Foreign-object detection method and apparatus based on dual-frequency signal

Abstract

The present invention relates to a foreign-object detection method and apparatus based on a dual-frequency signal. The method comprises: controlling a transmitting coil assembly to transmit a first-frequency transmitted signal and a second-frequency transmitted signal at the same time, wherein the frequency of the first-frequency transmitted signal is higher than the frequency of the second-frequency transmitted signal; acquiring a first-frequency received signal corresponding to the first-frequency transmitted signal, and acquiring a second-frequency received signal corresponding to the second-frequency transmitted signal; by using one of the first-frequency received signal and the second-frequency received signal as an abscissa, and using the other one as an ordinate, performing drawing in a rectangular plane coordinate system, so as to obtain a detection signal trajectory diagram; preprocessing the detection signal trajectory diagram to obtain a closed detection graph; and determining whether the detection graph is similar to a standard graph, wherein if the detection graph is similar to the standard graph, a detection result indicates that there are no metallic foreign objects, otherwise, the detection result indicates that there are metallic foreign objects. Compared with the prior art, the present invention can solve the problem of the detection accuracy of metallic foreign objects being low in products that are extremely humid and have a high salt content.

Description

A method and apparatus for foreign object detection based on dual-frequency signals Technical Field

[0001] This invention relates to the field of food testing, and in particular to a method and apparatus for detecting foreign objects based on dual-frequency signals. Background Technology

[0002] Metal detectors rely on changes in electromagnetic fields to detect metals. Specifically, they detect metals by detecting the degree of disturbance to the electromagnetic field within the probe.

[0003] Metal detectors typically operate within the 50kHz to 1MHz range, and generally operate at a single frequency, meaning that only one electromagnetic field of that frequency exists within the probe window at any given time. Different operating frequencies are suitable for different types of items being tested. For damp items with high salt content (high conductivity), lower operating frequencies (e.g., 50-300kHz) are suitable. Conversely, for dry items with low salt content (low conductivity), higher operating frequencies (e.g., 300-1MHz) are suitable.

[0004] In addition, items that are very moist and have a very high salt content, such as sauces, braised meat, pickles, and marinated products, are difficult to detect satisfactorily even if the operating frequency is reduced. This is because the conductivity of the tested items is high enough to cover the signal generated by the electromagnetic field disturbance inside the probe caused by large-sized metal foreign objects.

[0005] Although some existing technologies have proposed multi-frequency signal detection methods, these technologies generally use signal ratios as criteria, which still cannot detect products that are very moist and have a very high salt content, such as sauces and pickles.

[0006] Summary of the Invention

[0007] The purpose of this invention is to provide a foreign object detection method and device based on dual-frequency signals. By using the dual-frequency received signals as the horizontal and vertical coordinates respectively, the presence of foreign objects is determined based on whether the closed pattern enclosed by the signals is similar to a standard pattern. This can solve the problem of low detection accuracy of metallic foreign objects in products that are very humid or have a very high salt content.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] A foreign object detection method based on dual-frequency signals, comprising:

[0010] The control coil assembly simultaneously emits a first frequency transmission signal and a second frequency transmission signal, wherein the frequency of the first frequency transmission signal is higher than the frequency of the second frequency transmission signal.

[0011] Acquire a first frequency received signal corresponding to a first frequency transmitted signal, and acquire a second frequency received signal corresponding to a second frequency transmitted signal;

[0012] Using one of the first frequency received signal and the second frequency received signal as the abscissa and the other as the ordinate, a detection signal trajectory diagram is drawn in a Cartesian coordinate system.

[0013] The detection signal trajectory diagram is preprocessed to obtain a closed detection pattern;

[0014] Determine whether the detected image and the standard image are similar. If they are similar, the detection result is that there is no metal foreign object; otherwise, it is that there is a metal foreign object.

[0015] The determination of whether the detected image and the standard image are similar specifically involves:

[0016] The detected image is input into the trained similarity detection model to obtain the similarity detection result.

[0017] The determination of whether the detected image and the standard image are similar specifically includes:

[0018] Extract the feature vector of the detected image;

[0019] The extracted feature vectors are input into the trained similarity detection model to obtain similarity detection results.

