A food packaging detection method based on image fusion

Abstract

This invention relates to the field of plastic seal inspection technology, and more particularly to a food plastic seal inspection method based on image fusion, comprising: acquiring multimodal images of food plastic seal packaging; determining a predicted noise type based on the visible light texture features of the infrared distribution map and the visible light image; determining a denoising method based on the predicted noise type; performing denoising preprocessing on each of the multimodal images according to the denoising method to output a corresponding image to be fused; fusing each image to be fused to produce a test fused image; determining the fusion accuracy based on the feature similarity area ratio between the test fused image and each image to be fused; and reducing the filtering intensity of the visible light image if the fusion accuracy does not meet the requirements. This invention quantifies the influence of environmental factors on the visible light texture features of the visible light image, thereby improving the accuracy of plastic seal inspection.

Description

Technical Field

[0001] This invention relates to the field of plastic seal inspection technology, and in particular to a food plastic seal inspection method based on image fusion. Background Technology

[0002] Food sealing is a packaging method that uses plastic film to seal food through technologies such as heat pressing and vacuum adsorption, primarily to isolate it from air and moisture to extend shelf life. However, in actual production, defects such as incomplete sealing, leaks, or wrinkles may occur due to mechanical deviations, material contamination, or process fluctuations, affecting the sealing effect. Currently, visual or pressure testing is commonly used to monitor sealing quality, but factors such as the reflection of transparent films, production line vibration, or temperature changes can interfere with testing accuracy, leading to misjudgments or missed detections.

[0003] Chinese Patent Publication No. CN113252695B discloses a method and device for detecting defects in plastic sealing films based on image processing. The detection method includes the following steps: conveying the material wrapped in the plastic sealing film to the image acquisition area of ​​a camera; performing real-time image acquisition; locating the plastic sealing film area and cropping the plastic sealing film; image enhancement; simultaneously extracting features from the plastic sealing film using two methods and analyzing the results; judging whether the plastic sealing film is damaged based on the analysis results of the two methods; if at least one of the analysis results indicates that the plastic sealing film is damaged, then the plastic sealing film is judged to be damaged; otherwise, the plastic sealing film is judged not to be damaged. This invention combines two methods to determine whether the plastic sealing film is damaged, which can effectively reduce the missed detection rate of plastic sealing film defects, alarm for detected defects, and avoid product quality degradation due to missed detection of plastic sealing film defects. It is evident that the image processing-based method and device for detecting defects in plastic sealing films suffer from the problem that environmental factors can affect the visible light texture features of visible light images. Summary of the Invention

[0004] Therefore, this invention provides a food plastic seal detection method based on image fusion to overcome the problem in the prior art where environmental factors affect the visible light texture features of visible light images.

[0005] To achieve the above objectives, the present invention provides a food plastic seal detection method based on image fusion, comprising:

[0006] Acquire multimodal images of food plastic-sealed packaging, including visible light images, infrared distribution maps, and X-ray images;

[0007] The predicted noise type is determined based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image;

[0008] The denoising method is determined based on the estimated noise type, including determining the denoising region and the image segmentation edge thickness based on the overlap between the heat diffusion range and the texture offset range in the infrared distribution map.

[0009] Alternatively, the filtering intensity of the visible light image can be determined based on the offset distance between the visible light image and the infrared distribution map after comparison with the edge mask;

[0010] Each of the multimodal images is preprocessed for denoising according to the denoising method described above to output the corresponding image to be fused.

[0011] Each of the images to be fused is fused to generate a fused image for test purposes;

[0012] The fusion accuracy is determined based on the proportion of feature similarity area between the fused test image and each of the images to be fused.

[0013] If the fusion accuracy does not meet the requirements, the filtering intensity of the visible light image is reduced;

[0014] The visible light image is filtered according to the stated filtering intensity to output a standard visible light image;

[0015] The standard visible light image is fused with the infrared distribution map image to be fused and the X-ray image to be fused to output the target fused image of food plastic packaging.

[0016] Furthermore, the texture association feature ratio is the ratio of the area of ​​the region in the infrared distribution map that has the same infrared texture features as the visible light texture features in the visible light image to the area of ​​the infrared distribution map.

