WO2022016328A1 - 金属异物检测方法、装置及终端设备 - Google Patents

金属异物检测方法、装置及终端设备 Download PDF

Info

Publication number
WO2022016328A1
WO2022016328A1 PCT/CN2020/103066 CN2020103066W WO2022016328A1 WO 2022016328 A1 WO2022016328 A1 WO 2022016328A1 CN 2020103066 W CN2020103066 W CN 2020103066W WO 2022016328 A1 WO2022016328 A1 WO 2022016328A1
Authority
WO
WIPO (PCT)
Prior art keywords
hyperspectral
metal
target area
foreign body
reflectance
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/103066
Other languages
English (en)
French (fr)
Inventor
田勇
李正
田劲东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen University
Original Assignee
Shenzhen University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen University filed Critical Shenzhen University
Priority to PCT/CN2020/103066 priority Critical patent/WO2022016328A1/zh
Priority to CN202080001313.2A priority patent/CN113260882B/zh
Publication of WO2022016328A1 publication Critical patent/WO2022016328A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V8/00—Prospecting or detecting by optical means
    • G01V8/10—Detecting, e.g. by using light barriers
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1765—Method using an image detector and processing of image signal

Definitions

  • the present application relates to the technical field of object detection, and in particular, to a metal foreign object detection method, device and terminal equipment.
  • IPT Inductive Power Transfer
  • the IPT system includes a primary side power transmitting device and a secondary side power receiving device.
  • a primary side power transmitting device When applied to fields such as electric vehicle wireless charging, there will be a large space area (energy transmission area) between the primary side power transmitting device and the secondary side power receiving device. ), if there are metal foreign objects (such as metal fragments, coins, cans, etc.) entering the space area, it may hinder the energy transmission or reduce the energy transmission efficiency, and even cause a combustion accident due to the heating of the metal foreign objects.
  • metal foreign objects such as metal fragments, coins, cans, etc.
  • method one by adding a piezoelectric substrate in the transmitting coil, and exciting surface acoustic waves to detect temperature changes, thereby finding metal foreign objects; method two; , Set up several detection coils on the primary coil and/or the secondary coil, and determine whether there is metal foreign matter on the primary coil and/or the secondary coil by measuring and monitoring the current and voltage changes on the detection coil.
  • method 1 has shortcomings: when the size of the metal foreign body is too small, the oscillation frequency of the surface acoustic wave in the local area of the piezoelectric substrate changes very little, and the metal foreign body cannot be accurately detected by the passing method, and the detection accuracy is low; The magnetic field will affect the surface acoustic wave, making it impossible to set a uniform temperature threshold to determine whether there is a metal foreign body in each area.
  • the second method also has shortcomings: when the size of the metal foreign body is small, the detection coil cannot accurately detect it; when there is no metal foreign body, because the magnetic flux of each detection coil is inconsistent, and the output voltage value of each coil is different, there is no way to determine. Unified threshold, threshold parameter settings are many and difficult to obtain.
  • One of the purposes of the embodiments of the present application is to provide a method, device and terminal equipment for detecting metal foreign objects, aiming at solving the problems of inability to detect small-sized metal foreign objects and complicated adjustment.
  • a metal foreign body detection method including:
  • the target area is a charging area of a wireless charging device
  • an alarm signal is generated and sent to the wireless charging device.
  • the acquiring the hyperspectral reflectance curve of each pixel in the hyperspectral image includes:
  • a hyperspectral reflectance curve of each pixel point in the hyperspectral image is obtained according to the hyperspectral reflectance image.
  • the inputting all the hyperspectral reflectance curves into a metal detection model to detect properties of objects located in the target area includes:
  • the feature data point set is input into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • the feature data point set obtained by mapping all the hyperspectral reflectance curves to a high-dimensional feature space includes:
  • the inputting the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area includes:
  • the feature data point set is input into the classification function in the metal detection model to perform attribute detection on the feature data point set, and the attribute detection result of the object located in the target area is obtained.
  • the method before inputting all the hyperspectral reflectance curves into a metal detection model to detect properties of objects located in the target area, the method further includes:
  • the training data includes hyperspectral reflectance information corresponding to several types of foreign body samples;
  • the several types of foreign body samples include metallic materials and non-metallic materials;
  • the classification function is generated according to the support vector and the confidence level corresponding to the support vector.
  • the method before acquiring the training data, the method further includes:
  • the training data is generated by annotating the hyperspectral reflectance information of various types of foreign matter samples.
  • the determining, according to the sample feature data corresponding to each of the two types of foreign body samples, a separation hyperplane between each of the two types of the foreign body samples includes:
  • the separation hyperplane includes a weight coefficient and a deviation coefficient.
  • the method further includes:
  • the training data is input into the metal detection model for iterative training, so that the weight coefficient and the deviation coefficient satisfy a preset constraint condition.
  • a metal foreign body detection device including:
  • a hyperspectral image acquisition module for acquiring hyperspectral images of a target area; the target area is a charging area of a wireless charging device;
  • a reflectance curve acquisition module configured to acquire the hyperspectral reflectance curve of each pixel in the hyperspectral image
  • a detection module configured to input all the hyperspectral reflectance curves into a metal detection model to detect the properties of objects located in the target area;
  • the alarm module is configured to generate an alarm signal and send it to the wireless charging device if the attribute detection result of any of the objects is metal.
  • the reflectance curve acquisition module includes:
  • an information acquisition unit configured to acquire the ambient light intensity of the target area and the dark current of the photographing device used for photographing the hyperspectral image
  • a correction unit configured to correct the hyperspectral image according to the ambient light intensity and the dark current to obtain a hyperspectral reflectance image
  • a reflectance curve obtaining unit configured to obtain a hyperspectral reflectance curve of each pixel point in the hyperspectral image according to the hyperspectral reflectance image.
  • the detection module includes:
  • a feature data point set calculation unit used to map all the hyperspectral reflectance curves to a high-dimensional feature space to obtain a feature data point set
  • a detection unit configured to input the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • a terminal device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the first aspect when executing the computer program Method for the detection of foreign bodies in metals.
  • the beneficial effects of the metal foreign object detection method, device and terminal device are: collecting a hyperspectral image of a target area; the target area is a charging area of a wireless charging device; by collecting the hyperspectral image of the target area Properties of objects located within the target area are detected. Obtain the hyperspectral reflectance curve of each pixel in the hyperspectral image; since the ambient light in the target area has an influence on the imaging of the hyperspectral image, the hyperspectral reflectance curve is obtained after eliminating the ambient light in the hyperspectral image, thereby improving the The detection accuracy of foreign metal objects.
  • the device realizes attribute detection of foreign objects located in the target area, that is, the charging area of the wireless charging device. Because the hyperspectral image contains rich spectral information, it can distinguish the physical properties of different materials. According to the hyperspectral image of the target area, the material detection of foreign objects in the target area is not limited by the size of metal foreign objects. The problem of metal foreign objects improves the detection accuracy of metal foreign objects; and there is no need to set different thresholds according to different environments.
  • FIG. 1 is a schematic flowchart of a metal foreign body detection method provided in Embodiment 1 of the present application;
  • FIG. 2 is a schematic flowchart of a training process of a metal detection model provided in Embodiment 2 of the present application;
  • FIG. 3 is a schematic structural diagram of a metal foreign body detection device provided in Embodiment 3 of the present application.
  • FIG. 4 is a schematic structural diagram of a terminal device provided in Embodiment 4 of the present application.
  • FIG. 1 it is a schematic flowchart of the metal foreign body detection method provided in Embodiment 1 of the present application.
  • This embodiment can be applied to the application scenario of detecting foreign metal objects located in the charging area of the wireless charging device.
  • the method can be performed by a metal foreign object detection device, and the device can be a terminal device, a smart terminal, a tablet or a PC, etc.; in this application
  • the metal foreign body detection device is used as the execution body for description, and the method specifically includes the following steps:
  • a piezoelectric substrate is added in the transmitting coil to generate excitation surface acoustic waves to detect temperature changes; and several detection coils are arranged on the coil of the wireless charging device to detect the current and voltage changes on the coils to detect the change of the wireless charging device.
  • the embodiment of the present application collects hyperspectral images of the target area (that is, the charging area of the wireless charging device), and detects the properties of foreign objects located in the target area according to the collected hyperspectral images.
  • the hyperspectral image of the area is used to detect the material of foreign objects in the target area, which is not limited by the size of metal foreign objects, overcomes the problem that small-sized metal foreign objects cannot be detected, and improves the detection accuracy of metal foreign objects.
  • S110 Collect a hyperspectral image of a target area; the target area is a charging area of a wireless charging device.
  • the wireless charging device includes a primary-side power transmitter and a secondary-side power receiver, and there will be a large space area (ie, charging area) between the primary-side power transmitter and the secondary-side power receiver, charging is performed on the wireless charging device.
  • charging area ie, charging area
  • metal foreign objects such as metal fragments, coins, cans, etc.
  • the foreign objects located in the charging area are detected by metal foreign objects, and the charging area of the wireless charging device can be preset as the target area.
  • the hyperspectral image of the target area is collected by the photographing device.
  • the metal foreign object detection device includes a photographing device, and the photographing device may be a hyperspectral image collector.
  • the acquired hyperspectral image of the target area contains hyperspectral information of foreign objects currently located in the target area.
  • the hyperspectral image of the target area contains the characteristic information of the foreign objects located in the target area (that is, the characteristics of the properties corresponding to each foreign object)
  • the hyperspectral image can be obtained by obtaining the hyperspectral image.