[0020] The extraction of the feature vector of the detected image includes:

[0021] Obtain the centroid of the detected image, and calculate the distance from each point on the detected image to the centroid.

[0022] Select the point with the largest distance as the starting point, take the vector from the center of gravity to the starting point as the starting detection vector, take the center of gravity as the rotation center, and rotate the starting detection vector along the first direction according to the pre-configured angular step size to obtain multiple second detection vectors;

[0023] All intersections of the second detection vector and the detection pattern are taken as sampling points, and the sampling points are sorted in ascending order according to the rotation angle of the corresponding second detection vector relative to the initial detection vector along the first direction.

[0024] Based on the distance from each sampling point to the centroid, and combined with the sorting result of the sampling points, a sequence is constructed, and the distance from the centroid to the starting point is added to both the beginning and end of the sequence to obtain the feature vector corresponding to the detected image.

[0025] The similarity detection model is a classification model, and the model output is a binary variable, representing similarity or dissimilarity.

[0026] A foreign object detection device based on dual-frequency signals includes a memory, a processor, and a program stored in the memory. When the processor executes the program, it performs the following steps:

[0027] The control coil assembly simultaneously emits a first frequency transmission signal and a second frequency transmission signal, wherein the frequency of the first frequency transmission signal is higher than the frequency of the second frequency transmission signal.

[0028] Acquire a first frequency received signal corresponding to a first frequency transmitted signal, and acquire a second frequency received signal corresponding to a second frequency transmitted signal;

[0029] Using one of the first frequency received signal and the second frequency received signal as the abscissa and the other as the ordinate, a detection signal trajectory diagram is drawn in a Cartesian coordinate system.

[0030] The detection signal trajectory diagram is preprocessed to obtain a closed detection pattern;

[0031] Determine whether the detected image and the standard image are similar. If they are similar, the detection result is that there is no metal foreign object; otherwise, it is that there is a metal foreign object.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. By using the dual-frequency received signals as the horizontal and vertical coordinates respectively, and then judging whether there are foreign objects based on whether the closed shape enclosed by the signals is similar to the standard shape, the problem of low detection accuracy of metal foreign objects in products with very humid conditions and very high salt content can be solved.

[0034] 2. Using machine learning models to classify similarity can simplify program design when dealing with large samples.

[0035] 3. Using distance sequences as feature vectors as input to the model can greatly improve the convergence speed and classification accuracy with small sample sizes. Attached Figure Description

[0036] Figure 1 is a schematic diagram of the main steps of the method of the present invention;

[0037] Figure 2 is a schematic diagram of the detection signal trajectory. Detailed Implementation

[0038] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0039] A foreign object detection method based on dual-frequency signals, as shown in Figure 1, includes:

[0040] The control coil assembly simultaneously emits a first frequency transmission signal and a second frequency transmission signal, wherein the frequency of the first frequency transmission signal is higher than the frequency of the second frequency transmission signal;

[0041] Acquire a first frequency received signal corresponding to a first frequency transmitted signal, and acquire a second frequency received signal corresponding to a second frequency transmitted signal;

[0042] Using one of the first frequency received signal and the second frequency received signal as the abscissa and the other as the ordinate, a detection signal trajectory diagram is drawn in a Cartesian coordinate system.

[0043] The detection signal trajectory diagram is preprocessed to obtain a closed detection pattern;

[0044] Determine whether the detected image and the standard image are similar. If they are similar, the detection result is that there is no metal foreign object; otherwise, it is that there is a metal foreign object.

[0045] By using dual-frequency received signals as the horizontal and vertical coordinates respectively, and then determining whether there are foreign objects based on whether the closed pattern enclosed by the signals is similar to a standard pattern, the problem of low detection accuracy of metallic foreign objects in products with very humid conditions and very high salt content can be solved.

[0046] The second frequency transmission signal has a lower frequency and is considered a low-frequency signal, while the first frequency transmission signal has a higher frequency and is considered a high-frequency signal, as shown in Figure 2. In this embodiment, the low-frequency signal is used as the horizontal axis and the high-frequency signal is used as the vertical axis. The low-frequency signal and the high-frequency signal are synchronized in time to obtain the detection signal trajectory diagram. Generally, the detection signal trajectory diagram will appear in two opposite quadrants, such as the first quadrant and the third quadrant, and each quadrant is closed.