[0017] Further, the predicted noise type is determined based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, including:

[0018] Infrared texture features from the infrared distribution map and visible light texture features from the visible light image are extracted respectively.

[0019] The area of ​​regions where the infrared texture features and the visible light texture features are identical is calculated.

[0020] The proportion of texture-related features is calculated based on the area and the area of ​​the infrared distribution map;

[0021] If the proportion of the texture-related features is greater than the preset second proportion, then the estimated noise type is determined to be the first type;

[0022] If the proportion of the texture-related features is greater than a preset first proportion and less than or equal to the preset second proportion, then the estimated noise type is determined to be the second type.

[0023] If the proportion of the texture-related features is less than or equal to the preset first proportion, then the estimated noise type is determined to be a standard type.

[0024] Furthermore, the denoising region and image segmentation edge thickness are adjusted based on the overlap between the heat diffusion range and texture offset range in the infrared distribution map, including:

[0025] If the estimated noise type is the first type, then obtain the heat diffusion range and the texture offset range;

[0026] The degree of overlap is calculated based on the heat diffusion range and the texture offset range;

[0027] Compare the degree of overlap with the preset degree of overlap;

[0028] The denoising region is adjusted to a non-diffusion region based on the trigger condition that the overlap degree is greater than the preset overlap degree, and the thickness of the image segmentation edge is increased.

[0029] Furthermore, the image segmentation edge thickness is positively correlated with the overlap degree.

[0030] Furthermore, the filtering intensity of the visible light image is determined based on the offset distance between the visible light image and the infrared distribution map after comparison with the edge mask, including:

[0031] If the estimated noise type is the second type, then the offset distance is obtained;

[0032] Compare the offset distance with the preset offset distance;

[0033] If the offset distance is greater than the preset offset distance, the filtering intensity of the visible light image is increased.

[0034] Furthermore, the filtering intensity of the visible light image is positively correlated with the offset distance.

[0035] Further, the fusion accuracy is determined based on the proportion of feature similarity area between the test fused image and each of the images to be fused, including:

[0036] The similarity area of ​​several features between the test fused image and each of the images to be fused is obtained respectively;

[0037] The proportion of several feature similarities is calculated based on the area of ​​each of the aforementioned features and the area of ​​the corresponding image to be fused.

[0038] A fusion error is determined if the proportion of similar area of ​​at least one of the features is less than the preset proportion of similar area of ​​the features.

[0039] Furthermore, the filtering intensity of the visible light image is negatively correlated with the proportion of the area of ​​feature similarity.

[0040] Furthermore, the feature similarity area ratio is the ratio of each feature similarity area to the area of ​​the corresponding single image to be fused.

[0041] Compared with the prior art, the beneficial effects of the present invention are that by determining the estimated noise type based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, the accuracy of detecting plastic-sealed food packaging is insufficient due to the similarity between the texture of the packaging printing pattern and the actual defects, such as scratches, caused by relying solely on visible light texture features. The present invention comprehensively analyzes the infrared texture features of the infrared distribution map and the visible light texture features in the visible light image to analyze whether there is a correlation between the visible light texture features in the visible light image and temperature changes in the environmental factors. This allows for the analysis of the corresponding noise type based on the degree of correlation between the visible light texture features in the visible light image and temperature changes in the environmental factors, providing a basis for subsequent noise reduction processing and improving the accuracy of plastic-sealed packaging detection.

[0042] Furthermore, by determining the denoising area based on the overlap between the heat diffusion range and the texture offset range in the infrared distribution map and increasing the image segmentation edge thickness, the problem of altered surface texture details of food packaging caused by temperature increases from independent devices near the food packaging is overcome. Increasing the image segmentation edge thickness increases the width of the strip region for image segmentation between the diffusion and non-diffusion areas, preventing temperature increases from independent devices from interfering with the surrounding non-diffusion areas and improving the accuracy of plastic seal detection.

[0043] Furthermore, by determining the filtering intensity of the visible light image based on the offset distance between the visible light image and the infrared distribution map after comparison with the edge mask, the problem of excessive influence of temperature changes in environmental factors on visible light is solved, thus improving the accuracy of plastic seal inspection.

