An intelligent detection method and system for a metal detector

Abstract

The present invention relates to an intelligent detection method and system for a metal detector. The method includes: preparing pre-recorded images of products, including at least images of qualified products; obtaining feature vectors of the products at multiple frequencies; obtaining feature images based on the feature vectors; after the metal detector processes the feature images, comparing them with the pre-recorded images, and outputting the detection determination results of the products. Compared with the prior art, the present invention has the advantages of realizing the detection of non-metallic foreign objects and small-sized metals, high detection accuracy, and realizing intelligent detection based on feature images, etc.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, and in particular to an intelligent detection method and system for a metal detector. Background Art

[0002] A metal detector is an instrument that detects foreign metal objects in the object being tested. The metal detector detects foreign objects (especially metal objects) based on the changes in the electromagnetic field in the probe window. The products passing through the probe are also conductive due to the moisture and salt contained in them, which will significantly change the electromagnetic field in the probe (the so-called product effect). The signal generated can sometimes drown out the signal of foreign metal objects. An important task of a metal detector is to distinguish the signal of foreign metal objects from the effect caused by the product and remove the products containing foreign objects.

[0003] Metal detectors can currently only detect metal foreign bodies, but there are far more foreign bodies mixed into actual products than just metal. Common non-metallic foreign bodies include stones, glass, soil, paper, plastic products, etc., which cannot be detected by current solutions. However, these foreign bodies will also bring risks after being mixed into products.

[0004] At present, the commonly used means for detecting non-metallic foreign bodies are to use X-ray machines, visual inspection equipment, etc., but they also have their own limitations. For example, X-ray machines are limited to some high-density foreign bodies (such as stones, glass, etc.), and it is difficult to work on low-density foreign bodies. Visual inspection equipment can detect various foreign bodies on the surface of products, but it is still unable to detect foreign bodies inside the products. Moreover, the cost of these spectral inspection equipment is generally high at this stage.

[0005] How to improve detection accuracy has become a technical problem that needs to be solved. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent detection method and system for a metal detector in order to overcome the defects of the above-mentioned prior art.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] According to one aspect of the present invention, there is provided an intelligent detection method for a metal detector, the method comprising:

[0009] 1) Prepare pre-recorded images of products, including at least images of qualified products;

[0010] 2) Obtaining feature vectors of multiple frequencies of the product;

[0011] 3) obtaining a feature image based on the feature vector in step 2);

[0012] 4) After the metal detector processes the characteristic image, it compares it with the pre-recorded image in step 1) and outputs the detection determination result of the product.

[0013] Preferably, the signals for the characteristic vectors of the multiple frequencies are obtained by adjusting the phase of the carrier signal or the phase of the transmitted signal;

[0014] The characteristic vector formed by the two signals Changing with time t is expressed as:

[0015]

[0016] where and are two unit vectors in the two-dimensional coordinate system, and DT1(t) and DT2(t) are the information representing the change within a period of time of the disturbance generated by the probe.

[0017] Preferably, the product characteristic image is a binary closed curve image formed based on the vector end curve of the characteristic vector.

[0018] Preferably, the results after performing correlation operations on the characteristic images of multiple frequencies form a new characteristic image, where the correlation operations include difference operation, addition operation, multiplication operation, and fusion operation.

[0019] Preferably, the processing of the characteristic image by the metal detector includes image preprocessing, image feature extraction, and target recognition detection; the metal detector compares the characteristic image with the image of a qualified product and outputs the detection determination result of the product according to the similarity.

[0020] More preferably, the detection and judgment results of the metal detector include the detection of qualified products, unqualified products containing non-metal foreign objects, unqualified products containing metal foreign objects, missing parts, and multiple parts.

[0021] Preferably, the pre-recorded image also includes images of unqualified products containing non-metal foreign objects and images of unqualified products containing metal foreign objects.

[0022] Preferably, the intelligent detection of the metal detector is used in parallel with the traditional threshold detection.

[0023] According to another aspect of the present invention, there is provided an intelligent detection system for a metal detector, which system includes an actuator, and a probe, a balance signal generation module, a signal processing circuit, and an AD sampling module connected in sequence, and the system further includes a signal processing circuit, an intelligent detection module, and a judgment module;

[0024] The intelligent detection module generates a feature image based on the feature vectors output by the AD sampling module, and processes the feature image to distinguish the differences from the feature image of the standard qualified product, where the processing includes image preprocessing, image feature extraction, and target recognition and detection; the judgment module determines the output result of the intelligent detection module, outputs the detection judgment result, and then notifies the execution mechanism to make corresponding response actions.

