Safety warning automatic recognition system based on machine vision technology

Abstract

The present invention discloses a safety warning automatic identification system based on machine vision technology, which relates to the technical field of safety warning, including a data acquisition unit, an environment warning unit, a cargo warning unit, a personnel warning unit, a behavior warning unit and a display output unit. The present invention can scan the faces of all personnel in the current storage environment through the personnel warning unit. Once it is found that the facial information of the personnel is not similar to the facial data stored in the storage device, it will be monitored to determine the actual distance between the personnel and the cargo. If the personnel is not found to be close to the cargo, there is no need to give a warning prompt, thereby avoiding waste of manpower. At the same time, when the personnel is close, a warning prompt can be given in time to ensure the safety of the cargo. The actions of the personnel are monitored by the behavior warning unit, which can ensure the safety of the personnel themselves on the one hand, and avoid the impact of the personnel on the cargo on the other hand.

Description

Technical Field

[0001] The present invention relates to the technical field of safety warning, and in particular to a safety warning automatic recognition system based on machine vision technology. Background Art

[0002] Machine vision technology is to replace artificial vision with common machine vision. In large-scale repetitive industrial production processes, the use of machine vision detection methods can greatly improve production efficiency and automation. Safety warning is a means of responding to emergencies. Setting up safety warnings in a storage environment can effectively protect the safety of goods. The invention patent with publication number CN111459083A discloses "a storage safety warning system, including a data storage unit, a front-end data acquisition unit, a storage data monitoring unit, a working time setting unit, an image recognition unit and an alarm unit. The data storage unit is used to input and store the number of items stored in the storage unit; the front-end data acquisition unit collects and stores the input data; the storage data monitoring unit is used to monitor the changes in data in the front-end data acquisition unit; the working time setting unit is used to monitor the changes in the number of storage data monitoring units within the working time; the image recognition unit is used to monitor the changes in the images of different products in the storage data monitoring unit; the alarm unit is used to alarm when the number of storage monitoring units in the working time setting unit is reduced to less than 3. The present invention warns the number and image of stored goods in the warehouse to ensure that the number of stored goods is sufficient.";

[0003] The above-mentioned existing technology solves the problem of lack of resource network for controlling and managing logistics in the warehousing environment. However, when the system is running, it cannot automatically analyze the optimal temperature range according to the storage temperature range of the goods. It completely relies on the operator to set it by himself, which will waste a lot of time. In addition, it is impossible to provide early warning for defective goods, resulting in secondary damage during the transportation process. At the same time, it is impossible to monitor the behavior of personnel and cannot guarantee the safety of personnel themselves and goods. Summary of the invention

[0004] The purpose of the present invention is to provide a safety warning automatic identification system based on machine vision technology to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a safety warning automatic identification system based on machine vision technology, comprising a data acquisition unit, an environmental warning unit and a display output unit;

[0006] A cargo early warning unit, after obtaining the cargo external related picture information data and image sample data through the data acquisition unit, the cargo early warning unit extracts the related image features, builds a detection model using a defect detection algorithm, completes the model training through the image sample data, and transmits the cargo external related picture information data to the model for analysis and calculation. If the analysis determines that there are defects on the outside of the cargo, the cargo is numbered and marked as defective. After determining the defect ratio and location related information, the information is stored in a database, and a search engine is used to query the database according to the number, and the complete cargo related information is transmitted to the processor. If the processor calculates that the validity period of the cargo is less than thirty days, the cargo related information is retrieved, and the related information and early warning prompts are transmitted to the processor;

[0007] A personnel warning unit, wherein the personnel warning unit uses an infrared detector to photograph the faces of people in the storage environment and transmits the faces to the processor through an interface, and the processor uses a feature processing algorithm to extract the contours of the acquired image information and the image information in the storage device, and calculates the registration value of the collected image information and the image information in the storage device through the feature processing algorithm. If it is determined that the person is not entered in the storage device, the person is photographed by the monitoring device, and the image is analyzed by the image measuring device to determine the actual distance between the person and the goods. When the distance is less than three meters but the person does not touch the goods, the current distance information and the warning prompt are transmitted to the processor;

[0008] A behavior warning unit is provided. The behavior warning unit uses monitoring equipment to take pictures of people in a storage environment. The obtained image information is analyzed by a processor. The human body contour in the image information is extracted using a feature matrix. After the contour features are obtained through analysis, an abnormal behavior recognition algorithm is used to determine the current behavior of the person based on the contour features. If it is determined that the person jumps, crawls, or lies flat, the information related to the abnormal behavior and warning prompts are transmitted to the processor.