  • the hyperspectral reflectance curve of each pixel is used for property detection of foreign objects located in the target area. Specifically, according to the spectral wave range corresponding to the hyperspectral image, the reflectance corresponding to several bands divided by preset rules for each pixel point on the hyperspectral image in the spectral wave range is obtained, and each pixel point on the hyperspectral image is collected.
  • the metal foreign object detection device includes a hyperspectral image processor, and the hyperspectral image processor can obtain the hyperspectral reflectance curve of each pixel point in the hyperspectral image according to the hyperspectral image, and the hyperspectral image processor is connected to the photographing device. communication connection.
  • the spectral wave range corresponding to the collected hyperspectral image is 450nm-950nm; the preset rule is to divide a band every 10nm in the spectral wave range, then the spectral wave range is divided according to the preset rule There are 51 continuous bands. According to the spectral information contained in the hyperspectral image, the reflectance of each pixel on the hyperspectral image corresponding to each band is obtained; the 51 reflectances corresponding to each pixel on the hyperspectral image are collected to obtain the hyperspectral image. The hyperspectral reflectance curve of each pixel in the image.
  • the hyperspectral reflectance curve of all pixels in the hyperspectral image consists of 2448*2048*51 feature data point composition.
  • the photographing device since the photographing device is affected by the ambient light of the current target area and the dark current inside the photographing device when collecting the hyperspectral image of the target area, the spectral information contained in the collected hyperspectral image is biased .
  • the spectral information contained in the collected hyperspectral image is biased .
  • the white plate calibration method In order to correct the deviation of the hyperspectral image, it can be corrected by the white plate calibration method, and the specific process of obtaining the hyperspectral reflectance curve of each pixel in the hyperspectral image includes steps 11 to 13:
  • Step 11 Obtain the ambient light intensity of the target area and the dark current of the photographing device used for photographing the hyperspectral image
  • the ambient light of the target area when the hyperspectral image is collected can be detected by means of an illumination intensity detection sensor, etc., to obtain the ambient light intensity of the target area; according to the parameter setting of the photographing device, the value of the ambient light in the photographing device for photographing the hyperspectral image can be obtained by calculating. dark current.
  • Step 12 correcting the hyperspectral image according to the ambient light intensity and the dark current to obtain a hyperspectral reflectance image
  • Step 13 Obtain a hyperspectral reflectance curve of each pixel in the hyperspectral image according to the hyperspectral reflectance image.
  • the hyperspectral reflectance image obtained by calibrating the hyperspectral image according to the ambient light intensity and dark current includes pixels corresponding to several bands divided by preset rules within the spectral wave range corresponding to the hyperspectral image. Reflectance, collect several reflectances corresponding to each pixel on the hyperspectral reflectance image to obtain the hyperspectral reflectance curve of each pixel in the hyperspectral image.
  • the hyperspectral reflectance curve of each pixel point in the hyperspectral image After obtaining the hyperspectral reflectance curve of each pixel point in the hyperspectral image, in order to realize the attribute detection of foreign objects located in the charging area according to the characteristic information corresponding to the hyperspectral reflectance curve of each pixel point, all hyperspectral reflectance can be detected.
  • the rate curve is input to a pre-trained metal detection model to detect properties of objects located within the target area.
  • the metal detection model may be composed of an SVM classifier or a deep learning neural network.
  • the specific process of inputting all the hyperspectral reflectance curves into the metal detection model to detect the properties of the objects located in the target area includes: mapping all the hyperspectral reflectance curves to a high-dimensional feature space to obtain Feature data point set; inputting the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • hyperspectral images are high-dimensional data
  • each pixel has hundreds of adjacent spectral channels.
  • the traditional classification method of the classifier is to find a separation surface by modeling the density of different classes.
  • the high-dimensional density estimation process is affected by the Hughes effect, and high-dimensional data cannot be classified directly in the two-dimensional space.
  • a feature data point set can be obtained by mapping all hyperspectral reflectance curves to a high-dimensional feature space, and then the feature data point set can be input into the metal detection model to obtain the feature data point set of the object located in the target area.
  • Property detection result can be obtained by mapping all hyperspectral reflectance curves to a high-dimensional feature space, and then the feature data point set can be input into the metal detection model to obtain the feature data point set of the object located in the target area.
  • the metal detection model includes a kernel function and a classification function. Then map all hyperspectral reflectance curves to a high-dimensional feature space to obtain a feature data point set, and then input the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • the specific process can be as follows: : map all the hyperspectral reflectance curves into the kernel function in the metal detection model to a high-dimensional feature space to obtain the feature data point set; input the feature data point set into the classification in the metal detection model The function performs attribute detection on the feature data point set to obtain the attribute detection result of the object located in the target area.
  • hyperspectral images are linearly inseparable due to their inherent characteristics and the influence of the external environment.
  • mapping all the hyperspectral reflectance curves into the high-dimensional feature space by inputting the kernel function in the metal detection model to the high-dimensional feature space, the feature data point set is obtained, and the hyperspectral reflectance curve is mapped to the high-dimensional feature space, so that the obtained feature data
  • the point set becomes separable (classified).
  • the classification function can be constructed by a plurality of support vector machine-based classifiers (SVM); the attribute detection result includes metal and non-metal.
  • the device After obtaining the property detection results of the objects (foreign objects) located in the target area, if the property detection result of any object is metal, it means that there are metal foreign objects in the target area (that is, the charging area of the wireless charging device). At this time, metal foreign objects are detected.
  • the device generates an alarm signal and sends it to the wireless charging device, so that the wireless charging device can adjust the charging strategy according to the alarm information sent by the metal foreign object detection device, such as: stop charging or reduce charging power and other measures.
  • the wireless charging device performs the charging operation normally.
  • the hyperspectral image of the target area can be collected again according to the preset hyperspectral image collection period, And re-execute steps S120 to S140.
  • a hyperspectral image of a target area is collected; the target area is a charging area of a wireless charging device; objects located in the target area are detected by collecting the hyperspectral image of the target area. properties are checked. Obtain the hyperspectral reflectance curve of each pixel in the hyperspectral image; since the ambient light in the target area has an influence on the imaging of the hyperspectral image, the hyperspectral reflectance curve is obtained after eliminating the ambient light in the hyperspectral image, thereby improving the The detection accuracy of foreign metal objects.
  • FIG. 2 is a schematic flowchart of the training process of the metal detection model provided in the second embodiment of the present application.
  • this embodiment also provides a training process of a metal detection model, thereby further improving the accuracy of metal foreign object detection.
  • the method includes:
  • the metal detection model can be trained by using the training data including the hyperspectral reflectance information corresponding to several types of foreign body samples, so that the The trained metal detection model can detect the attributes corresponding to several types of foreign objects.
  • the training data including the hyperspectral reflectance information corresponding to several types of foreign body samples, so that the The trained metal detection model can detect the attributes corresponding to several types of foreign objects.
  • several types of foreign object samples included in the training data include metallic materials and non-metallic materials. Since wireless charging devices are mostly arranged outdoors, according to statistics, metal foreign objects that commonly appear in the charging area of wireless charging devices include: coins, steel nails, cans and other objects.
  • the foreign objects to be detected in the charging area of the wireless charging device specified in the art can be obtained as several types of foreign object samples included in the training data, such as 13 kinds of foreign objects required to be detected by the detection standard SAE J2954TM: paper clips, A4 paper clipped with paper clips, staples , A4 paper with staples, coins, steel nails, cans, aluminum sheets, steel sheets, copper coils, steel wool, metal foil with paper on the back and four forms of metal wire.
  • the hyperspectral reflectance information corresponding to several types of foreign body samples it is also necessary to obtain hyperspectral images corresponding to the several types of foreign body samples before acquiring the training data; Extract the hyperspectral reflectance information of various types of the foreign body samples from the hyperspectral images of the system; perform attribute annotation on the hyperspectral reflectance information of various types of the foreign body samples to generate the training data.
  • hyperspectral images corresponding to several types of foreign body samples are collected by a photographing device.
  • the photographing device may be a hyperspectral image collector.
  • the hyperspectral image corresponding to each type of foreign body sample contains the hyperspectral reflectance information of this type of foreign body sample.
  • For each type of foreign body sample according to the spectral wave range of the hyperspectral image corresponding to this type of foreign body sample, several bands divided by preset rules for each pixel point of this type of foreign body sample on the hyperspectral image within the spectral wave range are obtained
  • For the corresponding reflectance, a plurality of reflectances corresponding to all pixels of the foreign matter sample on the hyperspectral image are collected to obtain the hyperspectral reflectance information of the foreign matter sample of this type.
  • the hyperspectral reflectance information of various types of foreign body samples can be extracted from the hyperspectral images corresponding to the various types of foreign body samples.
  • the training data is generated from the hyperspectral reflectance information of various foreign body samples.
  • the hyperspectral reflectance information with attribute labeling obtained after attribute labeling of the hyperspectral reflectance information of various foreign body samples may be a foreign body sample data set, and the foreign body sample data set is divided according to the ratio of 7:3 for training data and test data.
  • the test data is input into the metal detection model to test the metal detection model, adjust the parameters in the metal detection model, and improve the detection accuracy of the metal detection model.
