Online detection method and system for foreign matters in meat products

Abstract

The invention discloses a meat product foreign matter on-line detection method and system, and belongs to the field of visual detection.The detection method comprises the steps that a fixed detection area is set, a multi-mode sensor is installed above the fixed detection area, and the multi-mode sensor collects data and transmits the data to a data processing center through a wireless network; the data processing center decompresses and preprocesses the received data and fuses the image data twice to obtain a final fused image; according to the feature information in the final fusion, extracting fusion feature data; analyzing the feature data by using a support vector machine, and judging whether foreign matters exist or not and classifying the types of the foreign matters; a detection result is fed back to a production line control system in real time, and meat products containing foreign matters are positioned and removed. According to the method, the multi-mode sensor and an efficient data analysis technology are combined, and the accuracy and the real-time performance of meat product foreign matter detection are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of visual inspection, and in particular to an online detection method and system for foreign matter in meat products. Background Art

[0002] During the production and processing of meat products, foreign matter will inevitably be mixed in. The sources of these foreign matter are varied, including exogenous foreign matter such as metal, plastic, sand and stone, as well as endogenous foreign matter such as meat and poultry hair and meat decay.

[0003] However, existing detection technologies still have certain limitations in practical applications. Traditional metal detectors are mainly used to detect metal foreign bodies, but their detection capabilities for non-metallic foreign bodies such as plastic, glass, sand and stone are relatively weak, which can easily lead to missed detections. Although X-ray imaging technology can detect a variety of foreign bodies, its recognition accuracy for small-sized foreign bodies is insufficient, and it is easy to misdetect certain high-density foreign bodies (such as bones and meat, where the density difference is small). Another common detection method is visual image recognition technology. Although it has certain advantages in detecting surface foreign bodies, it still has certain bottlenecks in efficiently processing large quantities of products and its application in high-speed production lines. Due to limitations of factors such as lighting, image quality, and sensor resolution, the accuracy and stability of existing visual inspection systems are still difficult to meet high standards for food safety.

[0004] In addition, although the existing multi-sensor fusion technology can improve the accuracy of foreign body detection by combining different types of sensors, it still faces problems such as data redundancy, low fusion efficiency, and poor real-time performance in data processing. The amount of data collected by multiple sensors is huge, and how to avoid data redundancy and calculation delays while ensuring high accuracy is still a major challenge facing current technology.

[0005] Therefore, although the existing non-destructive testing technology has improved the foreign body detection capability in meat production to a certain extent, there are still many technical bottlenecks that cannot fully and efficiently meet the production line's needs for real-time, accuracy and cost control. This has also prompted the continuous research and development of more advanced testing technologies and methods in order to achieve more accurate and efficient foreign body detection and further ensure consumer food safety. Summary of the invention

[0006] One of the purposes of the present invention is to provide an online foreign body detection method for meat products to solve the problem of low foreign body detection capability in meat product production in the prior art.

[0007] The present invention is implemented by the following technical scheme. A method for online detection of foreign matter in meat products comprises the following steps: S100, setting a fixed detection area, installing a multimodal sensor above the detection area, and the meat product stops when passing through the detection area. At this time, the multimodal sensor is started to detect the meat product in the detection area from multiple dimensions, and the data collected by the multimodal sensor is pre-processed by a wireless network node arranged in the detection area and then sent to a data processing center, wherein the multimodal sensor is composed of a mutually coupled terahertz detector, an X-ray imager, a metal detector and a visual image sensor; S200, after receiving the sent data, the data processing center first decompresses the data, integrates the data with the same timestamp into a group, and extracts the terahertz image data and the visual image data for image preprocessing, fuses the pre-processed terahertz image data and the visual image data for the first time, and fuses the image obtained by the first fusion with the X-ray data for the second time to obtain the final fused image; S300, according to the terahertz image data, the terahertz image data and the visual image data are fused for the first time, and the image obtained by the first fusion is fused for the second time with the X-ray data to obtain the final fused image. The reflection intensity and frequency characteristics in the Hertz image data, the density distribution in the X-ray data, the color and shape characteristics in the visual image are combined with the signal strength of the metal detector to extract features from the final fused image to obtain fused feature data; S400, the fused feature data is analyzed by a support vector machine, the support vector machine maps the data to a high-dimensional space, finds the optimal classification hyperplane, classifies whether the meat product contains foreign matter and the type of foreign matter, extracts feature vectors based on the second fused image and combined with the fused feature data, constructs an optimal hyperplane to maximize the interval between the two types of data, foreign matter and no foreign matter, uses the Lagrange multiplier method and the dual problem solution to obtain the optimal parameters, and finally classifies through the optimal hyperplane to determine whether there is foreign matter and determine the type of foreign matter; S500, according to the analysis results of step S400, locate the specific location according to the timestamp after the foreign matter is found, and classify according to the type of foreign matter, and at the same time feed back the detection results to the production line control system in real time, and the production line removes the meat products containing foreign matter according to the feedback.