[0044] Furthermore, the fusion accuracy is determined based on the proportion of similar areas between the test fused image and each of the images to be fused. If the fusion accuracy does not meet the requirements, the filtering intensity of the visible light image is reduced. This solves the problem that the test fused image formed by fusing each of the images to be fused loses the features of the images to be fused due to the high filtering intensity, which in turn causes the test fused image to fail to accurately reflect the sealing condition of the food plastic packaging surface, thus improving the accuracy of plastic sealing detection. Attached Figure Description

[0045] Figure 1This is an overall flowchart of the food plastic seal detection method based on image fusion according to an embodiment of the present invention;

[0046] Figure 2 This is a flowchart illustrating the process of determining the predicted noise type based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image in the food plastic seal detection method based on image fusion, according to an embodiment of the present invention.

[0047] Figure 3 This is a flowchart illustrating how the image fusion-based food packaging detection method adjusts the denoising region and the image segmentation edge thickness based on the overlap between the heat diffusion range and texture offset range in the infrared distribution map, according to an embodiment of the present invention.

[0048] Figure 4 This is a schematic diagram of the diffuse region, the non-diffuse region, and the strip-shaped region that segments the diffuse region and the non-diffuse region in the image fusion-based food sealing detection method according to an embodiment of the present invention.

[0049] In the figure, 1-diffusion region, 2-non-diffusion region, 3-strip region that segments the image between the diffusion region and the non-diffusion region. Detailed Implementation

[0050] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0051] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0052] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0053] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0054] Please see Figure 1 , Figure 2 , Figure 3 as well as Figure 4 As shown, these are respectively an overall flowchart of the food plastic seal detection method based on image fusion according to an embodiment of the present invention, a flowchart for determining the estimated noise type based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, a flowchart for adjusting the denoising area and the image segmentation edge thickness based on the overlap between the heat diffusion range and the texture offset range in the infrared distribution map, and a schematic diagram of the diffusion area, the non-diffusion area, and the strip-shaped area for image segmentation of the diffusion area and the non-diffusion area.

[0055] The present invention provides an image fusion-based food shrink wrapping detection method, comprising:

[0056] Step S1: Acquire multimodal images of the food plastic-sealed packaging, including visible light images, infrared distribution maps, and X-ray images;

[0057] Step S2: Determine the predicted noise type based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image;

[0058] Step S3: Determine the denoising method based on the estimated noise type, including determining the denoising region and image segmentation edge thickness based on the overlap between the heat diffusion range and texture offset range in the infrared distribution map.

[0059] Alternatively, the filtering intensity of the visible light image can be determined based on the offset distance between the visible light image and the infrared distribution map after comparison with the edge mask;

[0060] Step S4: Perform denoising preprocessing on each of the multimodal images according to the denoising method to output the corresponding image to be fused;

[0061] Step S5: Fuse each of the images to be fused to generate a fused image for testing;

[0062] Step S6: Determine the fusion accuracy based on the proportion of feature similarity area between the test fused image and each of the images to be fused;

[0063] Step S7: If the fusion accuracy does not meet the requirements, reduce the filtering intensity of the visible light image;

[0064] Step S8: Filter the visible light image according to the filtering intensity to output a standard visible light image;

[0065] Step S9: The standard visible light image is fused with the infrared distribution map image to be fused and the X-ray image to be fused to output the target fused image of the food plastic packaging.

[0066] Specifically, image fusion is a technique that integrates complementary information from multiple images (such as those taken by different sensors, focal lengths, or viewpoints) into a single high-quality image using algorithms. Its core lies in preserving key features (such as thermal targets in infrared images and textures in visible light) and eliminating redundancy (such as repetitive backgrounds), and it is divided into three levels: pixel-level (direct overlay or wavelet transform fusion), feature-level (matching key points such as edges), and decision-level (integrating high-level information).

[0067] Specifically, an edge mask is a binary template generated using image processing techniques to accurately extract or preserve edge regions in an image. Its principle is to first use an edge detection algorithm to identify contour lines in the original image, and then generate a black-and-white mask image—white areas (value 1) mark edge pixels, and black areas (value 0) represent non-edge parts.