  • the photographing device since the photographing device is affected by the current ambient light and the dark current inside the photographing device when collecting hyperspectral images of several types of foreign matter samples, the spectral information contained in the collected hyperspectral images is biased.
  • the spectral information contained in the collected hyperspectral images is biased.
  • the white plate calibration method In order to correct the deviation of the hyperspectral image, it can be corrected by the white plate calibration method, and the original reflected light intensity I of the hyperspectral image is obtained by performing deviation correction according to the ambient light intensity W and the dark current B of the shooting device used to capture the hyperspectral image.
  • the correction formula for the reflectance R of the hyperspectral image is: Among them, B is an all-black calibration image obtained by keeping the environment without light, representing the dark current in the hyperspectral camera, which is a constant; W is an all-white calibration hyperspectral image obtained by collecting hyperspectral images on a standard white calibration plate , the reflectivity is 99%, indicating the ambient light intensity. According to the above formula, the influence of ambient light and dark current in the original reflected light intensity of the hyperspectral image is eliminated, thereby improving the accuracy of metal foreign object detection.
  • the metal detection model includes a kernel function and a classification function. Since hyperspectral images are high-dimensional data, each pixel has hundreds of adjacent spectral channels. The traditional classification method of the classifier is to find a separation surface by modeling the density of different classes. However, the high-dimensional density estimation process is affected by the Hughes effect, and high-dimensional data cannot be classified directly in the two-dimensional space.
  • the kernel function in the metal detection model can be used to map the training data to a high-dimensional feature space to obtain sample feature data, which can become separable data.
  • the kernel function may be a radial basis function (RBF) or other function.
  • a nonlinear transformation function ie, a kernel function
  • binary SVMs classifiers
  • Each binary SVM classifies the sample feature data corresponding to any two types of foreign body samples in the sample feature data without repetition, and obtains a separation hyperplane between the two types of the foreign body samples.
  • the determination rule of the separation hyperplane is that the distance from the feature data point closest to the separation hyperplane to the separation hyperplane is the largest in the two types of foreign matter samples.
  • the specific process of determining the separation hyperplane between each two types of foreign body samples according to the sample characteristic data corresponding to each two types of foreign body samples includes: The sample characteristic data is input into a generalized linear model to obtain a separation hyperplane between each two types of foreign matter samples; the separation hyperplane includes a weight coefficient and a deviation coefficient.
  • the sample feature data corresponding to any two types of foreign body samples in the sample feature data can be classified without repetition through the generalized linear model of each binary SVM (classifier).
  • the separation hyperplane includes a weight coefficient and a deviation coefficient.
  • the extracted support vector and the corresponding credibility of the support vector are input based on support vector machine (SVM) for training to generate a classification function, so as to realize the training of the metal detection model.
  • SVM support vector machine
  • the metal detection model needs to be iteratively trained.
  • the training data is input into the metal detection model for iterative training, so that the weight coefficient and the deviation coefficient satisfy a preset constraint condition.
  • the preset constraints are where ⁇ i is a slack variable, which is introduced to make the points in extreme cases (that is, points that cannot be classified) also meet the classification conditions.
  • ⁇ i is a slack variable, which is introduced to make the points in extreme cases (that is, points that cannot be classified) also meet the classification conditions.
  • the goal of SVM-based classifiers is to find a hyperplane that minimizes the possibility of misclassification of the sample feature data on the training data. This problem can be defined as
  • the first formula in formula (1) is the equivalent problem of finding the maximum margin
  • the second formula is the constraint condition
  • the third formula is the definition of the slack variable.
  • C is the penalty coefficient, which represents the penalty of the SVM for the wrongly classified samples. Because the slack variable was introduced before, the penalty coefficient is introduced here.
  • the kernel function When the data points are mapped to the high-dimensional feature space, their inner product The calculation amount of , is usually very large, and the use of the kernel function can make the calculation of this inner product in the initial low-dimensional space, which greatly reduces the amount of calculation.
  • the metal detection model can identify multiple properties of metal and non-metal objects (for example: ferromagnetic metals, non-ferromagnetic metals, non-metals, etc.), we use OAR (one-against-rest, term) technology M two-class SVM classifiers are designed, in which the mth classifier is used to separate the points belonging to the m attribute category from the rest, and then combine the maximum output of the M discriminant functions to complete the multi-classification problem. which is,
  • the embodiment of the present application also provides a metal foreign object detection device 3, which includes:
  • the hyperspectral image acquisition module 301 is used for acquiring hyperspectral images of a target area; the target area is a charging area of a wireless charging device;
  • a reflectance curve acquisition module 302 configured to acquire the hyperspectral reflectance curve of each pixel in the hyperspectral image
  • a detection module 303 configured to input all the hyperspectral reflectance curves into a metal detection model to detect properties of objects located in the target area;
  • the alarm module 304 is configured to generate an alarm signal and send it to the wireless charging device if the attribute detection result of any of the objects is metal.
  • the reflectance curve acquisition module includes:
  • an information acquisition unit configured to acquire the ambient light intensity of the target area and the dark current of the photographing device used for photographing the hyperspectral image
  • a correction unit configured to correct the hyperspectral image according to the ambient light intensity and the dark current to obtain a hyperspectral reflectance image
  • a reflectance curve obtaining unit configured to obtain a hyperspectral reflectance curve of each pixel point in the hyperspectral image according to the hyperspectral reflectance image.
  • the detection module includes:
  • a feature data point set calculation unit used to map all the hyperspectral reflectance curves to a high-dimensional feature space to obtain a feature data point set
  • a detection unit configured to input the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • the detection module 303 includes:
  • mapping unit for mapping all the hyperspectral reflectance curves to a high-dimensional feature space to obtain a set of feature data points
  • a detection unit configured to input the feature data point set into the metal detection model to obtain the attribute detection result of the object located in the target area.
  • a metal foreign object detection device collects a hyperspectral image of a target area; the target area is a charging area of a wireless charging device; objects located in the target area are detected by collecting the hyperspectral image of the target area properties are checked. Obtain the hyperspectral reflectance curve of each pixel in the hyperspectral image; since the ambient light in the target area has an influence on the imaging of the hyperspectral image, the hyperspectral reflectance curve is obtained after eliminating the ambient light in the hyperspectral image, thereby improving the The detection accuracy of foreign metal objects.
  • the device realizes the attribute detection of foreign objects located in the target area, that is, the charging area of the wireless charging device. Since the hyperspectral image contains rich spectral information, it can distinguish the physical properties of different materials. According to the hyperspectral image of the target area, the material detection of foreign objects in the target area is not limited by the size of metal foreign objects. The problem of metal foreign objects improves the detection accuracy of metal foreign objects; and there is no need to set different thresholds according to different environments.
  • FIG. 4 is a schematic structural diagram of a terminal device provided in Embodiment 4 of the present application.
  • the terminal device includes: a processor 41, a memory 42, and a computer program 43 stored in the memory 42 and executable on the processor 41, such as a program for a metal foreign object detection method.
  • the processor 41 executes the computer program 43, the steps in the above embodiments of the metal foreign object detection method are implemented, for example, steps S110 to S140 shown in FIG. 1 .
  • the computer program 43 may be divided into one or more modules, and the one or more modules are stored in the memory 42 and executed by the processor 41 to complete the present application.
  • the one or more modules may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 43 in the terminal device.
  • the computer program 43 can be divided into a hyperspectral image module, a reflectance curve acquisition module, a detection module and an alarm module, and the specific functions of each module are as follows:
  • a hyperspectral image module for collecting hyperspectral images of a target area; the target area is a charging area of a wireless charging device;
  • a reflectance curve acquisition module configured to acquire the hyperspectral reflectance curve of each pixel in the hyperspectral image
  • a detection module configured to input all the hyperspectral reflectance curves into a metal detection model to detect the properties of objects located in the target area;
  • the alarm module is configured to generate an alarm signal and send it to the wireless charging device if the attribute detection result of any of the objects is metal.
  • the terminal device may include, but is not limited to, a processor 41 , a memory 42 and a computer program 43 stored in the memory 42 .
  • FIG. 4 is only an example of a terminal device, and does not constitute a limitation on the terminal device. It may include more or less components than the one shown in the figure, or combine some components, or different components, such as
  • the terminal device may also include an input and output device, a network access device, a bus, and the like.
  • the processor 41 may be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the memory 42 may be an internal storage unit of the terminal device, such as a hard disk or a memory of the terminal device.
  • the memory 42 can also be an external storage device, such as a plug-in hard disk equipped on a terminal device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash memory card (Flash Card), etc.
  • the memory 42 may also include both an internal storage unit of the terminal device and an external storage device.
  • the memory 42 is used for storing the computer program and other programs and data required by the metal foreign object detection method.
  • the memory 42 can also be used to temporarily store data that has been output or is to be output.
  • the integrated modules/units if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium.
  • the computer program includes computer program code
  • the computer program code may be in the form of source code, object code, executable file or some intermediate form, and the like.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, removable hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.
  • the content contained in the computer-readable media may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable media Electric carrier signals and telecommunication signals are not included.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Geophysics (AREA)
  • Evolutionary Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