[0009] Preferably, the data acquisition unit includes an environmental data acquisition module, a cargo data acquisition module, a personnel information acquisition module and a sample information acquisition module. The environmental data acquisition module detects the temperature and humidity in the storage environment using a temperature and humidity sensor, and transmits the relevant data obtained by the detection to a storage device. The air composition in the current storage environment is detected by a gas detector, and the gas type and content-related data obtained by the detection are transmitted to the storage device. After the cargo data acquisition module obtains the data related to the validity period and storage temperature range of the cargo, it transmits it to the storage device, uses an industrial camera to shoot the outside of the cargo in the detection area, and transmits the obtained image information to the host computer. The personnel information acquisition module uses a CCD camera to shoot a person to obtain a facial image, and after obtaining the personal information of the person, the facial image is associated with the personal information and then transmitted to the storage device. After the sample information acquisition module collects multiple groups of normal cargo packaging and defective cargo packaging image sample data, it transmits them to the storage device. The host computer model is IPC-610L equipment of Aikong Industrial Control, and the CCD camera is specifically a digital camera with a charge-coupled device image sensor.

[0010] Preferably, the environmental early warning unit includes a range statistics module, an initial range generation module, an optimal temperature analysis module and a range setting module. After the range statistics module obtains the storage temperature range related data of all the goods in the data acquisition unit, it counts the storage temperature ranges of these goods. The initial range generation module selects the maximum value of the storage temperature range of these goods as the highest value of the initial temperature range, and the minimum value of the storage temperature range as the lowest value of the initial temperature range, so as to determine the initial temperature range. The optimal temperature analysis module uses a processor to calculate the difference of the storage temperature range of each kind of goods, sorts the goods according to the difference, and then compares the storage temperature range of a single goods with the initial temperature range in turn, analyzes and obtains a new temperature range, and after the comparison is completed, determines the new temperature range as the optimal temperature range. The range setting module sets the optimal temperature range, the optimal humidity range, and the gas type and content range, and the set optimal temperature range has a higher priority than the optimal temperature range generated by the processor. The processor model is Intel 13400F processor.

[0011] Preferably, the cargo early warning unit includes an information acquisition module, a detection model building module and a defect detection module. After the information acquisition module obtains the external image information data and image sample data of the cargo through the data acquisition unit, the upper computer is used to identify and analyze the image information and extract relevant image features. After the detection model building module uses the defect detection algorithm to build a detection model, the image sample data is input into the model, and the weight of the model is updated through the loss function to complete the training of the model. The defect detection module transmits the external image information data of the cargo to the model, and the model is analyzed and calculated. If the analysis determines that there is no defect on the outside of the cargo, the cargo is numbered and marked as defect-free, and the information is stored in the database. If the analysis determines that there is a defect on the outside of the cargo, the cargo is numbered and marked as defective. After determining the defect ratio and location-related information, the information is stored in the database. The defect detection algorithm is specifically as follows:

[0012] ;

[0013] in, Indicates that the input sample belongs to The probability of the class, represents the output of the fully connected layer elements, represents the output of the fully connected layer elements, represents the total number of categories, represents the input vector, Indicates the number, Indicates a parameter.

[0014] Preferably, the cargo early warning unit also includes a cargo extraction module and an expiration date analysis module. After the cargo extraction module obtains the cargo-related number, it uses a search engine to query the database according to the number, and transmits the complete cargo-related information to the processor. After the expiration date analysis module obtains the cargo expiration date-related data through the data acquisition unit, if the processor calculates that the expiration date is less than thirty days, it uses a search engine to retrieve the cargo-related information from the database, and transmits the relevant information and early warning prompts to the processor. The search engine is specifically a Navicat database queryer.

[0015] Preferably, the personnel warning unit includes an image acquisition module, a registration analysis module and a distance processing module. The image acquisition module uses an infrared detector to shoot the faces of people in the storage environment, and transmits the acquired image information to the processor through an interface. The registration analysis module uses a feature processing algorithm to uniformly process the grayscale of the acquired image information and the image information in the storage device through the processor, and then extracts the contour of the image information. The feature processing algorithm is used to calculate the registration value of the collected image information and the image information in the storage device according to the contour extraction feature. If the value is lower than 0.5, If the person is not recorded in the storage device, the distance processing module determines that the person is not recorded in the storage device, and analyzes the image after shooting the person through the monitoring device to determine the actual distance between the person and the goods. When the distance is greater than three meters, no prompt will be given. When the distance is less than three meters but the person has not touched the goods, the current distance information and warning prompt will be transmitted to the processor. When the person has touched the goods, an alarm will be directly sent to the monitoring center. The feature processing algorithm includes a grayscale processing algorithm, and the grayscale processing algorithm is specifically:

[0016] ;

[0017] in, represents infrared and visible light face images, Indicates the maximum grayscale value of infrared and visible light images, Indicates the minimum grayscale value of infrared and visible light images, represents the normalized or initial grayscale values ​​of infrared and visible light images, Indicates the grayscale range of the infrared and visible light images.