[0008] Furthermore, the terahertz detector includes a terahertz wave transmitter and a terahertz wave receiver. The terahertz wave transmitter is used to transmit terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate terahertz image data for obtaining imaging data inside the meat product; the X-ray imager is used to detect foreign matter with large density differences, and the X-ray image collected by the X-ray imager is matched with the terahertz image data to confirm the type and location of the foreign matter; the metal detector is used to detect metal foreign matter and thus quickly identify the metal component; the visual image sensor is a high-definition visual camera, which is used to obtain the appearance image of the meat product.

[0009] Furthermore, preprocessing includes checking whether there are duplicate records in the data based on the timestamp of the data, deleting the duplicate data, and then classifying the data according to the specific type of sensor; dividing the data of different sensors into different time periods according to the timestamp, and then compressing and sending the data in the order of the time periods, and finally sending the compressed data to the data processing center through the TCP / IP protocol.

[0010] Furthermore, image preprocessing includes obtaining a clear image through filtering, smoothing and denoising, and enhancing the image through grayscale stretching, while further enhancing and denoising the enhanced image to remove small blocks of noise interference in the image and perform separate distortion correction on the visual image data.

[0011] Furthermore, feature extraction includes analyzing the reflection intensity and frequency information of terahertz waves through the data of terahertz sensors, capturing the electromagnetic properties of materials, and thus identifying different types of foreign matter; further analyzing the physical properties of foreign matter through density differences in X-ray data, and identifying its possible location and form; identifying foreign matter attached to the surface of meat products through color and shape characteristics of visual image data; and metal detectors providing intensity information of metal foreign matter for precise positioning of metal particles; and finally, fusing the data to obtain accurate feature data.

[0012] On the other hand, the present invention also provides an online detection system for foreign matter in meat products, the detection system includes a multimodal sensor module, a wireless network node module, a data processing module and an execution module, wherein the multimodal sensor module is composed of a mutually coupled terahertz detector, an X-ray imager, a metal detector and a visual image sensor, and is configured so that when the meat product passes through the detection area, the multimodal sensor installed above the detection area is started to detect the meat product in the detection area from multiple dimensions; the wireless network node module is arranged in the detection area and connected to the multimodal sensor module, and is configured to pre-process the data collected by the multimodal sensor and then send it to the data processing module; the data processing module is connected to the wireless network node module, and is configured to decompress the received data, integrate the data with the same timestamp into a group, and extract the terahertz image data and the visual image data for image preprocessing, fuse the preprocessed terahertz image data and the visual image data for the first time, and fuse the image obtained by the first fusion with the X-ray data for the second time The final fused image is extracted from the final fused image according to the reflection intensity and frequency characteristics in the terahertz image data, the density distribution in the X-ray data, the color and shape characteristics in the visual image and the signal strength of the metal detector to obtain fused feature data, and the fused feature data is analyzed by a support vector machine. The support vector machine maps the data to a high-dimensional space to find the optimal classification hyperplane to classify whether the meat product contains foreign matter and the type of foreign matter, extracts feature vectors according to the second fused image and in combination with the fused feature data, and constructs an optimal hyperplane to maximize the interval between the two types of data, foreign matter and no foreign matter, uses the Lagrange multiplier method and the dual problem solution to obtain the optimal parameters, and finally classifies through the optimal hyperplane to determine whether there is foreign matter and determine the type of foreign matter; the execution module is connected to the data processing module and is configured to locate to a specific position according to the timestamp according to the judgment result of the data processing module, and classify according to the type of foreign matter, and at the same time feed back the detection result to the production line control system in real time, and the production line removes the meat products containing foreign matter according to the feedback.