一种金属异物检测方法、装置及终端设备，该方法包括：采集目标区域的高光谱图像；目标区域为无线充电设备的充电区域（S110）；获取高光谱图像中各个像素点的高光谱反射率曲线（S120）；将所有高光谱反射率曲线输入金属检测模型检测位于目标区域内的物体的属性（S130）；若任一物体的属性检测结果为金属，则生成报警信号并发送至无线充电设备（S140）。解决无法检测小尺寸的金属异物以及调节繁琐的问题。

Description

金属异物检测方法、装置及终端设备 技术领域
本申请涉及物体检测的技术领域，尤其涉及一种金属异物检测方法、装置及终端设备。
背景技术
感应电能传输(Inductive Power Transfer,IPT)技术利用电磁感应原理将电能以非接触的方式由电源端传送到用电设备端，即实现了无线电能传输，具有安全可靠、灵活便捷、环境友好、可全天候工作等优点，因而近年来受到了广泛关注。IPT系统包括原边电能发射装置以及副边电能接收装置，当应用于电动汽车无线充电等领域时，原边电能发射装置和副边电能接收装置之间将存在较大的空间区域(能量传输区域)，如果有金属异物(如金属碎片、硬币、易拉罐等)进入该空间区域时，将可能阻碍能量传输或者降低能量传输效率，甚至由于金属异物被加热而引起燃烧事故。
在现有技术中主要通过两种方式对位于无线充电区域内的金属异物进行检测；方法一，通过在发射线圈内增加压电基板，激励声表面波检测温度变化，从而发现金属异物；方法二，在原边线圈和/或副边线圈上设置若干检测线圈，通过测量、监控检测线圈上的电流电压变化判断原边线圈和/或副边线圈上是否存在金属异物。但方法一具有不足之处：当金属异物尺寸过小时压电基板局部区域的声表面波的振荡频率变化很小，无法通过方准确检测到金属异物，检测准确度低；在发射线圈工作时产生的磁场会对声表面波产生影响，导致无法设定一个统一的温度阈值来判定各个区域是否存在金属异物。且方法二也具有不足之处：当金属异物的尺寸较小时，检测线圈无法准确检测；在没有金属异物时，由于每个检测线圈磁通量不一致，进而各个线圈输出电压值大小不一，没有办法确定统一的阈值，阈值参数设置较多且不易获取。
技术问题
本申请实施例的目的之一在于：提供一种金属异物检测方法、装置及终端设备，旨在解决无法检测小尺寸的金属异物以及调节繁琐的问题。
技术解决方案
为解决上述技术问题，本申请实施例采用的技术方案是：
第一方面，提供了一种金属异物检测方法，包括：
采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
获取所述高光谱图像中各个像素点的高光谱反射率曲线；
将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
在一个实施示例中，所述获取所述高光谱图像中各个像素点的高光谱反射率曲线，包括：
获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍摄装置的暗电流；
根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
根据所述高光谱反射率图像得到所述高光谱图像中各个像素点的高光谱反射率曲线。
在一个实施示例中，所述将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性，包括：
将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
在一个实施示例中，所述将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集，包括：
将所有所述高光谱反射率曲线输入所述金属检测模型中的核函数映射至高维特征空间，得到所述特征数据点集；
所述将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果，包括：
将所述特征数据点集输入所述金属检测模型中的分类函数对所述特征数据点集进行属性检测，得到位于所述目标区域内的物体的所述属性检测结果。
在一个实施示例中，在将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性之前，还包括：
获取训练数据；所述训练数据包括若干类异物样品对应的高光谱反射率信息；所述若干类异物样品包括金属材料和非金属材料；
通过所述核函数将所述训练数据映射至高维特征空间得到样本特征数据；
根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面；
从每一所述分离超平面中提取出支持向量并计算所述支持向量的可信度；
根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数。
在一个实施示例中，在获取训练数据之前，还包括：
获取所述若干类异物样品对应的高光谱图像；
从各类所述异物样品对应的所述高光谱图像中提取各类所述异物样品的高光谱反射率信息；
对各类所述异物样品的高光谱反射率信息进行属性标注生成所述训练数据。
在一个实施示例中，所述根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面，包括：
将每两类所述异物样品对应的所述样本特征数据输入广义线性模型得到每两类所述异物样品之间的分离超平面；所述分离超平面包括权重系数和偏离系数。
在一个实施示例中，在根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数之后，还包括：
将所述训练数据输入所述金属检测模型进行迭代训练，以使所述权重系数和所述偏离系数满足预设的约束条件。
第二方面，提供了一种金属异物检测装置，包括：
高光谱图像采集模块，用于采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
反射率曲线获取模块，用于获取所述高光谱图像中各个像素点的高光谱反射率曲线；
检测模块，用于将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
报警模块，用于若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
在一个实施示例中，所述反射率曲线获取模块包括：
信息获取单元，用于获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍摄装置的暗电流；
校正单元，用于根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
反射率曲线获取单元，用于根据所述高光谱反射率图像得到所述高光谱图像中各个像素点的高光谱反射率曲线。
在一个实施示例中，所述检测模块包括：
特征数据点集计算单元，用于将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
检测单元，用于将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
第三方面，提供一种终端设备，包括：存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现第一方面中金属异物检测方法。
有益效果
本申请实施例提供的一种金属异物检测方法、装置及终端设备的有益效果在于：采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；通过采集目标区域的高光谱图像对位于所述目标区域内的物体的属性进行检测。获取所述高光谱图像中各个像素点的高光谱反射率曲线；由于目标区域的环境光对高光谱图像的成像具有影响，消除高光谱图像中的环境光后得到高光谱反射率曲线，从而提高金属异物的检测准确度。将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；若任一所述物体对的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备，实现对位于所述目标区域即无线充电设备的充电区域内的异物的属性检测。由于高光谱图像包含丰富的光谱信息，能够区分不同材料的物理性质，根据目标区域的高光谱图像进行目标区域内的异物的材质检测，不受限于金属异物的尺寸，克服无法检测小尺寸的金属异物的问题，提高金属异物检测精确度；且不需要根据不同环境设置不同阈值。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或示范性技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图。
图1是本申请实施例一提供的金属异物检测方法的流程示意图；
图2是本申请实施例二提供的金属检测模型的训练过程的流程示意图；
图3是本申请实施例三提供的金属异物检测装置的结构示意图；
图4是本申请实施例四提供的终端设备的结构示意图。
本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
需说明的是，本申请的全文及上述附图中的术语“包括”以及它们任何变形，意图在于覆盖不排他的包含。例如包含一系列步骤或单元的过程、方法或系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。术语“上”、“下”、“左”、“右”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本申请的限制，对于本领域的普通技术人员而言，可以根据具体情况理解上述术语的具体含义。术语“第一”、“第二”仅用于便于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明技术特征的数量。“多个”的含义是两个或两个以上，除非另有明确具体的限定。
为了说明本申请所述的技术方案，以下结合具体附图及实施例进行详细说明。
实施例一
如图1所示，是本申请实施例一提供的金属异物检测方法的流程示意图。本实施例可适用于检测位于无线充电设备的充电区域内的金属异物的应用场景，该方法可以由金属异物检测装置执行，该装置可为终端设备、智能终端、平板或PC等；在本申请实施例中以金属异物检测装置作为执行主体进行说明，该方法具体包括如下步骤：
现有技术中采用在发射线圈内增加压电基板产生激励声表面波检测温度变化；以及在无线充电设备的线圈上设置若干检测线圈，检测线圈上的电流电压变化的方法来检测无线充电设备的充电区域中是否存在有金属异物。但上述两种金属异物检测方法均无法检测出充电区域中小尺寸的金属异物，导致检测准确度低。为解决这一技术问题，本申请实施例通过采集目标区域(即无线充电设备的充电区域)的高光谱图像，根据采集到的高光谱图像对位于目标区域内的异物的属性进行检测，根据目标区域的高光谱图像进行目标区域内的异物的材质检测，不受限于金属异物的尺寸，克服无法检测小尺寸的金属异物的问题，提高金属异物检测精确度。
S110、采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域。
由于无线充电设备包含原边电能发射装置和副边电能接收装置，且原边电能发射装置和副边电能接收装置之间将存在较大的空间区域(即充电区域)，在无线充电设备执行充电操作的过程中，如果有金属异物(如金属碎片、硬币、易拉罐等)进入该充电区域时，将可能阻碍能量传输或者降低能量传输效率，甚至由于金属异物被加热而引起燃烧事故，因此需对位于充电区域内的异物进行金属异物检测，可以预设无线充电设备的充电区域为目标区域。在无线充电设备执行充电操作的过程中，通过拍摄装置采集目标区域的高光谱 图像。可选的，金属异物检测装置包括拍摄装置，该拍摄装置可以为高光谱图像采集器。采集到的目标区域的高光谱图像包含当前位于目标区域内的异物的高光谱信息。
S120、获取所述高光谱图像中各个像素点的高光谱反射率曲线。