[0018] Preferably, the behavior warning unit includes a feature matrix determination module, a contour feature analysis module and an abnormal behavior determination module. The feature matrix determination module uses monitoring equipment to take pictures of people in a storage environment, and the obtained image information is analyzed by a processor to determine the feature matrix and delete useless points in the image. The contour feature analysis module uses the feature matrix to extract the human body contour in the image information, and after analyzing the contour features, uses an abnormal behavior recognition algorithm to determine the current behavior of the person according to the contour features. If the abnormal behavior determination module determines that the person does not jump, crawl or lie flat, no prompt will be given. If it is determined that the person jumps, crawls or lies flat, the behavior abnormality related information and warning prompt will be transmitted to the processor, and the timer will be started for timing. If the prompt is thirty seconds later, the person still has abnormal behavior, an alarm will be directly sent to the monitoring center.

[0019] Preferably, the display output unit includes a warning level classification module, an environmental warning display module, a cargo warning display module, a personnel warning display module and a behavior warning display module. The warning level classification module divides the warning levels into three categories, namely the first level, the second level and the third level. When the warning level is the first level, a prompt will be given through a visual interface. When the warning level is the second level, a prompt will be given through a visual interface and a loudspeaker will be used for broadcasting. When the warning level is the third level, a corresponding solution will be generated using deep learning technology, and the solution will be output through a visual interface and a loudspeaker will be used for broadcasting. After the environmental warning display module obtains the current storage environment temperature, humidity, gas type and content related data information and warning prompts, it determines the The level of the warning prompt is determined, and corresponding operations are performed according to the level. After the cargo warning display module obtains the complete information related to the cargo, it will be output through a visual interface. If the cargo is marked as defective, the warning prompt level is determined according to the proportion and location of the defects in the cargo, and corresponding operations are performed according to the level. When obtaining information related to the validity period of the cargo and the warning prompt, the level of the warning prompt is determined based on these data, and corresponding operations are performed according to the level. After the personnel warning display module uses the processor to obtain the current distance information and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level. After the behavior warning display module obtains information related to abnormal behavior and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. The present invention collects information on storage environment, goods, personnel and other aspects through a data acquisition unit, which is convenient for subsequent data analysis. The environmental early warning unit is used to collect statistics on the storage temperature range of the goods, and the processor automatically analyzes the optimal temperature range. The operator can also independently set the optimal temperature range, the optimal humidity range, and the gas type and content range according to their own needs. When the relevant data of the storage environment exceeds the set range, an early warning prompt can be issued in time, which is convenient for the operator to take measures to intervene as soon as possible. The cargo early warning unit can detect the outside of the goods. If there are defects on the outside of the goods, they can be discovered in time through the detection model. When storing the relevant information of the goods, the defective goods will be marked. When the goods are extracted, an early warning prompt can be issued in time to prevent the goods from being damaged again during transportation due to defects in the packaging, which greatly ensures the safety of the goods. In addition, the unit will also issue an early warning prompt for goods with a validity period of less than thirty days to avoid expiration of the goods.

[0022] 2. The present invention can scan the faces of all personnel in the current storage environment through the personnel warning unit. Once it is found that the facial information of a person is not similar to the facial data stored in the storage device, it will be monitored to determine the actual distance between the person and the goods. If the person is not found to be close to the goods, there is no need to issue a warning prompt, thus avoiding waste of manpower. At the same time, when a person approaches, a warning prompt can be issued in time to ensure the safety of the goods. The actions of the person are monitored through the behavior warning unit, which can ensure the safety of the person himself on the one hand, and prevent the person from affecting the goods on the other hand. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Provides a schematic diagram of the overall system flow for an embodiment of the present invention;

[0024] Figure 2 An internal module block diagram of a data acquisition unit provided by an embodiment of the present invention;

[0025] Figure 3 An internal module block diagram of an environmental early warning unit provided by an embodiment of the present invention;

[0026] Figure 4 A block diagram of the internal modules of the cargo warning unit provided by an embodiment of the present invention;

[0027] Figure 5 A block diagram of the internal modules of a personnel warning unit provided in an embodiment of the present invention;

[0028] Figure 6 A block diagram of the internal modules of the behavior warning unit provided by an embodiment of the present invention;

[0029] Figure 7 This is a block diagram of the internal modules of the display output unit provided by an embodiment of the present invention.