[0013] Furthermore, the terahertz detector in the multimodal sensor module includes a terahertz wave transmitter and a terahertz wave receiver. The terahertz wave transmitter is used to transmit terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate terahertz image data for obtaining imaging data inside the meat product; the X-ray imager is used to detect foreign objects with large density differences, and the X-ray image collected by the X-ray imager is matched with the terahertz image data to confirm the type and location of the foreign object; the metal detector is used to detect metal foreign objects so as to quickly identify the metal components; the visual image sensor is a high-definition visual camera, which is used to obtain the appearance image of the meat product.

[0014] Furthermore, the preprocessing in the wireless network node module includes checking whether there are duplicate records in the data based on the timestamp of the data, deleting the duplicate data, and then classifying the data according to the specific type of the sensor; dividing the data of different sensors into different time periods according to the timestamp, and then compressing and sending the data in the order of the time periods, and finally sending the compressed data to the data processing center through the TCP / IP protocol.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0016] 1. The present invention enhances the accuracy of foreign body detection in meat products through multi-sensor fusion. Through the combination of multiple sensors such as terahertz waves, X-ray imaging, metal detectors and visual image sensors, meat products can be fully detected from multiple angles. The advantages of each sensor are maximized through this multi-modal data acquisition method. And by fusing terahertz images, X-ray images, visual images and metal detector data, the type and position of foreign bodies can be more comprehensively and accurately identified and located, reducing the possible false detection and missed detection of a single sensor.

[0017] 2. The present invention has the characteristics of efficient data transmission and processing. It uses wireless network nodes for data transmission, avoids complex wiring problems, and improves the flexibility and scalability of the system. At the same time, through wireless transmission, the information exchange between sensors and data centers is more convenient. The data is pre-processed and compressed during the transmission process, which can effectively reduce the bandwidth occupation during the transmission process and improve the real-time performance and data transmission efficiency of the system. By removing duplicate data and small noise interference, the accuracy of subsequent analysis can be guaranteed.

[0018] 3. The present invention performs filtering, smoothing, denoising, grayscale stretching and other processing on terahertz images and visual images, which can improve image quality, reduce noise interference, and thus improve the accuracy of foreign body detection. In particular, distortion correction of visual images can avoid image distortion caused by imaging device errors, thereby improving the reliability of classification results. And by performing two rounds of fusion on the processed images, different types of sensor data are effectively combined to extract more meaningful features. The second fusion image provides more comprehensive feature data for the SVM classification algorithm, which helps to improve classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0020] Figure 1 This is a flowchart provided for Example 1 of the present invention.

[0021] Figure 2 This is a system block diagram provided for Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0023] Embodiment 1,

[0024] This embodiment discloses an online detection method for foreign matter in meat products. This method combines the data collected by the terahertz transmitter with other detection technologies (such as X-rays, metal detectors, visual image processing, etc.), thereby effectively improving the accuracy and reliability of foreign matter detection in meat products. Through data fusion and machine learning analysis, judgments can be quickly drawn and applied to the production line in real time, improving detection efficiency and ensuring product safety and quality. With the advancement of technology and the reduction of hardware costs, this multi-modal, multi-source data fusion detection system is expected to become an important technical means in the field of food safety in the future.

[0025] Figure 1 The flowchart of the method of this embodiment is shown. It can be seen from the figure that this embodiment includes the following steps:

[0026] Step 1: Select a fixed inspection area on the conveyor belt and install a terahertz detector, X-ray imager, metal detector, and visual image sensor above the inspection area.