[0068] Specifically, the predicted noise type is determined based on the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, including:

[0069] Infrared texture features from the infrared distribution map and visible light texture features from the visible light image are extracted respectively.

[0070] The area of ​​regions where the infrared texture features and the visible light texture features are identical is calculated.

[0071] The proportion of texture-related features is calculated based on the area and the area of ​​the infrared distribution map;

[0072] If the proportion of the texture-related features is greater than the preset second proportion, then the estimated noise type is determined to be the first type;

[0073] If the proportion of the texture-related features is greater than a preset first proportion and less than or equal to the preset second proportion, then the estimated noise type is determined to be the second type.

[0074] If the proportion of the texture-related features is less than or equal to the preset first proportion, then the estimated noise type is determined to be a standard type.

[0075] Specifically, the texture association feature ratio is the ratio of the area of ​​the region in the infrared distribution map that has the same infrared texture features as the visible light texture features in the visible light image to the area of ​​the infrared distribution map, and the area of ​​the infrared distribution map is the same size as the visible light image.

[0076] Specifically, the first type refers to changes in the surface texture details of food packaging caused by an increase in temperature from a separate device located near the food packaging.

[0077] The second type refers to changes in the surface texture details of food packaging due to changes in ambient temperature. Both the first and second types refer to the influence of temperature factors on the surface texture details of food packaging.

[0078] Specifically, the standard type refers to the influence of non-temperature factors on the texture details of the surface of food packaging.

[0079] Optionally, when the food plastic packaging is placed under external conditions of temperature: 35℃~39℃, the preset first proportion is generally taken in the range of [40%, 50%], and the preset first proportion is generally taken in the range of [60%, 80%].

[0080] Preferably, the food plastic-sealed packaging is placed under external conditions of 35℃~39℃, with the first preferred embodiment having a proportion of 45% and the second preferred embodiment having a proportion of 70%.

[0081] Those skilled in the art will understand that [40%, 50%], [60%, 80%], 45%, and 70% are several optional and preferred embodiments of the food plastic-sealed packaging under external conditions of 35℃~39℃. In actual application or implementation, those skilled in the art can make adaptive adjustments to the preset first percentage and the preset second percentage according to the actual application environment and application scenario.

[0082] In practice, the predicted noise type is determined by combining the infrared texture features of the infrared distribution map with the visible light texture features of the visible light image. This addresses the issue that the accuracy of detecting plastic-sealed food packaging solely based on visible light texture features is insufficient because the texture of the printed packaging pattern is similar to that of actual defects, such as scratches. By comprehensively analyzing the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, the correlation between the visible light texture features in the visible light image and temperature changes in the environmental factors is analyzed. This allows for the determination of the corresponding noise type based on the degree of correlation between the visible light texture features in the visible light image and temperature changes in the environmental factors, providing a basis for subsequent noise reduction processing and improving the accuracy of plastic-sealed packaging detection.

[0083] Specifically, the denoising region and image segmentation edge thickness are adjusted based on the overlap between the heat diffusion range and texture offset range in the infrared distribution map, including:

[0084] If the estimated noise type is the first type, then obtain the heat diffusion range and the texture offset range;

[0085] The degree of overlap is calculated based on the heat diffusion range and the texture offset range;

[0086] Compare the degree of overlap with the preset degree of overlap;

[0087] Based on the triggering condition that the overlap is greater than the preset overlap, the denoising region is adjusted to non-diffusion region 2, and the image segmentation edge thickness is increased.

[0088] Specifically, the overlapping area where the heat diffusion range and texture offset range meet the overlap degree greater than the preset overlap degree is denoted as diffusion area 1, the area outside diffusion area 1 on the surface of food plastic packaging is denoted as non-diffusion area 2, and the image segmentation edge thickness is the width of the strip-shaped area 3 formed by image segmentation of diffusion area 1 and non-diffusion area 2.

[0089] Specifically, the overlap between the heat diffusion range and the texture offset range in the infrared distribution map is the ratio of the area of ​​the overlapping region of the heat diffusion range and the texture offset range to the area of ​​the texture offset range.