在采集到目标区域的高光谱图像之后，由于高光谱图像中的高光谱反射率包含位于目标区域内的异物的特征信息(即每一异物对应的属性的特性)，可以通过获取高光谱图像中各个像素点的高光谱反射率曲线以对位于目标区域内的异物进行属性检测。具体地，根据该高光谱图像对应的光谱波范围，得到高光谱图像上每一像素点在该光谱波范围内按预设规则划分的若干波段对应的反射率，集合高光谱图像上各个像素点对应的若干反射率得到高光谱图像中各个像素点的高光谱反射率曲线。可选的，金属异物检测装置包括高光谱图像处理器，可通过高光谱图像处理器根据高光谱图像获取高光谱图像中各个像素点的高光谱反射率曲线，该高光谱图像处理器与拍摄装置通信连接。
详细举例说明，若采集到的高光谱图像对应的光谱波范围为450nm～950nm；预设规则为在该光谱波范围内每间隔10nm划分一个波段，则在该光谱波范围内按预设规则划分有51个连续的波段，根据高光谱图像包含的光谱信息得到高光谱图像上每一像素点在每一波段对应的反射率；集合高光谱图像上各个像素点对应的51个反射率得到高光谱图像中各个像素点的高光谱反射率曲线。若高光谱图像的大小是2448*2048，它的每个像素点又对应51个波段的反射率值，因此高光谱图像中所有像素点的高光谱反射率曲线由2448*2048*51个特征数据点构成。
在一个实施示例中，由于拍摄装置在采集目标区域的高光谱图像时会受到当前目标区域的环境光以及该拍摄装置内部的暗电流影响，导致采集到的高光谱图像所包含的光谱信息存在偏差。为校正高光谱图像的偏差可以通过白板校正法进行校正，则获取所述高光谱图像中各个像素点的高光谱反射率曲线的具体过程包括步骤11至步骤13：
步骤11、获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍摄装置的暗电流；
可以通过光照强度检测传感器等方式检测目标区域在高光谱图像采集时的环境光，得到目标区域的环境光强度；根据拍摄装置的参数设置计算得到用于拍摄所述高光谱图像的拍摄装置中的暗电流。
步骤12、根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
由于高光谱图像的反射率与高光谱图像的原始反射光强度和环境光强度有关，根据获得的环境光强度和暗电流对高光谱图像进行偏差校正得到高光谱反射率图像。具体地，根 据环境光强度W和用于拍摄所述高光谱图像的拍摄装置的暗电流B对高光谱图像的原始反射光强度I进行偏差校正得到高光谱图像的反射率R的校正公式为：
Figure PCTCN2020103066-appb-000001
根据上述公式消除高光谱图像的原始反射光强度中环境光和暗电流的影响，得到高光谱反射率图像，从而提高金属异物检测的准确度。
步骤13、根据所述高光谱反射率图像得到所述高光谱图像中各个像素点的高光谱反射率曲线。
由于根据环境光强度和暗电流对高光谱图像进行校正得到的高光谱反射率图像，包含高光谱图像中各个像素点在高光谱图像对应的光谱波范围内按预设规则划分的若干波段对应的反射率，集合高光谱反射率图像上各个像素点对应的若干反射率得到高光谱图像中各个像素点的高光谱反射率曲线。
S130、将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性。
在获得高光谱图像中各个像素点的高光谱反射率曲线之后，为实现根据各个像素点的高光谱反射率曲线对应的特征信息对位于充电区域内的异物进行属性检测，可以将所有高光谱反射率曲线输入预先训练好的金属检测模型检测位于所述目标区域内的物体的属性。可选的，该金属检测模型可为SVM分类器或深度学习的神经网络构成。
在一个实施示例中，将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性的具体过程包括：将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
具体地，由于高光谱图像是一种高维数据，每个像素有几百个相邻的光谱通道。传统的分类器的分类方法为通过对不同类的密度进行建模，找到一个分离曲面。然而，高维密度估计过程受到休斯效应的影响，直接在二维空间无法对高维数据进行分类。为解决这一问题，可以通过将所有高光谱反射率曲线映射至高维特征空间得到特征数据点集，然后将特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
在一个实施示例中，金属检测模型包括核函数和分类函数。则将所有高光谱反射率曲线映射至高维特征空间得到特征数据点集，然后将特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果具体过程可为：将所有所述高光谱反射率曲线输入所述金属检测模型中的核函数映射至高维特征空间，得到所述特征数据点集； 将所述特征数据点集输入所述金属检测模型中的分类函数对所述特征数据点集进行属性检测，得到位于所述目标区域内的物体的所述属性检测结果。
具体地，高光谱图像由于其自身的固有特性和外界环境的影响，是线性不可分割的。通过将所有高光谱反射率曲线输入所述金属检测模型中的核函数映射至高维特征空间，得到所述特征数据点集，实现将高光谱反射率曲线映射至高维特征空间，使得到的特征数据点集变为可分割(分类)。将可分割的特征数据点集输入金属检测模型中的分类函数，以使金属检测模型中的分类函数对输入的特征数据点集进行属性分类，得到位于所述目标区域内的物体在高光谱图像中对应的像素点的属性；根据位于所述目标区域内的物体在高光谱图像中对应的所有像素点的属性得到该物体的属性检测结果。从而实现对位于目标区域内的异物(物体)的属性检测。可选的，该分类函数可由多个基于支持向量机的分类器(SVM)构建；该属性检测结果包括金属和非金属。
S140、若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
得到位于目标区域内的物体(异物)的属性检测结果后，若任一物体的属性检测结果为金属，则说明目标区域(即无线充电设备的充电区域)内具有金属异物，此时金属异物检测装置生成报警信号并发送至无线充电设备，以使该无线充电设备根据金属异物检测装置发送的报警信息，进行充电策略调整，例如：停止充电或降低充电功率等措施。
若位于目标区域内的所有物体的属性检测结果均为非金属，则无线充电设备正常执行充电操作。为确保无线充电设备在执行充电操作的过程中也能获知充电区域是否存在金属异物，在金属异物检测装置执行步骤S140之后还可以根据预设高光谱图像采集周期再次采集目标区域的高光谱图像，并重新执行步骤S120至步骤S140。
本申请实施例提供的一种金属异物检测方法，采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；通过采集目标区域的高光谱图像对位于所述目标区域内的物体的属性进行检测。获取所述高光谱图像中各个像素点的高光谱反射率曲线；由于目标区域的环境光对高光谱图像的成像具有影响，消除高光谱图像中的环境光后得到高光谱反射率曲线，从而提高金属异物的检测准确度。将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；若任一所述物体对的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备，实现对位于所述目标区域即无线充电设备的充电区域内的异物的属性检测。由于高光谱图像包含丰富的光谱信息，能够区分不同材料的物理性质，根据目标区域的高光谱图像进行目标区域内的异物的材质检测，不受限于金属异物的尺寸，克服无法检测小尺寸的金属异物的问题，提高金属异物检测精确度；且不 需要根据不同环境设置不同阈值。
实施例二
如图2所示的是本申请实施例二提供的金属检测模型的训练过程的流程示意图。在实施例一的基础上，本实施例还提供了金属检测模型的训练过程，从而进一步提高金属异物检测的准确率。该方法具体包括：
S210、获取训练数据；所述训练数据包括若干类异物样品对应的高光谱反射率信息；所述若干类异物样品包括金属材料和非金属材料；
由于每一类异物样品对应的高光谱反射率信息包含该类异物样品对应的属性的特性，可以采用包括若干类异物样品对应的高光谱反射率信息的训练数据对金属检测模型进行训练，以使训练得到的金属检测模型能够实现对若干类异物样品对应的属性进行检测。具体的，为使得训练得到的金属检测模型不但能够对金属异物进行检测还能对非金属异物进行检测，训练数据中包含的若干类异物样本包括金属材料和非金属材料。由于无线充电设备多布置于户外，经统计常见出现于无线充电设备的充电区域的金属异物包括：硬币，钢钉，易拉罐等物体。可获取本领域规定的无线充电设备的充电区域需检测的异物作为训练数据包含的若干类异物样品，例如检测标准SAE J2954TM要求检测的13种异物：回形针，夹有回形针的A4纸，订书针，订有订书针的A4纸，硬币，钢钉，易拉罐，铝片，钢片，铜质线圈，钢丝绒，背面是纸的金属箔和四种形态的金属线。
在一个实施示例中，为实现获取若干类异物样品对应的高光谱反射率信息，还需在获取训练数据之前，获取所述若干类异物样品对应的高光谱图像；从各类所述异物样品对应的所述高光谱图像中提取各类所述异物样品的高光谱反射率信息；对各类所述异物样品的高光谱反射率信息进行属性标注生成所述训练数据。
具体的，通过拍摄装置采集若干类异物样品对应的的高光谱图像。可选的，该拍摄装置可以为高光谱图像采集器。每类异物样品对应的高光谱图像包含该类异物样品的高光谱反射率信息。对于每一类异物样品，根据该类异物样品对应的高光谱图像的光谱波范围，得到高光谱图像上该类异物样品的每一像素点在该光谱波范围内按预设规则划分的若干波段对应的反射率，集合高光谱图像上该类异物样品的所有像素点对应的若干反射率得到该类所述异物样品的高光谱反射率信息。从而实现从各类所述异物样品对应的所述高光谱图像中提取各类所述异物样品的高光谱反射率信息。
在得到各类异物样品的高光谱反射率信息后，需对各类异物样品的高光谱反射率信息进行属性标注，以得到各类异物样品所属的属性对应的高光谱反射率信息；根据属性标注 后的各类异物样品的高光谱反射率信息生成训练数据。
可选的，对各类异物样品的高光谱反射率信息进行属性标注后得到的具有属性标注的高光谱反射率信息可为异物样品数据集，将该异物样品数据集按照7：3的比例划分为训练数据和测试数据。在金属检测模型训练好后，将测试数据输入金属检测模型，以对金属检测模型进行测试，调整金属检测模型中的参数，提高金属检测模型的检测准确度。
在一个实施示例中，由于拍摄装置在采集若干类异物样品的高光谱图像时会受到当前环境光以及该拍摄装置内部的暗电流影响，导致采集到的高光谱图像所包含的光谱信息存在偏差。为校正高光谱图像的偏差可以通过白板校正法进行校正，根据环境光强度W和用于拍摄所述高光谱图像的拍摄装置的暗电流B对高光谱图像的原始反射光强度I进行偏差校正得到高光谱图像的反射率R的校正公式为：
Figure PCTCN2020103066-appb-000002
其中，B为通过保持环境无光而获得的全黑校准图像，表示高光谱相机中的暗电流，为常数；W是通过在标准白色校准板上采集高光谱图像获得的全白色校准高光谱图像，反射率为99％，表示环境光强度。根据上述公式消除高光谱图像的原始反射光强度中环境光和暗电流的影响，从而提高金属异物检测的准确度。
S220、通过所述核函数将所述训练数据映射至高维特征空间得到样本特征数据；
金属检测模型包括核函数和分类函数。由于高光谱图像是一种高维数据，每个像素有几百个相邻的光谱通道。传统的分类器的分类方法为通过对不同类的密度进行建模，找到一个分离曲面。然而，高维密度估计过程受到休斯效应的影响，直接在二维空间无法对高维数据进行分类。为解决这一问题，可以通过金属检测模型中的核函数将训练数据映射至高维特征空间得到样本特征数据，变为可分割数据。可选的，该核函数可为径向基函数(RBF)等函数。
详细举例说明，若对样本特征数据中任意两类异物样品对应的高光谱反射率信息x 1,...x N∈R D进行分类，任意两类异物样品对应的高光谱反射率信息对应的标签为y 1,...y N＝±1(两类标签+1和-1)，将这些高光谱反射率信息输入一个非线性变换函数(即核函数)