[0047] In other embodiments, the opposite can be used, with the low-frequency signal as the vertical axis and the high-frequency signal as the horizontal axis.

[0048] For cases where the signal trajectory diagram itself is closed, the pattern in that quadrant can be directly used as the detection pattern. For example, in Figure 2, the part in the first quadrant can be directly used as the detection pattern. Of course, in some other embodiments, the patterns in the first quadrant and the third quadrant can be used as two detection patterns respectively. In this case, it is necessary to determine whether each detection pattern is similar to its corresponding standard pattern. Only when the judgment results of the two detection patterns are similar can it be considered that there is no foreign object.

[0049] When the signal estimation map is not closed, image inpainting is required. In some cases, re-detection is necessary, while in other cases, some image inpainting models can be used to perform inpainting using interpolation methods.

[0050] In addition, for some signal trajectory maps that are entangled, it is necessary to extract the outer contour of the signal trajectory map as the detection pattern.

[0051] In one embodiment, determining whether the detected image and the standard image are similar specifically involves:

[0052] The detected image is input into the trained similarity detection model to obtain the similarity detection result.

[0053] The similarity detection model is a classification model, and the model output is a binary variable, representing whether the model is similar or dissimilar.

[0054] Using machine learning models to classify similarity can simplify program design when dealing with large samples.

[0055] Of course, in another type of embodiment, determining whether the detected image and the standard image are similar specifically includes:

[0056] Extract the feature vector of the detected image;

[0057] The extracted feature vectors are input into the trained similarity detection model to obtain similarity detection results.

[0058] Extract the feature vector of the detected image, including:

[0059] Obtain the centroid of the detected image, and calculate the distance from each point on the detected image to the centroid.

[0060] Select the point with the largest distance as the starting point, take the vector from the centroid to the starting point as the initial detection vector, take the centroid as the rotation center, and rotate the initial detection vector along the first direction according to the pre-configured angular step size to obtain multiple second detection vectors;

[0061] All intersections of the second detection vector and the detection pattern are taken as sampling points, and the sampling points are sorted in ascending order according to the rotation angle of the corresponding second detection vector relative to the initial detection vector along the first direction.

[0062] Based on the distance from each sampling point to the centroid, and combined with the sorting result of the sampling points, a sequence is constructed, and the distance from the centroid to the starting point is added to both the beginning and end of the sequence to obtain the feature vector corresponding to the detected image.

[0063] Therefore, using distance sequences as feature vectors as input to the model can greatly improve the convergence speed and classification accuracy with small sample sizes.

[0064] In this embodiment, the angle step size is 2 degrees, the first direction is clockwise, and the length of the resulting feature vector is 181.

[0065] Furthermore, in such embodiments, the centroid can be determined using existing centroid extraction algorithms.

[0066] In such embodiments, the similarity detection model is also a classification model, and the model output is a binary variable, representing similarity or dissimilarity.

[0067] For the classification model, a conventional neural network model can be used. In addition, in some embodiments of this type, a sequence consisting of the ratios of each element in the feature vector of the detected image and each element in the feature vector of the standard image can be calculated. Then, the variance of all elements in the sequence is calculated, and it is determined whether the variance is greater than a pre-configured threshold. If it is, they are considered dissimilar; otherwise, they are considered similar.

[0068] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A foreign matter detection method based on a dual-frequency signal, characterized by, The method comprises the following steps: controlling the transmitting coil assembly to simultaneously emit a first frequency transmitting signal and a second frequency transmitting signal, wherein the frequency of the first frequency transmitting signal is higher than the frequency of the second frequency transmitting signal; obtaining a first frequency receiving signal corresponding to the first frequency transmitting signal and a second frequency receiving signal corresponding to the second frequency transmitting signal; plotting a detection signal trajectory graph in a plane rectangular coordinate system by taking one of the first frequency receiving signal and the second frequency receiving signal as the abscissa and the other as the ordinate; preprocessing the detection signal trajectory graph to obtain a closed detection graph; judging whether the detection graph and a standard graph are similar, and if yes, the detection result is that no metal foreign matter is contained, otherwise, the detection result is that metal foreign matter is contained.