[0030] In the figure: 1. Data acquisition unit; 101. Environmental data acquisition module; 102. Cargo data acquisition module; 103. Personnel information acquisition module; 104. Sample information acquisition module; 2. Environmental early warning unit; 201. Range statistics module; 202. Initial range generation module; 203. Optimal temperature analysis module; 204. Range setting module; 3. Cargo early warning unit; 301. Information acquisition module; 302. Detection model building module; 303. Defect detection module; 304. Cargo extraction module; 305. Validity period analysis module; 4. Personnel warning unit; 401. Image acquisition module; 402. Registration analysis module; 403. Distance processing module; 5. Behavior warning unit; 501. Feature matrix determination module; 502. Contour feature analysis module; 503. Abnormal behavior determination module; 6. Display output unit; 601. Warning level classification module; 602. Environmental warning display module; 603. Cargo warning display module; 604. Personnel warning display module; 605. Behavior warning display module. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] See also Figures 1 to 7 , the present invention provides a technical solution: a safety warning automatic identification system based on machine vision technology, including a data acquisition unit 1, an environmental warning unit 2 and a display output unit 6;

[0033] The cargo warning unit 3 obtains the cargo external image information data and the image sample data through the data acquisition unit 1, extracts the relevant image features, builds a detection model using the defect detection algorithm, completes the model training through the image sample data, and transmits the cargo external image information data to the model for analysis and calculation. If the analysis determines that there are defects on the outside of the cargo, the cargo is numbered and marked as defective. After determining the defect ratio and location related information, the information is stored in the database, and the search engine is used to query the database according to the number, and the complete cargo related information is transmitted to the processor. If the processor calculates that the validity period of the cargo is less than thirty days, the cargo related information is retrieved, and the related information and warning prompts are transmitted to the processor;

[0034] Personnel warning unit 4, personnel warning unit 4 uses an infrared detector to shoot the face of the person in the storage environment and transmits it to the processor through an interface, and the processor uses a feature processing algorithm to extract the contour of the acquired image information and the image information in the storage device, and calculates the registration value of the collected image information and the image information in the storage device through the feature processing algorithm. If it is determined that the person is not entered in the storage device, the person is photographed by the monitoring device, and the image is analyzed by the image measurement device to determine the actual distance between the person and the goods. When the distance is less than three meters but the person does not touch the goods, the current distance information and warning prompts are transmitted to the processor;

[0035] The behavior warning unit 5 uses the monitoring equipment to take pictures of people in the storage environment, and the obtained image information is analyzed by the processor, and the human body contour in the image information is extracted using the feature matrix. After the contour features are obtained through analysis, the abnormal behavior recognition algorithm is used to determine the current behavior of the person based on the contour features. If it is determined that the person jumps, crawls, or lies flat, the information related to the abnormal behavior and the warning prompt are transmitted to the processor.

[0036] The data collection unit 1 includes an environmental data collection module 101, a cargo data collection module 102, a personnel information collection module 103 and a sample information collection module 104. The environmental data collection module 101 uses a temperature and humidity sensor to detect the temperature and humidity in the storage environment, and transmits the relevant data obtained by the detection to a storage device. The gas detector detects the air composition in the current storage environment, and transmits the gas type and content related data obtained by the detection to the storage device. After the cargo data collection module 102 obtains the data related to the validity period and storage temperature range of the cargo, it transmits it to the storage device. , using an industrial camera to shoot the outside of the goods in the detection area, and transmitting the obtained image information to the host computer, the personnel information acquisition module 103 uses a CCD camera to shoot the personnel to obtain a facial image, and after obtaining the personal information of the personnel, the facial image is associated with the personal information and then transmitted to the storage device, the sample information acquisition module 104 collects multiple groups of normal goods packaging and defective goods packaging image sample data, and transmits them to the storage device, the host computer model is IPC-610L equipment of Aikong Industrial Control, and the CCD camera is specifically a digital camera with a charge-coupled device image sensor;