[0027] The terahertz detector includes a terahertz transmitter and a receiver. The terahertz transmitter is responsible for emitting terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate imaging data. The terahertz detector is used to obtain imaging data inside meat products, and the data information obtained by terahertz is used as the central data for detection and analysis.

[0028] X-ray imagers are used to detect foreign bodies with large density differences, such as metal, bone, glass, etc. X-ray images can be combined with terahertz imaging images to confirm the type and location of foreign bodies.

[0029] Metal detectors are used to detect the presence of metallic foreign matter, especially in situations that are very sensitive to metallic foreign matter, so as to quickly identify the metal component.

[0030] High-definition visual cameras are used to obtain images of the appearance of meat products, especially to identify more obvious foreign objects such as plastic, broken glass, etc. Computer vision technology is used to process images and identify foreign objects attached to the surface of meat products.

[0031] Adjust the terahertz detector, X-ray imager, metal detector and visual image sensor to align with the center point of the fixed detection area. At the same time, couple the above detectors in hardware. When detecting, the meat products on the conveyor belt stop when passing the center point of the detection area. At this time, the above detectors will be started at the same time due to the coupling in hardware, and the meat products in the detection area will be detected from multiple dimensions.

[0032] The terahertz transmitter of the terahertz detector emits terahertz waves. When the terahertz waves pass through meat products, the areas with foreign matter will show reflection signals or transmission signals different from those in normal areas. The terahertz receiver converts these signals into terahertz image digital signals and sends them to the background database. The terahertz image signals of the terahertz receiver are used as basic data to reflect the internal structure and potential foreign matter of the meat products.

[0033] The X-ray imager obtains X-ray data through X-ray imaging and also sends it to the background database to display areas with density changes inside meat products. For example, foreign objects such as metal and bones have a higher density. Other detectors are not effective for foreign objects, but they can be well reflected by X-ray imagers.

[0034] The metal detector will generate a signal response to the metal material, which is also sent to the background database for subsequent data processing.

[0035] The visual image sensor acquires the surface image of the meat product through a high-resolution camera and uses computer vision technology to extract the features of foreign objects in the image (e.g., shape, color, etc.). It also sends the collected data to the background database for subsequent data processing.

[0036] Considering the difficulty of factory wiring, in this implementation, the collected data can be transmitted wirelessly. The specific contents include the following:

[0037] Corresponding wireless network nodes are deployed near the location of the fixed detection area. The wireless network nodes receive the information sent by the above sensors and pre-process the data information sent by the above sensors.

[0038] Preprocessing includes checking whether there are duplicate records in the data according to the timestamp, deleting the duplicate data, and then classifying the data according to the type of sensor; then dividing the data of different sensors into different time periods according to the timestamp, and then compressing and sending the data in the order of the time periods, and finally sending the compressed data to the data processing center through the TCP / IP protocol.

[0039] Step 2:

[0040] After receiving the sent data, the data processing center first decompresses the data.

[0041] Then the terahertz image data, X-ray data, metal detector data and visual image data with the same timestamp are integrated into a group, and the terahertz image data and visual image data in the same group are preprocessed, filtered, smoothed and denoised to obtain a clear image, and the image is enhanced by grayscale stretching. The denoising can be further enhanced to remove small noise interference in the image, and distortion correction can also be performed on the visual image data.

[0042] Then, the processed terahertz image data and visual image data are fused for the first time, and the image obtained by the first fusion is fused for the second time with the X-ray data to obtain the final fused image.

[0043] The data processing center extracts meaningful features from the second fusion image to obtain feature data based on the reflection intensity and frequency characteristics in the terahertz image data, the density distribution in the X-ray data, the color and shape characteristics in the visual image, and the signal strength of the metal detector.