Figure PCTCN2020103066-appb-000003
映射至高维特征空间得到这两类异物样品对应的样本特征数据
Figure PCTCN2020103066-appb-000004
S230、根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面；
为实现对样本特征数据中每类异物样品对应的样本特征数据进行分类，可采用多个二分类SVM(分类器)对样本特征数据中每两类所述异物样品对应的样本特征数据进行分类。 每一二分类SVM(分类器)不重复的对样本特征数据中任意两类异物样品对应的所述样本特征数据进行分类，得到这两类所述异物样品之间的分离超平面。具体地，该分离超平面的确定规则为两类异物样品中距离该分离超平面最近的特征数据点到该分离超平面的距离最大。
在一个实施示例中，根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面的具体过程包括：将每两类所述异物样品对应的所述样本特征数据输入广义线性模型得到每两类所述异物样品之间的分离超平面；所述分离超平面包括权重系数和偏离系数。
具体地，可通过每一二分类SVM(分类器)的广义线性模型不重复的对样本特征数据中任意两类异物样品对应的所述样本特征数据进行分类。将每两类所述异物样品对应的所述样本特征数据输入广义线性模型得到每两类所述异物样品之间的分离超平面；所述分离超平面包括权重系数和偏离系数。可选的，该广义线性模型可为：g(z)＝w Tz+b＝0；其中g(z)为每两类所述异物样品之间的分离超平面；w为权重系数；b为偏离系数。
S240、从每一所述分离超平面中提取出支持向量并计算所述支持向量的可信度；
将位于每一分离超平面中的数据点提取出来作为支持向量，并将提取得到的每一支持向量输入置信度计算公式得到每一支持向量的可信度。可选的，从每一分离超平面中提取出使每两类异物样品对应的样本特征数据分离间隔最大的数据点作为支持向量，并将提取到的每一支持向量输入置信度计算公式得到每一支持向量的可信度。
S250、根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数。
将提取得到的支持向量以及支持向量对应的可信度输入基于支持向量机(SVM)进行训练生成分类函数，从而实现对金属检测模型的训练。
在一个实施示例中，在根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数之后，为提高金属检测模型的检测精度，还需对金属检测模型进行迭代训练。将所述训练数据输入所述金属检测模型进行迭代训练，以使所述权重系数和所述偏离系数满足预设的约束条件。
为使得金属检测模型的检测结果更加准确，每两类异物样品对应的样品特征数据之间的分离超平面需满足两类异物样品对应的样品特征数据中距离该分离超平面最近的特征数据点到该分离超平面的距离margin最大，即为训练数据中每两类所述异物样品之间的分离超平面中的权重系数w和偏离系数b需满足的预设的约束条件。
可选的，该预设的约束条件为
Figure PCTCN2020103066-appb-000005
其中ξ i是松弛变量，引入它 是为了使一下处于极端情况的点(即无法被分类的点)也能满足分类的条件。基于SVM的分类器的目标是找到一个使训练数据上的样品特征数据被错误分类的可能性最小的超平面，这个问题可以被定义为
Figure PCTCN2020103066-appb-000006
公式(1)中的第一式是求最大margin的等价问题，第二式是约束条件，第三式是松弛变量的定义。C是惩罚系数，代表着SVM对错分的样本的惩罚，因为之前引入了松弛变量，所以这里引入惩罚系数。对(1)式使用拉格朗日乘数法，得到(2)式
Figure PCTCN2020103066-appb-000007
其中α i和μ i是拉格朗日乘子。最后对(2)式求权重系数w、偏离系数b和ξ i的偏导数，并使其等于0。得到下式
Figure PCTCN2020103066-appb-000008
这就计算得到了权重系数w，之后通过式(4)得到偏离系数b，
Figure PCTCN2020103066-appb-000009
将式(3)代入式(2)和式(1)中的问题，可以被转化为其对偶形式，
Figure PCTCN2020103066-appb-000010
其中
Figure PCTCN2020103066-appb-000011
是核函数。当数据点被映射到高维特征空间后，其内积
Figure PCTCN2020103066-appb-000012
的计算量通常很大，使用核函数可以使这种内积的计算在初始的低维空间进行，这就大大降低了计算量。我们在这里选择高斯核函数(式(6))，减少参数的使用。
Figure PCTCN2020103066-appb-000013
为了实现金属检测模型能够对金属和非金属属性物体的多个属性(例如：铁磁性金属、非铁磁性金属、非金属等属性)进行识别，我们使用OAR(one-against-rest，术语)技术设计了M个二分类的SVM分类器，其中第m个分类器用于将属于m属性这一类的点与其余点分开，然后结合这M个判别函数的最大输出，从而完成多分类问题。即，
Figure PCTCN2020103066-appb-000014
结合上式，因为金属检测模型中一共训练了M个二分类器，判断输入的数据点x属于哪一类属性时，就将这一点代入这M个判别函数中，输出值最大的那一个判别函数所对应的类别就是这一点x的类别(即属性)。
实施例三
如图3所示的是本申请实施例三提供的金属异物检测装置。在实施例一或二的基础上，本申请实施例还提供了一种金属异物检测装置3，该装置包括：
高光谱图像采集模块301，用于采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
反射率曲线获取模块302，用于获取所述高光谱图像中各个像素点的高光谱反射率曲线；
检测模块303，用于将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
报警模块304，用于若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
在一个实施示例中，所述反射率曲线获取模块包括：
信息获取单元，用于获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍摄装置的暗电流；
校正单元，用于根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
反射率曲线获取单元，用于根据所述高光谱反射率图像得到所述高光谱图像中各个像 素点的高光谱反射率曲线。
在一个实施示例中，所述检测模块包括：
特征数据点集计算单元，用于将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
检测单元，用于将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
在一个实施示例中，检测模块303包括：
映射单元，用于将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
检测单元，用于将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
本申请实施例提供的一种金属异物检测装置，采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；通过采集目标区域的高光谱图像对位于所述目标区域内的物体的属性进行检测。获取所述高光谱图像中各个像素点的高光谱反射率曲线；由于目标区域的环境光对高光谱图像的成像具有影响，消除高光谱图像中的环境光后得到高光谱反射率曲线，从而提高金属异物的检测准确度。将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；若任一所述物体对的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备，实现对位于所述目标区域即无线充电设备的充电区域内的异物的属性检测。由于高光谱图像包含丰富的光谱信息，能够区分不同材料的物理性质，根据目标区域的高光谱图像进行目标区域内的异物的材质检测，不受限于金属异物的尺寸，克服无法检测小尺寸的金属异物的问题，提高金属异物检测精确度；且不需要根据不同环境设置不同阈值。
实施例四
图4是本申请实施例四提供的终端设备的结构示意图。该终端设备包括：处理器41、存储器42以及存储在所述存储器42中并可在所述处理器41上运行的计算机程序43，例如用于金属异物检测方法的程序。所述处理器41执行所述计算机程序43时实现上述金属异物检测方法实施例中的步骤，例如图1所示的步骤S110至S140。
示例性的，所述计算机程序43可以被分割成一个或多个模块，所述一个或者多个模块被存储在所述存储器42中，并由所述处理器41执行，以完成本申请。所述一个或多个模块可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程序43在所述终端设备中的执行过程。例如，所述计算机程序43可以被分割成高光谱图像 模块、反射率曲线获取模块、检测模块和报警模块，各模块具体功能如下：
高光谱图像模块，用于采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
反射率曲线获取模块，用于获取所述高光谱图像中各个像素点的高光谱反射率曲线；
检测模块，用于将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
报警模块，用于若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
所述终端设备可包括，但不仅限于，处理器41、存储器42以及存储在所述存储器42中的计算机程序43。本领域技术人员可以理解，图4仅仅是终端设备的示例，并不构成对终端设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述终端设备还可以包括输入输出设备、网络接入设备、总线等。
所述处理器41可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器42可以是所述终端设备的内部存储单元，例如终端设备的硬盘或内存。所述存储器42也可以是外部存储设备，例如终端设备上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。进一步地，所述存储器42还可以既包括终端设备的内部存储单元也包括外部存储设备。所述存储器42用于存储所述计算机程序以及金属异物检测方法所需的其他程序和数据。所述存储器42还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘 述。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
所述集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。其中，所述计算机程序包括计算机程序代码，所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质可以包括：能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是，所述计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，计算机可读介质不包括电载波信号和电信信号。
以上仅为本申请的可选实施例而已，并不用于限制本申请。对于本领域的技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的权利要求范围之内。