[0025] Preferably, the system further includes a threshold detection module parallel to the intelligent detection module. The threshold detection module uses a preset threshold as the judgment criterion. If the sampling signal exceeds the threshold, it is determined that a foreign object has been detected.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1) The present invention obtains a feature image based on the feature vectors of multiple frequencies of the product, thereby obtaining more feature information, and uses image processing methods for processing, improving the detection accuracy.

[0028] 2) When the pre-recorded image of the present invention contains an image of a non-metallic foreign object, the non-metallic foreign object can be detected from the product, solving the detection problem of non-metallic abnormalities that could not be solved before.

[0029] 3) The present invention is parallel to the traditional threshold detection, further improving the detection accuracy.

[0030] 4) Based on the differences in the feature images, the metal detector can use image features and image processing algorithms to detect smaller metal foreign objects, improving the detection accuracy and adaptability, thereby realizing the intelligent detection of various foreign objects in the product.

[0031] 5) The present invention broadens the application scope of detection and can be used for the detection of foreign objects in products, the detection of missing parts and multiple parts in products, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the acquisition process of the balance signal in the present invention;

[0033] Figure 2 It is a schematic diagram of the processing method flow of the balance signal in the present invention;

[0034] Figure 3 It is a schematic diagram of the feature vector in the present invention;

[0035] Figure 4 It is a schematic diagram of the feature image formed by the tip curve of the feature vector in the present invention;

[0036] Figure 5 It is a schematic diagram of a part of the feature image exceeding the threshold rectangular window in the present invention;

[0037] Figure 6 Schematic diagram of all feature images within the threshold rectangular window in the present invention;

[0038] Figure 7 Schematic diagram of the feature image at the first frequency in the present invention;

[0039] Figure 8 Schematic diagram of the feature image at the second frequency in the present invention;

[0040] Figure 9 Schematic diagram of the feature image of a qualified product in the present invention;

[0041] Figure 10 Schematic diagram of the feature image including non - metal in the present invention;

[0042] Figure 11 Schematic diagram of the feature image including metal in the present invention;

[0043] Figure 12 Schematic diagram of the feature image of a missing part in the present invention;

[0044] Figure 13 Schematic diagram of the feature image of multiple parts in the present invention;

[0045] Figure 14 Schematic diagram of the parallel detection system for threshold detection and intelligent detection in the present invention;

[0046] Figure 15 Schematic diagram of the intelligent detection process in the present invention. Detailed implementation manners

[0047] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] This embodiment relates to an intelligent detection method for a metal detector.

[0049] The metal detector realizes the detection of metal foreign objects by processing the balance signal. Generally, the balance signal can be obtained in the following manner. The transmitted signal is amplified by the power amplifier circuit and then sent into the transmitting coil through the transmitting transformer, generating an electromagnetic field of a certain intensity in the probe. The receiving coil senses the change of this electromagnetic field, and its output enters the receiving amplifier circuit through the receiving transformer and outputs the balance signal, as Figure 1 shown.

[0050] For balanced signals, common processing methods such as Figure 2 (including but not limited to), the balanced signal and the quadrature carrier signal generated by the carrier signal generator are input into a multiplier. The output signal of the multiplier is filtered and amplified to obtain the DT1 and DT2 signals for sampling, which are then sent to the CPU after AD sampling. Since the balanced signal and the quadrature carrier signal generated by the carrier signal generator are not from the same source, the phase of the balanced signal can be changed by adjusting the phase of the transmitted signal (while the phase of the carrier signal remains fixed), or conversely, the phase of the transmitted signal is fixed and the phase of the carrier signal is adjusted. In the CPU, DT1 and DT2 are processed. Generally, a preset threshold is used as the judgment criterion. If DT1 or DT2 exceeds the threshold, it is determined that a foreign object has been detected.

[0051] The essence of the above scheme is to obtain the quantization information (DT1(t) and DT2(t)) of two characteristics of the disturbance generated when the product passes through the probe at a certain moment t by certain means, and then process this quantization information according to certain judgment rules (exceeding the thresholds THS1 and THS2) to obtain the detection result.

[0052] Obviously, this processing method loses the change information of the two characteristics of the disturbance generated when the product passes through the probe within a period of time, that is, (DT1(t) and DT2(t)).