[0044] Specifically, to extract meaningful feature packages, the electromagnetic properties of the material are first captured through the reflection intensity and frequency information of the terahertz wave, thereby identifying different types of foreign matter, such as metal and plastic. Subsequently, the physical properties of the foreign matter are further analyzed through the density differences in the X-ray image to identify its possible location and form. At the same time, the color and shape characteristics of the visual image can help identify foreign matter attached to the surface of meat products, such as broken glass or foreign particles. The metal detector provides the intensity information of the metal foreign matter, which helps to accurately locate the metal particles. By fusing these data from different sensors, more comprehensive and accurate feature data can be extracted, thus providing a reliable basis for subsequent classification and detection.

[0045] It should be noted that multi-sensor fusion images can provide richer spatial information and help distinguish foreign objects of different types that are close in position. For example, the combination of terahertz images and X-ray images can further confirm the type of foreign objects by comprehensively considering density and electromagnetic characteristics.

[0046] Comparison of foreign matter and meat: The fused image data can comprehensively consider the advantages of different sensors and compare the different physical and chemical properties of meat and foreign matter, thereby improving the accuracy of detection and reducing false detection and missed detection. By identifying and classifying various data features of foreign matter, including reflection intensity, frequency characteristics, density distribution, color, shape, metal detection signal, etc. These features are crucial for subsequent classification and foreign matter detection through support vector machines (SVM), which can significantly improve detection accuracy and ensure the reliability of detection results.

[0047] Step 3: Input the fused feature data into the classification algorithm for data analysis.

[0048] The classification algorithm adopted in this embodiment is a support vector machine (SVM), which maps the data into a high-dimensional space to find the optimal classification hyperplane to classify whether the meat products contain foreign matter and the type of foreign matter.

[0049] Specifically, it includes the following:

[0050] The feature vector is extracted based on the second fused image and combined with the fused feature data.

[0051] By constructing the optimal hyperplane, the interval between the two types of data, foreign objects and no foreign objects, is maximized.

[0052] The optimal parameters are obtained using the Lagrange multiplier method and the dual problem solution.

[0053] Finally, the optimal hyperplane is used for classification to determine whether there is a foreign body and the type of foreign body.

[0054] Step 4: After data analysis, determine in real time whether there are foreign objects in the meat products. If foreign objects are found, the specific location can be located according to the timestamp and classified according to the type of foreign objects, such as metal, bone, plastic, etc.

[0055] At the same time, the test results are fed back to the production line control system in real time. If foreign matter is detected, the production line can be automatically adjusted to remove the meat products containing foreign matter, thereby ensuring the quality and safety of meat products.

[0056] Embodiment 2,

[0057] In this embodiment, a foreign body online detection system for meat products is disclosed. Figure 2 The system block diagram of this embodiment is shown. It can be seen from the figure that this embodiment includes a multimodal sensor module, a wireless network node module, a data processing module and an execution module.

[0058] Among them, the multimodal sensor module is composed of mutually coupled terahertz detectors, X-ray imagers, metal detectors and visual image sensors. It is configured so that when the meat products pass through the detection area, they stop and the multimodal sensor installed above the detection area is started to detect the meat products in the detection area from multiple dimensions.

[0059] Specifically, the terahertz detector in the multimodal sensor module emits terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate terahertz image data for obtaining imaging data inside the meat product.

[0060] The X-ray imager is used to detect foreign objects with large density differences. The X-ray image collected by the X-ray imager is matched with the terahertz image data to confirm the type and location of the foreign object.

[0061] The metal detector is used to detect metallic foreign matter so as to quickly identify the metal component.

[0062] The visual image sensor is a high-definition visual camera, which is used to obtain the appearance image of the meat product.

[0063] The wireless network node module is arranged in the detection area and connected to the multimodal sensor module, and is configured to pre-process the data collected by the multimodal sensor and then send it to the data processing module.

[0064] Specifically, the preprocessing in the wireless network node module includes checking whether there are duplicate records in the data based on the timestamp of the data, deleting the duplicate data, and then classifying the data according to the specific type of sensor; dividing the data of different sensors into different time periods according to the timestamp, and then compressing and sending the data in the order of the time periods, and finally sending the compressed data to the data processing center through the TCP / IP protocol.