[0090] Specifically, in the actual testing of food plastic-sealed packaging, the number of factors that cause texture shift in visible light images is much greater than the number of factors that cause heat diffusion or growth in infrared distribution images. Therefore, except in the extremely special case of a fire, the range of texture shift is larger than or at least not smaller than the range of heat diffusion.

[0091] Specifically, the heat diffusion range refers to the area of ​​heat increase in the infrared distribution map.

[0092] Specifically, texture offset range refers to the area enclosed by the offset texture in a visible light image.

[0093] Optionally, when the food plastic packaging is placed under external conditions of 35℃~39℃, the preset overlap degree is generally taken in the range of [70%, 90%].

[0094] Preferably, the food plastic-sealed packaging is placed under external conditions of 35℃~39℃, and the preferred embodiment of the preset overlap is 80%.

[0095] Those skilled in the art will understand that [70%, 90%] and 80% are several optional and preferred embodiments of the food plastic-sealed packaging under external conditions of temperature: 35℃~39℃. In actual application or implementation, those skilled in the art can make adaptive adjustments to the preset overlap degree according to the actual application environment and application scenario.

[0096] In practice, by determining the denoising area based on the overlap between the heat diffusion range and the texture offset range in the infrared distribution map and increasing the image segmentation edge thickness, the problem of altered surface texture details of food packaging caused by temperature increases from independent devices near the food packaging is overcome. Increasing the image segmentation edge thickness increases the width of the strip region 3 for image segmentation between the diffusion region 1 and the non-diffusion region 2, preventing temperature increases from independent devices from interfering with the surrounding non-diffusion region 2 and improving the accuracy of plastic seal detection.

[0097] Specifically, the image segmentation edge thickness is positively correlated with the degree of overlap.

[0098] In implementation, when the food plastic-sealed packaging is exposed to an external temperature of 35℃~39℃, if the overlap is greater than the preset overlap by less than 10%, the image segmentation edge thickness is adjusted to 1.1 times the current image segmentation edge thickness. If the overlap exceeds the preset overlap by more than 10%, the image segmentation edge thickness is adjusted to 1.1 times the current image segmentation edge thickness for every 10% exceeding the preset overlap. For example, in a possible embodiment, if the overlap exceeds the preset overlap by 20%, the image segmentation edge thickness is adjusted to 1.1 × 1.1 = 1.21 times the current image segmentation edge thickness.

[0099] Specifically, the filtering intensity of the visible light image is determined based on the offset distance between the visible light image and the infrared distribution map after comparison with the edge mask, including:

[0100] If the estimated noise type is the second type, then the offset distance is obtained;

[0101] Compare the offset distance with the preset offset distance;

[0102] If the offset distance is greater than the preset offset distance, the filtering intensity of the visible light image is increased.

[0103] Specifically, the offset distance between the visible light image and the infrared distribution map after edge masking is the maximum separation distance between the area occupied by the food plastic packaging surface image in the visible light image and the area occupied by the food plastic packaging surface image in the infrared distribution map with clear edges formed after edge masking.

[0104] Specifically, the maximum separation distance is the length of the longest straight line segment connecting the points on the edge curve of the area occupied by the food packaging surface image in the visible light image, the points on the edge curve of the area occupied by the food packaging surface image in the infrared distribution map with clear edges, and the geometric center point of the area occupied by the food packaging surface image in the visible light image.

[0105] Optionally, when the food plastic packaging is placed under external conditions of 35℃~39℃, the preset offset distance is generally taken in the range of [0.5mm, 2.0mm].

[0106] Preferably, in the case where the food plastic packaging is placed under external conditions of 35℃~39℃, the preferred embodiment of the preset offset distance is 0.8mm.

[0107] Those skilled in the art will understand that [0.5mm, 2.0mm] and 0.8mm are several optional and preferred embodiments of the food plastic packaging under external conditions of temperature: 35℃~39℃. In actual application or implementation, those skilled in the art can adaptively adjust the preset offset distance according to the actual application environment and application scenario.