[0053] The characteristics (DT1(t) and DT2(t)) of a specific product (when passing through the probe) jointly reflect the characteristics of this product and can be represented by a feature vector as Figure 3 shown. According to the method of obtaining DT1 and DT2 described above, DT1 and DT2 are relatively independent. Therefore, the change of this feature vector with t can be expressed as:

[0054]

[0055] where and are two unit vectors in the two-dimensional coordinate system.

[0056] The above-mentioned processing method of the detection signal (exceeding the threshold) judges through a rectangular window to judge the characteristic image M(t) formed by the vector end curve of in the above two-dimensional coordinate system within a period of time, as Figure 4 shown. If there is a part of the characteristic image outside the window, it is considered that a foreign object has been found, as Figure 5 and Figure 6That is to say, this processing method only looks at the relationship between the feature image and the window boundary, while losing the information contained in the image itself. The information contained in the image itself has more value and can reflect more (more complete) feature information of the product. This is also the advantage of the present invention. More complete feature information means better detection effect.

[0057] Actually, the two orthogonal signals generated by the carrier signal generator in the above metal detector circuit scheme can also generate two signals with a phase difference of 120°, or even add another path to form three signals with a phase difference of 120° between each pair. For example:

[0058]

[0059] Among them, and are the three unit vectors of the two-dimensional coordinate system, and z(t) is the third signal.

[0060] Theoretically, the components of the vector can be increased infinitely to obtain more feature information. Generally speaking, however, in order to facilitate the application of mature image processing algorithms, it is a good choice to use a two-dimensional coordinate system to obtain images.

[0061] In addition, by using the dual-frequency technology, the product features at another frequency can be obtained and then the image M′(t) formed by its vector end curve is obtained. Thus, we obtain two images M(t) and M′(t) that can reflect the product features, such as Figure 7 and Figure 8 , and various image processing methods can be conveniently used to process them, and then some non-metallic foreign objects and smaller-sized metal foreign objects can be distinguished from the product. Of course, this can be extended to more frequencies to obtain more images (M″(t), M″′(t), etc.).

[0062] The method for the metal detector to obtain the feature image M(t) of the object passing through the probe is introduced above. Below, we will discuss it taking M(t) as an example.

[0063] Due to the symmetry of the receiving coil of the probe, the image corresponding to the object passing through the probe is also symmetric (with appropriate circuit gain). For convenience, we only look at half of the content of M(t) below.

[0064] Such as Figure 9 , Figure 10 and Figure 11 , which are the qualified product M(t) without foreign objects, the unqualified product M1(t) containing non-metallic foreign objects (stones), and the unqualified product M2(t) containing metal foreign objects, respectively.

[0065] Prepare pre-recorded images of the product, including at least images of qualified products;

[0066] Pre-record images of qualified products for detection and determination by the metal detector. In addition, the pre-recorded images also include images of non-conforming products containing non-metallic foreign objects and images of non-conforming products containing metallic foreign objects, which are used to enhance the further refined determination of the detection results.

[0067] In addition, for application scenarios with requirements for detecting missing parts or multiple parts, such as detecting whether one is missing in a box of 12-pack cakes, or detecting whether one bottle is missing in a case of mineral water, or whether there is one more or one less in a bag of 20 chicken legs, the conventional method in the past was to use a weight checker to judge by weight. The solution in the present invention can achieve detection of missing parts (M3(t)) and multiple parts (M4(t)), as shown respectively in Figure 12 and Figure 13 .

[0068] If judged purely by the threshold detection method, then there may be a situation where there are missing parts (or multiple parts) but there are foreign objects, which may cause the signal to still be within the threshold range. By combining the method of intelligent image processing, this situation can be effectively avoided because the image features of missing parts / multiple parts are significantly different from the image features of containing foreign objects.

[0069] The above-mentioned M(t), M1(t), M2(t), M3(t), M4(t) cover most of the product feature images. Then, next, as long as the image information is input into the intelligent detection module and processed using various image algorithms to distinguish the differences from the standard qualified product feature images, various situations can be identified, thereby achieving the determination of the final result and notifying the actuator to make corresponding response actions. This detection can be parallel to the traditional threshold detection method, and the two are not contradictory, as shown in Figure 14 .

[0070] As mentioned above: The product feature images M(t) and M ′ (t) (and even the third and fourth frequencies) at different frequencies contain more information. Therefore, they can be respectively sent into the intelligent detection module for parallel processing. Further, even a difference operation can be performed on M(t) and M′(t), and the result M - (t) obtained from the difference operation can also be regarded as a kind of product feature image. Of course, all these operations require more computing resources, and the detection effect and computing efficiency should be balanced when actually applied. In addition, in addition to the difference operation, other relevant operations can be performed, and the relevant operations include but are not limited to addition operation, multiplication operation, and fusion operation.