[0065] The data processing module is connected to the wireless network node module, and is configured to decompress the received data, integrate the data with the same timestamp into a group, and extract the terahertz image data and the visual image data for image preprocessing, fuse the preprocessed terahertz image data and the visual image data for the first time, and fuse the image obtained by the first fusion with the X-ray data for the second time to obtain the final fused image.

[0066] According to the reflection intensity and frequency characteristics in the terahertz image data, the density distribution in the X-ray data, the color and shape characteristics in the visual image and the signal strength of the metal detector, features are extracted from the final fused image to obtain fused feature data.

[0067] The fused feature data is analyzed by a support vector machine, which maps the data to a high-dimensional space to find the optimal classification hyperplane, classify whether the meat product contains foreign matter and the type of foreign matter, extract feature vectors based on the second fusion image in combination with the fused feature data, maximize the interval between the two types of data, foreign matter and no foreign matter, use the Lagrange multiplier method and the dual problem solution to obtain the optimal parameters, and finally classify through the optimal hyperplane to determine whether there is foreign matter and the type of foreign matter.

[0068] The execution module is connected to the data processing module, and is configured to locate the specific position according to the timestamp based on the judgment result of the data processing module, and classify the foreign matter according to the type. At the same time, the detection result is fed back to the production line control system in real time. The production line removes meat products containing foreign matter based on the feedback.

[0069] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for online detection of foreign matter in meat products, characterized in that: The detection method comprises: S100, setting a fixed detection area, installing a multimodal sensor above the detection area, and the meat product stops when passing through the detection area. At the same time, the multimodal sensor is started to detect the meat product in the detection area from multiple dimensions, and the data collected by the multimodal sensor is pre-processed by the wireless network nodes arranged in the detection area and then sent to the data processing center, The multimodal sensor is composed of a mutually coupled terahertz detector, an X-ray imager, a metal detector and a visual image sensor; S200, after receiving the sent data, the data processing center first decompresses the data and integrates the data with the same timestamp into a group. The terahertz image data and visual image data are extracted for image preprocessing, and the preprocessed terahertz image data and visual image data are first fused. The image obtained by the first fusion is fused with the X-ray data for a second time to obtain a final fused image; S300, extracting features from the final fused image to obtain fused feature data based on the reflection intensity and frequency features in the terahertz image data, the density distribution in the X-ray data, the color and shape features in the visual image, and the signal strength of the metal detector; S400, analyzing the fused feature data by a support vector machine, mapping the data to a high-dimensional space, finding an optimal classification hyperplane, classifying whether the meat product contains foreign matter and the type of the foreign matter, extracting a feature vector based on the second fusion image and in combination with the fused feature data, maximizing the interval between the two types of data, foreign matter and no foreign matter, using the Lagrange multiplier method and the dual problem solution to obtain the optimal parameters, and finally classifying by the optimal hyperplane to determine whether there is foreign matter and the type of foreign matter; S500. According to the analysis result of step S400, after the foreign matter is found, the specific position is located according to the timestamp, and the foreign matter is classified according to the type. At the same time, the detection result is fed back to the production line control system in real time. The production line removes the meat products containing foreign matter according to the feedback.

2. The method for online detection of foreign matter in meat products according to claim 1, characterized in that: The terahertz detector includes a terahertz wave transmitter and a terahertz wave receiver, the terahertz wave transmitter is used to transmit terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate terahertz image data for obtaining imaging data inside the meat product; The X-ray imager is used to detect foreign objects with large density differences. The X-ray image collected by the X-ray imager is matched with the terahertz image data to confirm the type and location of the foreign object; The metal detector is used to detect metallic foreign matter so as to quickly identify the metal component; The visual image sensor is a high-definition visual camera, which is used to obtain the appearance image of the meat product.

3. The method for online detection of foreign matter in meat products according to claim 1, characterized in that: The preprocessing includes checking whether there are duplicate records in the data according to the timestamp of the data, deleting the duplicate data, and then classifying the data according to the specific type of the sensor; The data from different sensors are divided into different time periods according to the timestamps, and then the data is compressed and sent in the order of the time periods. Finally, the compressed data is sent to the data processing center via TCP / IP protocol.