[0037] The environmental early warning unit 2 includes a range statistics module 201, an initial range generation module 202, an optimal temperature analysis module 203 and a range setting module 204. After the range statistics module 201 obtains the storage temperature range related data of all the goods in the data acquisition unit 1, it counts the storage temperature ranges of these goods. The initial range generation module 202 selects the maximum value of the storage temperature range of these goods as the highest value of the initial temperature range, and the minimum value of the storage temperature range as the lowest value of the initial temperature range, thereby determining the initial temperature range. The optimal temperature analysis module 203 uses a processor to calculate the difference of the storage temperature range of each kind of goods, sorts the goods according to the difference, and compares the storage temperature range of a single goods with the initial temperature range in turn, analyzes and obtains a new temperature range, and after the comparison is completed, determines the new temperature range as the optimal temperature range. The range setting module 204 sets the optimal temperature range, the optimal humidity range, and the gas type and content range, and the set optimal temperature range has a higher priority than the optimal temperature range generated by the processor, and the processor model is Intel 13400F processor;

[0038] The cargo early warning unit 3 includes an information acquisition module 301, a detection model building module 302 and a defect detection module 303. After the information acquisition module 301 obtains the external image information data and image sample data of the cargo through the data acquisition unit 1, the upper computer is used to identify and analyze these image information and extract relevant image features. After the detection model building module 302 uses the defect detection algorithm to build a detection model, the image sample data is input into the model, and the weight of the model is updated through the loss function to complete the training of the model. The defect detection module 303 transmits the external image information data of the cargo to the model, and the model is analyzed and calculated. If the analysis determines that there is no defect on the outside of the cargo, the cargo is numbered and marked as defect-free, and the information is stored in the database. If the analysis determines that there is a defect on the outside of the cargo, the cargo is numbered and marked as defective. After determining the defect ratio and location related information, the information is stored in the database. The defect detection algorithm is specifically as follows:

[0039] ;

[0040] in, Indicates that the input sample belongs to The probability of the class, represents the output of the fully connected layer elements, represents the output of the fully connected layer elements, represents the total number of categories, represents the input vector, Indicates the number, Indicates parameters;

[0041] The cargo early warning unit 3 also includes a cargo extraction module 304 and an effective period analysis module 305. After the cargo extraction module 304 obtains the cargo related number, it uses a search engine to query the database according to the number and transmits the complete cargo related information to the processor. After the effective period analysis module 305 obtains the cargo effective period related data through the data acquisition unit 1, if the processor calculates that the effective period is less than thirty days, it uses a search engine to retrieve the cargo related information in the database and transmits the related information and early warning prompts to the processor. The search engine is specifically a Navicat database query device.

[0042] The personnel warning unit 4 includes an image acquisition module 401, a registration analysis module 402 and a distance processing module 403. The image acquisition module 401 uses an infrared detector to shoot the faces of people in the storage environment, and transmits the acquired image information to the processor through an interface. The registration analysis module 402 uses a feature processing algorithm to uniformly process the acquired image information and the grayscale of the image information in the storage device through the processor, and then extracts the contour of the image information. The feature processing algorithm is used to calculate the registration value of the collected image information and the image information in the storage device according to the contour extraction feature. If the value is lower than 0.5, it is determined that the person is not recorded in the storage device, otherwise it is determined that the person has been recorded in the storage device. If the distance processing module 403 determines that the person is not recorded in the storage device, the person is photographed by the monitoring device, and the image is analyzed to determine the actual distance between the person and the goods. When the distance is greater than three meters, no prompt will be given. When the distance is less than three meters but the person has not touched the goods, the current distance information and the early warning prompt will be transmitted to the processor. When the person has touched the goods, an alarm will be directly sent to the monitoring center. The feature processing algorithm includes a grayscale processing algorithm, and the grayscale processing algorithm is specifically:

[0043] ;

[0044] in, represents infrared and visible light face images, Indicates the maximum grayscale value of infrared and visible light images, Indicates the minimum grayscale value of infrared and visible light images, represents the normalized or initial grayscale values ​​of infrared and visible light images, Indicates the grayscale range of the infrared and visible light images;

[0045] The feature processing algorithm also includes a contour extraction algorithm, which is specifically:

[0046] ;

[0047] in, Represents the directional characteristics of any point P on the edge contour, Represents pixel The gray value of Represents pixel The gray value of Represents pixel The gray value of Represents pixel The gray value of Indicates parameters;

[0048] The behavior warning unit 5 includes a feature matrix determination module 501, a contour feature analysis module 502 and an abnormal behavior determination module 503. After the feature matrix determination module 501 uses the monitoring equipment to shoot the personnel in the storage environment, the obtained image information is analyzed by the processor to determine the feature matrix and delete the useless points in the image. The contour feature analysis module 502 uses the feature matrix to extract the human body contour in the image information. After the contour features are obtained through analysis, the abnormal behavior recognition algorithm is used to determine the current behavior of the personnel according to the contour features. If the abnormal behavior determination module 503 determines that the personnel does not jump, crawl or lie flat, no prompt will be given. If it is determined that the personnel jump, crawl or lie flat, the abnormal behavior related information and warning prompt will be transmitted to the processor, and the timer will be started for timing. If the prompt is thirty seconds later, the personnel still has abnormal behavior, then the monitoring center will be directly alarmed. The abnormal behavior recognition algorithm is specifically as follows:

[0049] ;

[0050] in, Indicates the aspect ratio, Indicates the minimum bounding rectangle of the contour feature The coordinate value on the axis, Indicates the minimum bounding rectangle of the contour feature The coordinate value on the axis, Indicates the width of the minimum bounding rectangle. Indicates the height of the minimum enclosing rectangle;

[0051] The display output unit 6 includes a warning level classification module 601, an environmental warning display module 602, a cargo warning display module 603, a personnel warning display module 604 and a behavior warning display module 605. The warning level classification module 601 divides the warning levels into three categories, namely the first level, the second level and the third level. When the warning level is the first level, a prompt will be given through a visual interface. When the warning level is the second level, a prompt will be given through a visual interface and a loudspeaker will be used to broadcast. When the warning level is the third level, a corresponding solution is generated using deep learning technology, and it is output through a visual interface and broadcasted using a loudspeaker. After the environmental warning display module 602 obtains the current storage environment temperature, humidity, gas type and content related data information and warning prompts, it determines the The level of the warning prompt is determined, and corresponding operations are performed according to the level. After the cargo warning display module 603 obtains the complete information related to the cargo, it outputs it through a visual interface. If the cargo is marked as defective, the warning prompt level is determined according to the proportion and location of the defects in the cargo, and corresponding operations are performed according to the level. When the processor is used to obtain information related to the validity period of the cargo and the warning prompt, the level of the warning prompt is determined based on these data, and corresponding operations are performed according to the level. After the personnel warning display module 604 obtains the current distance information and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level. After the behavior warning display module 605 uses the processor to obtain information related to abnormal behavior and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level.

[0052] Working principle: The present invention uses the environmental data acquisition module 101 in the data acquisition unit 1 to obtain the temperature, humidity, gas type and content related data in the current storage environment, the cargo data acquisition module 102 obtains the validity period of the cargo and the storage temperature range related data, and takes pictures of the exterior of the cargo, and transmits the image information to the host computer, and takes pictures of the facial images of the personnel through the personnel information acquisition module 103, and transmits them to the storage device, and the sample information acquisition module 104 collects multiple groups of normal cargo packaging and defective cargo packaging image sample data and transmits them to the storage device, and uses the range statistics module 201 in the environmental early warning unit 2 to collect all cargo The storage temperature range of the goods is counted. After the initial range generation module 202 determines the initial temperature range, the optimal temperature analysis module 203 uses the goods storage temperature range to calculate the optimal temperature range. At the same time, the optimal temperature range, the optimal humidity range, and the gas type and content range can also be set through the range setting module 204. The information acquisition module 301 in the goods early warning unit 3 is used to identify and analyze the image information. The detection model building module 302 constructs a detection model and trains it to determine the weight of the model. The defect detection module 303 detects the outside of the goods and transmits the obtained information to the database. The goods extraction module 304 uses Use the serial number to query the cargo information. The validity period analysis module 305 uses the processor to calculate the validity period, and transmits the cargo information with a validity period of less than thirty days to the processor. The image acquisition module 401 in the personnel warning unit 4 uses an infrared detector to shoot the face of the person. The registration analysis module 402 calculates the registration value of the image information. The distance processing module 403 determines the actual distance between the person and the cargo. When the distance is less than three meters but the person does not touch the cargo, the current distance information and the warning prompt are transmitted to the processor. The feature matrix is ​​obtained through the feature matrix determination module 501 in the behavior warning unit 5 and useless points in the image are deleted. The contour feature analysis module 502 determines the current behavior of the personnel according to the contour features. When the personnel performs abnormal actions, the abnormal behavior determination module 503 transmits the abnormal behavior related information and warning prompts to the processor. The warning level classification module 601 in the display output unit 6 divides the warning level into three categories. The environmental warning display module 602 performs corresponding operations according to the warning level based on the storage environment information. The cargo warning display module 603 performs corresponding operations according to the warning level based on the cargo related information. The personnel warning display module 604 performs corresponding operations according to the warning level based on the current distance information. The behavior warning display module 605 performs corresponding operations according to the warning level based on the abnormal behavior related information.