[0071] Generally, as shown in Figure 15, after the product image is sent into the intelligent detection module, it will go through several links including image preprocessing, image feature extraction, and target detection / recognition, and then output a judgment result. In these several links, there are a large number of algorithms to choose from for each link, and we need to select a suitable algorithm according to the image features.

[0072] Since M(t) is a binary closed curve image, some suitable algorithms include but are not limited to: image stretching, gradient sharpening, Roberts operator, image reconstruction, geometric transformation, image cropping, feature point detection, feature point matching, CNN deep learning, and so on.

[0073] The above algorithms can be easily obtained through tools such as MATLAB and OPENCV and applied to the device. Especially in the feature extraction and target recognition links, tools such as OPENAI can also be used to improve the accuracy and adaptability of processing, and truly achieve intelligent detection.

[0074] This embodiment also relates to an intelligent detection system for a metal detector, which system includes a probe, a balance signal generation module, a signal processing circuit, an AD sampling module, a CPU module, a judgment module, and an actuator connected in sequence. The CPU module includes an intelligent detection module and a judgment module.

[0075] The intelligent detection module generates a feature image based on the feature vector output by the AD sampling module, and processes the feature image to distinguish the differences from the standard qualified product image, where the processing includes image preprocessing, image feature extraction, and target recognition detection.

[0076] The judgment module determines the output result of the intelligent detection module, outputs a detection judgment result, and then notifies the actuator to make corresponding response actions.

[0077] The system also includes a threshold detection module parallel to the intelligent detection module. The threshold detection module uses a preset threshold as the judgment standard. If the sampling signal exceeds the threshold, it is determined that a foreign object has been detected. The results of both detection modules are sent to the judgment module for detection result determination.

[0078] As described above, only the specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art in the technical field disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An intelligent detection method for a metal detector, characterized in that The method includes: 1) Prepare pre-recorded images of the product, including at least images of qualified products; 2) Obtain feature vectors of the product at multiple frequencies; 3) Obtain a feature image based on the feature vectors in step 2); 4) After the metal detector processes the feature image, compare it with the pre-recorded image in step 1), and output the detection determination result of the product, where the detection determination result includes qualified products, non-conforming products containing non-metallic foreign objects, non-conforming products containing metallic foreign objects, products with missing parts, and products with multiple parts; The feature vector formed by the two signals which changes with time t is expressed as: ， Among them, are two unit vectors of a two-dimensional coordinate system, and represent the variation information of the disturbance generated by the probe over a period of time; The signals for the feature vectors at multiple frequencies are obtained by adjusting the phase of the carrier signal or the phase of the transmitted signal; The feature image is a binary closed curve image formed by the tip curves of the feature vectors; The results of performing correlation operations on the feature images at multiple frequencies form a new feature image, where the correlation operations include subtraction operation, addition operation, multiplication operation, and fusion operation.

2. The intelligent detection method of a metal detector according to claim 1, characterized in that, The processing of the feature image by the metal detector includes image preprocessing, image feature extraction, and target recognition and detection; the metal detector compares the feature image with the image of the qualified product and outputs the detection determination result of the product according to the similarity.

3. The intelligent detection method of a metal detector according to claim 1, characterized in that The pre-recorded image also includes images of non-conforming products containing non-metallic foreign objects and non-conforming products containing metallic foreign objects.

4. An intelligent detection method for a metal detector according to claim 1, characterized in that, The intelligent detection of the metal detector is used in parallel with the traditional threshold detection.

5. A system for an intelligent detection method using the metal detector described in claim 1, the system comprising an actuator, and a probe, a balance signal generation module, a signal processing circuit, and an AD sampling module connected in sequence, characterized in that, The system further includes an intelligent detection module and a judgment module; The intelligent detection module generates a feature image based on the feature vectors output by the AD sampling module, processes the feature image, and distinguishes the differences from the feature image of the standard qualified product, where the processing includes image preprocessing, image feature extraction, and target recognition and detection; the judgment module determines the output result of the intelligent detection module, outputs the detection determination result, and then notifies the actuator to make corresponding response actions.

6. The system according to claim 5, wherein The system further includes a threshold detection module parallel to the intelligent detection module. The threshold detection module uses a preset threshold as the judgment criterion. If the sampling signal exceeds the threshold, it is determined that a foreign object has been detected.

Images