Claims (12)

  1. 一种金属异物检测方法，其特征在于，包括：
    采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
    获取所述高光谱图像中各个像素点的高光谱反射率曲线；
    将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
    若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
  2. 如权利要求1所述的金属异物检测方法，其特征在于，所述获取所述高光谱图像中各个像素点的高光谱反射率曲线，包括：
    获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍摄装置的暗电流；
    根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
    根据所述高光谱反射率图像得到所述高光谱图像中各个像素点的高光谱反射率曲线。
  3. 如权利要求1或2所述的金属异物检测方法，其特征在于，所述将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性，包括：
    将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
    将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
  4. 如权利要求3所述的金属异物检测方法，其特征在于，所述将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集，包括：
    将所有所述高光谱反射率曲线输入所述金属检测模型中的核函数映射至高维特征空间，得到所述特征数据点集；
    所述将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果，包括：
    将所述特征数据点集输入所述金属检测模型中的分类函数对所述特征数据点集进行属性检测，得到位于所述目标区域内的物体的所述属性检测结果。
  5. 如权利要求4所述的金属异物检测方法，其特征在于，在将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性之前，还包括：
    获取训练数据；所述训练数据包括若干类异物样品对应的高光谱反射率信息；所述若 干类异物样品包括金属材料和非金属材料；
    通过所述核函数将所述训练数据映射至高维特征空间得到样本特征数据；
    根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面；
    从每一所述分离超平面中提取出支持向量并计算所述支持向量的可信度；
    根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数。
  6. 如权利要求5所述的金属异物检测方法，其特征在于，在获取训练数据之前，还包括：
    获取所述若干类异物样品对应的高光谱图像；
    从各类所述异物样品对应的所述高光谱图像中提取各类所述异物样品的高光谱反射率信息；
    对各类所述异物样品的高光谱反射率信息进行属性标注生成所述训练数据。
  7. 如权利要求5所述的金属异物检测方法，其特征在于，所述根据每两类所述异物样品对应的所述样本特征数据确定每两类所述异物样品之间的分离超平面，包括：
    将每两类所述异物样品对应的所述样本特征数据输入广义线性模型得到每两类所述异物样品之间的分离超平面；所述分离超平面包括权重系数和偏离系数。
  8. 如权利要求7所述的金属异物检测方法，其特征在于，在根据所述支持向量以及所述支持向量对应的可信度生成所述分类函数之后，还包括：
    将所述训练数据输入所述金属检测模型进行迭代训练，以使所述权重系数和所述偏离系数满足预设的约束条件。
  9. 一种金属异物检测装置，其特征在于，包括：
    高光谱图像采集模块，用于采集目标区域的高光谱图像；所述目标区域为无线充电设备的充电区域；
    反射率曲线获取模块，用于获取所述高光谱图像中各个像素点的高光谱反射率曲线；
    检测模块，用于将所有所述高光谱反射率曲线输入金属检测模型检测位于所述目标区域内的物体的属性；
    报警模块，用于若任一所述物体的属性检测结果为金属，则生成报警信号并发送至所述无线充电设备。
  10. 根据权利要求9所述的金属异物检测装置，其特征在于，所述反射率曲线获取模块包括：
    信息获取单元，用于获取所述目标区域的环境光强度和用于拍摄所述高光谱图像的拍 摄装置的暗电流；
    校正单元，用于根据所述环境光强度和所述暗电流对所述高光谱图像进行校正得到高光谱反射率图像；
    反射率曲线获取单元，用于根据所述高光谱反射率图像得到所述高光谱图像中各个像素点的高光谱反射率曲线。
  11. 根据权利要求9或10所述的金属异物检测装置，其特征在于，所述检测模块包括：
    特征数据点集计算单元，用于将所有所述高光谱反射率曲线映射至高维特征空间得到特征数据点集；
    检测单元，用于将所述特征数据点集输入所述金属检测模型得到位于所述目标区域内的物体的所述属性检测结果。
  12. 一种终端设备，其特征在于，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现如权利要求1至8任一项所述金属异物检测方法的步骤。
PCT/CN2020/103066 2020-07-20 2020-07-20 金属异物检测方法、装置及终端设备 Ceased WO2022016328A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2020/103066 WO2022016328A1 (zh) 2020-07-20 2020-07-20 金属异物检测方法、装置及终端设备
CN202080001313.2A CN113260882B (zh) 2020-07-20 2020-07-20 金属异物检测方法、装置及终端设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/103066 WO2022016328A1 (zh) 2020-07-20 2020-07-20 金属异物检测方法、装置及终端设备