[0053] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0054] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The automatic safety warning recognition system based on machine vision technology is characterized by: A data acquisition unit (1), the data acquisition unit (1) comprising an environment data acquisition module (101), a cargo data acquisition module (102), a personnel information acquisition module (103) and a sample information acquisition module (104), the environment data acquisition module (101) using a temperature and humidity sensor to detect the temperature and humidity in a storage environment, and transmitting the relevant data obtained by the detection to a storage device, detecting the air composition in the current storage environment by means of a gas detector, and transmitting the relevant data on the gas type and content obtained by the detection to the storage device, the cargo data acquisition module (102) acquiring data related to the validity period and storage temperature range of the cargo, and transmitting the data to the storage device, using an industrial camera to photograph the outside of the cargo within the detection area, and transmitting the obtained image information to a host computer, the personnel information acquisition module (103) using a CCD camera to photograph a person to obtain a facial image, and after obtaining the personal information of the person, the facial image is associated with the personal information and transmitted to the storage device, the sample information acquisition module (104) collecting a plurality of groups of normal cargo packaging and defective cargo packaging image sample data, and transmitting the data to the storage device; An environmental early warning unit (2), the environmental early warning unit (2) comprising a range statistics module (201), an initial range generation module (202), an optimal temperature analysis module (203) and a range setting module (204), wherein the range statistics module (201) obtains the storage temperature range related data of all goods in the data acquisition unit (1), and then performs statistics on the storage temperature ranges of these goods, the initial range generation module (202) selects the maximum value of the storage temperature range of these goods as the maximum value of the initial temperature range, and the minimum value of the storage temperature range as the minimum value of the initial temperature range, thereby determining the initial temperature range, the optimal temperature analysis module (203) uses a processor to calculate the difference between the storage temperature ranges of each type of goods, sorts the goods according to the difference, and then compares the storage temperature range of each single goods with the initial temperature range in turn, analyzes and obtains a new temperature range, and after the comparison is completed, determines the new temperature range as the optimal temperature range, the range setting module (204) sets the optimal temperature range, the optimal humidity range, and the gas type and content range, and the set optimal temperature range has a higher priority than the optimal temperature range generated by the processor; A cargo early warning unit (3), wherein the cargo early warning unit (3) obtains cargo external related image information data and image sample data through the data acquisition unit (1), extracts related image features, builds a detection model using a defect detection algorithm, completes model training using the image sample data, transmits cargo external related image information data to the model for analysis and calculation, and if the analysis determines that there are defects on the cargo external, the cargo is numbered and marked as defective, and after determining the defect ratio and location related information, the information is stored in a database, and a search engine is used to query the database according to the number, and the complete cargo related information is transmitted to the processor. If the processor calculates that the cargo validity period is less than thirty days, the cargo related information is retrieved, and the related information and early warning prompt are transmitted to the processor. The defect detection algorithm is specifically as follows: ; in, Indicates that the input sample belongs to The probability of the class, represents the output of the fully connected layer elements, represents the output of the fully connected layer elements, represents the total number of categories, represents the input vector, Indicates the number, Represents parameters; A personnel warning unit (4), wherein the personnel warning unit (4) uses an infrared detector to photograph the face of a person in the storage environment and transmits the face to a processor through an interface, and the processor uses a feature processing algorithm to extract the contour of the acquired image information and the image information in the storage device, and uses the feature processing algorithm to calculate the registration value of the acquired image information and the image information in the storage device. If it is determined that the person is not entered in the storage device, the person is photographed by a monitoring device, and the image is analyzed by an image measuring device to determine the actual distance between the person and the goods. When the distance is less than three meters but the person does not touch the goods, the current distance information and the warning prompt are transmitted to the processor; A behavior warning unit (5), wherein the behavior warning unit (5) uses a monitoring device to take a picture of a person in a storage environment, obtains image information, and is analyzed by a processor. A feature matrix is ​​used to extract a human body contour in the image information. After the contour features are obtained through analysis, an abnormal behavior recognition algorithm is used to determine the current behavior of the person based on the contour features. If it is determined that the person is jumping, crawling, or lying flat, information related to the abnormal behavior and a warning prompt are transmitted to the processor; The display output unit (6) comprises an early warning level classification module (601), an environmental early warning display module (602), a cargo early warning display module (603), a personnel early warning display module (604) and a behavior early warning display module (605). The early warning level classification module (601) classifies the early warning levels into three categories, namely, the first level, the second level and the third level. When the early warning level is the first level, a prompt is given through a visual interface. When the early warning level is the second level, a prompt is given through a visual interface and a loudspeaker is used for broadcasting. When the early warning level is the third level, a corresponding solution is generated using deep learning technology, and the solution is output through a visual interface and a loudspeaker is used for broadcasting. The environmental early warning display module (602) obtains data information related to the current storage environment temperature, humidity, and gas type and content. After the warning prompt is given, the level of the warning prompt is determined based on these data, and corresponding operations are performed according to the level. After the cargo warning display module (603) obtains the complete information related to the cargo, it outputs it through a visual interface. If the cargo is marked as defective, the warning prompt level is determined based on the proportion and location of the defects in the cargo, and corresponding operations are performed according to the level. When obtaining information related to the validity period of the cargo and the warning prompt, the level of the warning prompt is determined based on these data, and corresponding operations are performed according to the level. After the personnel warning display module (604) obtains the current distance information and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level. After the behavior warning display module (605) obtains information related to abnormal behavior and the warning prompt, it determines the level of the warning prompt based on these data, and performs corresponding operations according to the level.