2. The foreign matter detection method based on a dual-frequency signal according to claim 1, characterized by, The judging whether the detection graph and the standard graph are similar specifically comprises: inputting the detection graph into a trained similarity detection model to obtain a similarity detection result.

3. The foreign matter detection method based on a dual-frequency signal according to claim 1, characterized by, The judging whether the detection graph and the standard graph are similar specifically comprises: extracting a feature vector of the detection graph; inputting the extracted feature vector into the trained similarity detection model to obtain a similarity detection result.

4. The foreign matter detection method based on a dual-frequency signal according to claim 3, characterized by, The extracting the feature vector of the detection graph comprises: obtaining a barycenter of the detection graph, and calculating distances from each point on the detection graph to the barycenter respectively; selecting a point with the largest distance as a starting point, taking a vector from the barycenter to the starting point as a starting detection vector, taking the barycenter as a rotation center, and rotating the starting detection vector along a first direction according to a preconfigured angle step to obtain a plurality of second detection vectors; taking intersection points of all the second detection vectors and the detection graph as sampling points, and sorting the sampling points according to rotation angles of the corresponding second detection vectors along the first direction compared with the starting detection vector from small to large; based on distances of the sampling points to the barycenter, combining a result of the sorting of the sampling points, constructing a sequence, and adding distances from the barycenter to the starting point at the head and the tail of the sequence to obtain a feature vector corresponding to the detection graph.

5. The foreign matter detection method based on a dual-frequency signal according to claim 2 or 3, characterized by, The similarity detection model is a classification model, and the model output is a binary variable, which respectively represents similarity or dissimilarity.

6. A foreign matter detection device based on a dual-frequency signal, comprising a memory, a processor, and a program stored in the memory, characterized by, The processor implements the following steps when executing the program: controlling the transmitting coil assembly to simultaneously emit a first frequency transmitting signal and a second frequency transmitting signal, wherein the frequency of the first frequency transmitting signal is higher than the frequency of the second frequency transmitting signal; obtaining a first frequency receiving signal corresponding to the first frequency transmitting signal and a second frequency receiving signal corresponding to the second frequency transmitting signal; plotting a detection signal trajectory graph in a plane rectangular coordinate system by taking one of the first frequency receiving signal and the second frequency receiving signal as the abscissa and the other as the ordinate; preprocessing the detection signal trajectory graph to obtain a closed detection graph; judging whether the detection graph and a standard graph are similar, and if yes, the detection result is that no metal foreign matter is contained, otherwise, the detection result is that metal foreign matter is contained.

7. The foreign object detection apparatus based on dual frequency signals according to claim 6, characterized in that, The judging whether the detection graph and the standard graph are similar specifically comprises: inputting the detection graph into a trained similarity detection model to obtain a similarity detection result.

8. The foreign object detection apparatus based on dual frequency signals according to claim 6, characterized in that, The judging whether the detection graph and the standard graph are similar specifically comprises: extracting a feature vector of the detection graph; inputting the extracted feature vector into the trained similarity detection model to obtain a similarity detection result. Input the extracted feature vector into the trained similarity detection model to obtain a similarity detection result.

9. The foreign object detection apparatus based on dual frequency signals according to claim 8, characterized in that, The feature vector of the detection graph is extracted, including: Obtaining the center of gravity of the detection graph, and respectively calculating the distances from each point on the detection graph to the center of gravity; Selecting a point with the largest distance as a starting point, taking a vector from the center of gravity to the starting point as a starting detection vector, taking the center of gravity as a rotation center, and rotating the starting detection vector in a first direction according to a preconfigured angle step to obtain a plurality of second detection vectors; Taking the intersection points of all the second detection vectors and the detection graph as sampling points, and sorting the sampling points according to the rotation angles of the corresponding second detection vectors along the first direction relative to the starting detection vector from small to large; Based on the distances of the sampling points to the center of gravity, and in combination with the sorting result of the sampling points, a sequence is constructed, and the distance from the center of gravity to the starting point is added at the head and tail of the sequence to obtain a feature vector corresponding to the detection graph.

10. The foreign object detection apparatus based on dual frequency signals according to claim 7 or 8, characterized in that, The similarity detection model is a classification model, and the model output is a binary variable, which respectively represents similarity or dissimilarity.

Images