtered image at coordinate (x, y), is the standard deviation of the Gaussian function, k is the size of the filter, i and j are loop variables used to traverse all elements of the filter, The input image is at coordinates The pixel value at is an element in the filter.

5. According to claim 4, the AI-based production scene personnel supervision and protection system is characterized by: When using an image segmentation algorithm to segment an image into multiple regions with similar attributes, the following steps are included: S21: Initialize the cluster center and select the initial cluster center evenly on the image; S22: Search field, defining a search field for each cluster center, wherein the size of the search field is related to the scaling ratio of the high-definition camera; S23: Clustering, in the search domain, the color distance and spatial distance between each pixel and the cluster center are calculated, and the color distance and spatial distance are integrated into a comprehensive distance metric by a weighted sum method; S24: updating the cluster center, and updating the cluster center of each search area according to the clustering result, wherein the updated cluster center selects the average color value of all pixels in the search area as the new cluster center; S25: Iterate and repeat steps S22-S24 until the cluster center no longer changes significantly.

6. According to claim 5, the AI-based production scene personnel supervision and protection system is characterized by: When the positioning module determines the position of the objects hoisted by the crane in the workshop according to the data of the data acquisition module and the data of the communication module, the principle steps are as follows: S31: signal transmission and reception, the UWB module transmits a UWB pulse signal, and after the UWB base station receives the pulse signal sent by the UWB module, it extracts the required data from the received pulse signal, wherein the required data includes the signal arrival time and signal strength; S32: Data processing and analysis: converting the extracted data into actual coordinate values ​​through a positioning algorithm, wherein the principle equation group of the positioning algorithm used is: ； ； ； in, is the coordinate of the uwb module, that is, the plane coordinate of the crane hoisting items in the workshop. , , and are the coordinates of four groups of UWB base stations, c is the speed of light, , , and They are the time when the pulse signals sent by the uwb module arrive at the four groups of uwb base stations; S33: Height coordinate analysis. After calculating the plane coordinates of the objects hoisted by the crane in the workshop, the position of the objects hoisted by the crane in the workshop is identified in the image according to the feature information obtained by the big data processing module, and then the height coordinates of the objects hoisted by the crane are identified.

7. According to claim 6, the AI-based production scene personnel supervision and protection system is characterized by: After calculating the plane coordinates of the objects hoisted by the crane in the workshop, the position of the objects hoisted by the crane in the workshop is identified in the image according to the feature information obtained by the big data processing module, and then the height coordinates of the objects hoisted by the crane in the workshop are identified, including the following steps: S41: extracting the feature information obtained by the data processing module, and identifying the image information of the objects hoisted by the crane in the workshop through AI; S42: Identify the bottom contour of the lifting object of the workshop crane, and calculate the height of the lifting object according to the projection position of the bottom contour of the lifting object of the workshop crane in the image, wherein the principle formula is: ， ； Where h is the height of the hoisted object, is the height of the HD camera, is the coordinate difference between the coordinates of the object hoisted by the workshop crane and one of the sets of high-definition cameras calculated in step S32, is the coordinate difference between the coordinates of the lifting object of the workshop crane and another set of high-definition cameras calculated in step S32, d is half the width of the lifting object, and S is the distance from the projection position of the bottom contour of the lifting object of the workshop crane in the image to the edge of the image; S43: At the same time, according to Calculate the width of the objects hoisted by the workshop crane, where Z is the distance between the two sets of high-definition cameras.

8. The AI-based production scene personnel supervision and protection system according to claim 7 is characterized by: The AI ​​analysis module analyzes whether there are people walking under the objects hoisted by the crane in the workshop based on the data results of the data processing module and the positioning module, including the following steps: S51: Divide the dangerous area according to the specific coordinates of the objects hoisted by the crane in the workshop, the height of the objects hoisted by the crane in the workshop, and the width of the objects hoisted by the crane in the workshop, which are identified by the positioning module; S52: Image recognition, using a recognition algorithm to perform target recognition on each segmented area and identify the area containing people; S53: Position determination and tracking: calculating the specific position coordinates of the person through the distance between the person's feet and the edge of the image, and tracking the person's image; S54: Risk assessment is performed based on the location information of personnel, information obtained by image tracking, the moving speed of the objects hoisted by the crane in the workshop, the moving direction of the objects hoisted by the crane in the workshop, and the designated dangerous areas.

9. The AI-based production scene personnel supervision and protection system according to claim 8 is characterized by: The control module controls the workshop crane according to the data results of the AI ​​analysis module, issues an early warning, and notifies the personnel under the workshop crane hoisting objects, including the following steps: S61: When a collision risk is detected, a prompt message is first issued to notify the personnel to avoid the collision, and the vehicles in the workshop are controlled to slow down; S62: If the personnel are not away from the crane hoisting objects in the workshop, the crane hoisting objects in the workshop is controlled to stop moving; S63: When no personnel enter under the workshop crane hoisting objects, control the workshop crane hoisting objects to slowly descend to 30 cm from the ground.

10. The AI-based production scene personnel supervision and protection system according to claim 9 is characterized by: When using the recognition algorithm to identify the target in each segmented area and identify the area containing people, the following steps are included: S71: Data preparation, collecting a large amount of image data containing people, and annotating the data, wherein which areas in the image contain images of people are annotated; S72: Feature extraction: extract features from the annotated image through a convolutional neural network model and automatically learn useful features in the image; S73: Model training, inputting the extracted features and annotation information into the convolutional neural network model at the same time, and performing iterative training; S74: Model application, using the trained model to identify and predict the extracted image features, and output the prediction result, where the prediction result is the area containing the person image.

Images