[0108] In implementation, when the food plastic-sealed packaging is placed under external conditions of 35℃~39℃, and the offset distance is within 0.1mm of the preset offset distance, the peak signal-to-noise ratio (PSNR) of the visible light image is adjusted to 1.1 times the current PSNR of the visible light image. When the offset distance exceeds 0.1mm of the preset offset distance, the PSNR of the visible light image is adjusted to 1.1 times the current PSNR of the visible light image for every 0.1mm exceeding the preset offset distance. For example, in one possible embodiment, the offset distance is 0.1mm greater than the preset offset distance, and the PSNR of the visible light image is adjusted to 1.1 times the current PSNR of the visible light image. Here, the peak signal-to-noise ratio (PSNR) characterizes the filtering intensity.

[0109] It is understandable that peak signal-to-noise ratio is an indicator used by those skilled in the art to characterize the strength of filtering, and it is a conventional technique in the field, so it will not be elaborated here.

[0110] In practice, the filtering intensity of the visible light image is determined by comparing the visible light image and the infrared distribution map after passing through an edge mask. This solves the problem of excessive influence of temperature changes in the environment on the visible light and improves the accuracy of plastic sealing inspection.

[0111] Specifically, the filtering intensity of the visible light image is positively correlated with the offset distance.

[0112] Specifically, determining the fusion accuracy based on the proportion of feature similarity area between the test fused image and each of the images to be fused includes:

[0113] The similarity area of ​​several features between the test fused image and each of the images to be fused is obtained respectively.

[0114] The proportion of several feature similarities is calculated based on the area of ​​each of the aforementioned features and the area of ​​the corresponding image to be fused.

[0115] A fusion error is determined if the proportion of similar area of ​​at least one of the features is less than the preset proportion of similar area of ​​the features.

[0116] In practice, the fusion accuracy is determined based on the proportion of similar areas between the test fused image and each of the images to be fused. If the fusion accuracy does not meet the requirements, the filtering intensity of the visible light image is reduced. This solves the problem that the test fused image formed by fusing each of the images to be fused loses the features of the images to be fused due to the high filtering intensity, which in turn makes the test fused image unable to accurately reflect the sealing condition of the food plastic packaging surface, thus improving the accuracy of plastic sealing detection.

[0117] Specifically, the area enclosed by the same features of the test fused image and an image to be fused is the feature similarity area between the test fused image and the image to be fused.

[0118] Optionally, when the food plastic packaging is placed under external conditions of 35℃~39℃, the pre-defined characteristic similarity area ratio is generally taken in the range of [80%, 95%].

[0119] Preferably, in the case where the food plastic-sealed packaging is placed under external conditions of 35℃~39℃, the preferred embodiment of the area ratio of the preset similar features is 90%.

[0120] Those skilled in the art will understand that [80%, 95%], 90% refers to several optional and preferred embodiments of the food plastic-sealed packaging under external conditions of 35℃~39℃. In actual application or implementation, those skilled in the art can adaptively adjust the pre-set feature similarity area ratio according to the actual application environment and application scenario.

[0121] Specifically, the filtering intensity of the visible light image is negatively correlated with the proportion of the area of ​​feature similarity.

[0122] In implementation, when the food packaging is exposed to an external temperature of 35℃~39℃, and the proportion of similar feature areas is less than 10% of the preset similar feature area proportion, the peak signal-to-noise ratio (PSNR) of the visible light image is adjusted to 0.9 times the current PSNR of the visible light image. When the proportion of similar feature areas is less than 10% of the preset similar feature area proportion, the PSNR of the visible light image is adjusted to 0.9 times the current PSNR of the visible light image for every 10% exceeding the preset proportion. For example, in one possible embodiment, if the proportion of similar feature areas is less than 20% of the preset similar feature area proportion, the PSNR of the visible light image is adjusted to 0.9 × 0.9 = 0.81 times the current PSNR of the visible light image. Here, the PSNR represents the filtering intensity.

[0123] Specifically, the feature similarity area ratio is the ratio of each feature similarity area to the area of ​​the corresponding single image to be fused.