4. The method for online detection of foreign matter in meat products according to claim 1, characterized in that: The image preprocessing includes obtaining a clear image through filtering, smoothing and denoising, and enhancing the image through grayscale stretching, and further enhancing and denoising the enhanced image to remove small noise interference in the image and perform separate distortion correction on the visual image data.

5. The method for online detection of foreign matter in meat products according to claim 1, characterized in that: The feature extraction includes analyzing the reflection intensity and frequency information of the terahertz wave through the data of the terahertz sensor, capturing the electromagnetic characteristics of the material, and thus identifying different types of foreign matter; Through the density differences in the X-ray data, the physical characteristics of the foreign body are further analyzed to identify its possible location and shape; Identify foreign matter attached to the surface of meat products through the color and shape characteristics of visual image data; The metal detector provides the strength information of the metal foreign body, which is used to accurately locate the metal particles; Finally, the data is fused to obtain accurate feature data.

6. An online detection system for foreign matter in meat products, characterized in that: The detection system includes a multimodal sensor module, a wireless network node module, a data processing module and an execution module, wherein: The multimodal sensor module is composed of mutually coupled terahertz detectors, X-ray imagers, metal detectors, and visual image sensors, and is configured so that when the meat product passes through the detection area, it stops, and the multimodal sensor installed above the detection area is activated to detect the meat product in the detection area from multiple dimensions; The wireless network node module is arranged in the detection area and connected to the multimodal sensor module, and is configured to pre-process the data collected by the multimodal sensor and then send it to the data processing module; The data processing module is connected to the wireless network node module and configured to decompress the received data, integrate the data with the same timestamp into a group, extract the terahertz image data and the visual image data for image preprocessing, fuse the preprocessed terahertz image data and the visual image data for the first time, fuse the image obtained by the first fusion with the X-ray data for the second time to obtain the final fused image, According to the reflection intensity and frequency characteristics in the terahertz image data, the density distribution in the X-ray data, the color and shape characteristics in the visual image and the signal strength of the metal detector, the features are extracted from the final fused image to obtain fused feature data. The fused feature data is analyzed by a support vector machine, the support vector machine maps the data to a high-dimensional space, finds an optimal classification hyperplane, classifies whether the meat product contains foreign matter and the type of foreign matter, extracts a feature vector based on the second fusion image and the fused feature data, maximizes the interval between the two types of data, foreign matter and no foreign matter, uses the Lagrange multiplier method and the dual problem solution to obtain the optimal parameters, and finally classifies by the optimal hyperplane to determine whether there is foreign matter and the type of foreign matter; The execution module is connected to the data processing module, and is configured to locate the specific position according to the timestamp based on the judgment result of the data processing module, and classify the foreign matter according to the type. At the same time, the detection result is fed back to the production line control system in real time. The production line removes meat products containing foreign matter based on the feedback.

7. The on-line detection system for foreign matter in meat products according to claim 1, characterized in that: The terahertz detector in the multimodal sensor module includes a terahertz wave transmitter and a terahertz wave receiver, the terahertz wave transmitter is used to transmit terahertz waves, and the terahertz receiver captures reflected waves or transmitted waves to generate terahertz image data for obtaining imaging data inside the meat product; The X-ray imager is used to detect foreign objects with large density differences. The X-ray image collected by the X-ray imager is matched with the terahertz image data to confirm the type and location of the foreign object; The metal detector is used to detect metallic foreign matter so as to quickly identify the metal component; The visual image sensor is a high-definition visual camera, which is used to obtain the appearance image of the meat product.

8. The on-line detection system for foreign matter in meat products according to claim 1, characterized in that: The preprocessing in the wireless network node module includes checking whether there are duplicate records in the data according to the timestamp of the data, deleting the duplicate data, and then classifying the data according to the specific type of the sensor; dividing the data of different sensors into different time periods according to the timestamp, and then compressing and sending the data in the order of the time periods, and finally sending the compressed data to the data processing center through the TCP / IP protocol.

Images