2. The automatic safety warning identification system based on machine vision technology according to claim 1 is characterized in that: The cargo early warning unit (3) comprises an information acquisition module (301), a detection model building module (302) and a defect detection module (303). After the information acquisition module (301) acquires cargo external related image information data and image sample data through the data acquisition unit (1), the image information is identified and analyzed by a host computer to extract relevant image features. After the detection model building module (302) builds a detection model using a defect detection algorithm, the image sample data is input into the model, and the weight of the model is updated through a loss function, thereby completing the training of the model. The defect detection module (303) transmits cargo external related image information data to the model, and the model performs analysis and calculation. If the analysis determines that there is no defect on the cargo external, the cargo is numbered and marked as defect-free, and the information is stored in a database. If the analysis determines that there is a defect on the cargo external, the cargo is numbered and marked as defective, and after determining the defect ratio and location related information, the information is stored in the database.

3. The automatic safety warning identification system based on machine vision technology according to claim 2 is characterized in that: The cargo early warning unit (3) further comprises a cargo extraction module (304) and an expiration date analysis module (305). After obtaining the cargo related number, the cargo extraction module (304) uses a search engine to search in a database according to the number and transmits the complete cargo related information to the processor. After obtaining the cargo related expiration date data through the data acquisition unit (1), if the processor calculates that the expiration date is less than thirty days, the expiration date analysis module (305) uses a search engine to retrieve the cargo related information from the database and transmits the related information and early warning prompt to the processor.

4. The automatic safety warning identification system based on machine vision technology according to claim 1 is characterized in that: The personnel warning unit (4) comprises an image acquisition module (401), a registration analysis module (402) and a distance processing module (403). The image acquisition module (401) uses an infrared detector to photograph the face of a person in a storage environment, and transmits the acquired image information to a processor through an interface. The registration analysis module (402) uses a feature processing algorithm to uniformly process the acquired image information and the grayscale of the image information in the storage device through the processor, and then extracts the contour of the image information. The feature processing algorithm is used to calculate the collected image information and the grayscale of the image information in the storage device according to the contour extraction feature. The registration value of the image information in the device, if the value is lower than 0.5, it is determined that the person is not recorded in the storage device, otherwise it is determined that the person has been recorded in the storage device. If the distance processing module (403) determines that the person is not recorded in the storage device, the person is photographed by the monitoring device, and the image is analyzed to determine the actual distance between the person and the goods. When the distance is greater than three meters, no prompt will be given. When the distance is less than three meters but the person has not touched the goods, the current distance information and warning prompt will be transmitted to the processor. When the person has touched the goods, an alarm will be directly sent to the monitoring center.

5. The automatic safety warning identification system based on machine vision technology according to claim 1 is characterized in that: The behavior warning unit (5) comprises a feature matrix determination module (501), a contour feature analysis module (502) and an abnormal behavior determination module (503). The feature matrix determination module (501) uses monitoring equipment to take pictures of people in a storage environment, and the obtained image information is analyzed by a processor to determine a feature matrix and delete useless points in the image. The contour feature analysis module (502) uses the feature matrix to extract the human body contour in the image information. After analyzing and obtaining the contour features, an abnormal behavior recognition algorithm is used to determine the current behavior of the person according to the contour features. If the abnormal behavior determination module (503) determines that the person does not jump, crawl or lie flat, no prompt will be given. If the person jumps, crawls or lies flat, the information related to the abnormal behavior and the warning prompt will be transmitted to the processor, and a timer will be started to count. If the person still has abnormal behavior after thirty seconds of prompting, an alarm will be directly sent to the monitoring center.

Images