[0124] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A food plastic wrapping detection method based on image fusion, characterized in that, The application comprises: Collecting multi-modal images of food plastic packaging, including visible light images, infrared distribution maps, and X-ray images; Determining an estimated noise type according to the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image; Determining a denoising method according to the estimated noise type, including determining a denoising area and an image segmentation edge thickness according to the coincidence degree of the heat diffusion range and the texture offset range in the infrared distribution map, Or, determining the filtering strength of the visible light image according to the offset distance after the visible light image and the infrared distribution map are compared through edge mask comparison; Respectively performing denoising preprocessing on each of the multi-modal images according to the denoising method to output corresponding to-be-fused images; Fusing each of the to-be-fused images to generate a test fusion image; Determining fusion accuracy according to the feature similarity area ratio of the test fusion image and each of the to-be-fused images; If the fusion accuracy does not meet the requirements, reducing the filtering strength of the visible light image; Filtering the visible light image according to the filtering strength to output a standard visible light image; Fusing the standard visible light image with the to-be-fused images of the infrared distribution map and the X-ray image to output a target fusion image of the food plastic packaging.

2. The image fusion based food plastic wrapping detection method according to claim 1, wherein, The texture correlation feature ratio is the ratio of the area of the region with the same infrared texture features in the infrared distribution map and the visible light texture features in the visible light image to the area of the infrared distribution map.

3. The image fusion based food plastic wrapping detection method according to claim 2, wherein, Determining an estimated noise type according to the infrared texture features of the infrared distribution map and the visible light texture features of the visible light image, including, Respectively extracting the infrared texture features in the infrared distribution map and the visible light texture features in the visible light image; Counting the area of the region with the same infrared texture features and visible light texture features; Calculating the texture correlation feature ratio according to the area and the area of the infrared distribution map; If the texture correlation feature ratio is greater than a preset second ratio, determining that the estimated noise type is a first type; If the texture correlation feature ratio is greater than a preset first ratio and less than or equal to the preset second ratio, determining that the estimated noise type is a second type; If the texture correlation feature ratio is less than or equal to the preset first ratio, determining that the estimated noise type is a standard type.

4. The image fusion based food plastic wrapping detection method according to claim 3, wherein, Adjusting the denoising area and the image segmentation edge thickness according to the coincidence degree of the heat diffusion range and the texture offset range in the infrared distribution map, including, If the estimated noise type is the first type, obtaining the heat diffusion range and the texture offset range; Calculating the coincidence degree according to the heat diffusion range and the texture offset range; Comparing the coincidence degree with a preset coincidence degree; According to the trigger condition that the coincidence degree is greater than the preset coincidence degree, adjusting the denoising area to a non-diffusion area and increasing the image segmentation edge thickness.

5. The image fusion based food plastic wrapping detection method according to claim 4, wherein, The image segmentation edge thickness and the coincidence degree have a positive correlation.

6. The image fusion based food plastic wrapping detection method according to claim 5, wherein, Determining the filtering strength of the visible light image according to the offset distance after the visible light image and the infrared distribution map are compared through edge mask comparison, including: If the estimated noise type is the second type, obtaining the offset distance; The offset distance is compared with a preset offset distance; If the offset distance is greater than the preset offset distance, the filtering intensity of the visible light image is increased.

7. The image fusion based food plastic wrapping detection method according to claim 6, wherein, The filtering intensity of the visible light image is in a positive correlation with the offset distance.

8. The image fusion based food plastic wrapping detection method according to claim 7, wherein, The fusion accuracy is determined according to a feature similar area ratio of the test fusion image and each of the to-be-fused images, comprising: A plurality of feature similar areas of the test fusion image and each of the to-be-fused images are respectively acquired; A feature similar area ratio is calculated according to a single feature similar area and an area of the corresponding to-be-fused image. If at least one of the feature similar area ratios is less than a preset feature similar area ratio, it is determined that the fusion is incorrect.

9. The image fusion based food plastic wrapping detection method according to claim 8, wherein, The filtering intensity of the visible light image is in a negative correlation with the feature similar area ratio.

10. The image fusion based food plastic wrapping detection method according to claim 9, wherein, The feature similar area ratio is a ratio of each feature similar area to an area of the corresponding single to-be-fused image.

Images