Publications (1)

Publication Number Publication Date
WO2022016328A1 true WO2022016328A1 (zh) 2022-01-27

Family

ID=77181280

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/103066 Ceased WO2022016328A1 (zh) 2020-07-20 2020-07-20 金属异物检测方法、装置及终端设备

Country Status (2)

Country Link
CN (1) CN113260882B (zh)
WO (1) WO2022016328A1 (zh)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114648534A (zh) * 2022-05-24 2022-06-21 成都理工大学 基于视频帧聚类的管网缺陷智能识别方法及装置、介质
US20230030303A1 (en) * 2021-07-29 2023-02-02 Electronics And Telecommunications Research Institute Image processing-based foreign substance detection method in wireless charging system and device performing method
CN115687928A (zh) * 2022-11-11 2023-02-03 广芯微电子（广州）股份有限公司 一种异物识别模型的构建方法、装置及异物识别方法
US20230267598A1 (en) * 2022-02-23 2023-08-24 Applied Materials Israel Ltd. Gray level ratio inspection
CN116665051A (zh) * 2023-05-29 2023-08-29 山东睿政信息科技有限公司 基于rgb图像重建高光谱图像的垃圾中金属再筛选方法
CN116993717A (zh) * 2023-09-25 2023-11-03 深圳市五轮科技股份有限公司 一种电子烟发热丝生产质量视觉检测方法
CN117691766A (zh) * 2022-09-09 2024-03-12 荣耀终端有限公司 一种无线充电异物温度检测系统及检测方法
CN119150028A (zh) * 2024-11-14 2024-12-17 江西升泰克电子科技有限公司 一种基于金属传感器的金属智能识别方法及系统
CN120181113A (zh) * 2025-05-21 2025-06-20 北京众驰伟业科技发展有限公司 一种全自动凝血测试仪的样本条码扫描系统

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114152621A (zh) * 2021-11-30 2022-03-08 联想(北京)有限公司 一种处理方法及处理装置、处理系统
CN115343774B (zh) * 2022-07-11 2025-04-11 深圳大学 一种检测金属物体的方法、装置、终端设备以及存储介质
WO2024011362A1 (zh) * 2022-07-11 2024-01-18 深圳大学 一种检测金属物体的方法、装置、终端设备以及存储介质
CN115165772B (zh) * 2022-07-25 2024-11-01 江苏大学 一种基于高光谱下籽粒含杂率检测装置及其检测方法
CN119428312A (zh) * 2025-01-10 2025-02-14 安徽省充换电有限责任公司 一种充电桩安全监控方法及系统
CN120932767B (zh) * 2025-10-11 2026-02-06 杭州国光旅游用品有限公司 一种多模态数据融合的湿巾中金属含量质检方法及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101710334A (zh) * 2009-12-04 2010-05-19 大连理工大学 基于图像哈希的大规模图像库检索方法
CN104568778A (zh) * 2015-01-15 2015-04-29 浙江理工大学 一种基于高光谱成像的纺织品成分鉴别方法
CN106412293A (zh) * 2016-09-29 2017-02-15 宇龙计算机通信科技(深圳)有限公司 一种异物检测方法、装置及终端设备
CN107860475A (zh) * 2017-11-03 2018-03-30 中惠创智无线供电技术有限公司 一种电动汽车及用于电动汽车的异物检测装置
CN207166228U (zh) * 2017-12-13 2018-03-30 桑堇馨 车载无线充电装置
US20190121401A1 (en) * 2017-10-20 2019-04-25 Canon Kabushiki Kaisha Electronic apparatus, wireless power supply system, and control methods thereof
CN210062694U (zh) * 2019-05-20 2020-02-14 温州大学 一种用于电动汽车无线充电的异物检测装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102609703B (zh) * 2012-03-05 2013-10-02 中国科学院对地观测与数字地球科学中心 一种高光谱图像中目标地物检测方法及装置
KR101634792B1 (ko) * 2014-10-27 2016-06-29 한국지질자원연구원 초분광 영상의 분광혼합분석에 의한 표적물질 탐지방법
US10902581B2 (en) * 2017-06-19 2021-01-26 Apeel Technology, Inc. System and method for hyperspectral image processing to identify foreign object
CN109146890B (zh) * 2018-07-16 2020-07-31 西安电子科技大学 基于滤波器的高光谱图像的异常目标检测方法
CN110363186A (zh) * 2019-08-20 2019-10-22 四川九洲电器集团有限责任公司 一种异常检测方法、装置及计算机存储介质、电子设备

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101710334A (zh) * 2009-12-04 2010-05-19 大连理工大学 基于图像哈希的大规模图像库检索方法
CN104568778A (zh) * 2015-01-15 2015-04-29 浙江理工大学 一种基于高光谱成像的纺织品成分鉴别方法
CN106412293A (zh) * 2016-09-29 2017-02-15 宇龙计算机通信科技(深圳)有限公司 一种异物检测方法、装置及终端设备
US20190121401A1 (en) * 2017-10-20 2019-04-25 Canon Kabushiki Kaisha Electronic apparatus, wireless power supply system, and control methods thereof
CN107860475A (zh) * 2017-11-03 2018-03-30 中惠创智无线供电技术有限公司 一种电动汽车及用于电动汽车的异物检测装置
CN207166228U (zh) * 2017-12-13 2018-03-30 桑堇馨 车载无线充电装置
CN210062694U (zh) * 2019-05-20 2020-02-14 温州大学 一种用于电动汽车无线充电的异物检测装置

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230030303A1 (en) * 2021-07-29 2023-02-02 Electronics And Telecommunications Research Institute Image processing-based foreign substance detection method in wireless charging system and device performing method
US12260565B2 (en) * 2021-07-29 2025-03-25 Electronics And Telecommunications Research Institute Image processing-based foreign substance detection method in wireless charging system and device performing method
US20230267598A1 (en) * 2022-02-23 2023-08-24 Applied Materials Israel Ltd. Gray level ratio inspection
US12175656B2 (en) * 2022-02-23 2024-12-24 Applied Materials Israel Ltd. Gray level ratio inspection
CN114648534A (zh) * 2022-05-24 2022-06-21 成都理工大学 基于视频帧聚类的管网缺陷智能识别方法及装置、介质
CN117691766A (zh) * 2022-09-09 2024-03-12 荣耀终端有限公司 一种无线充电异物温度检测系统及检测方法
CN117691766B (zh) * 2022-09-09 2025-08-29 荣耀终端股份有限公司 一种无线充电异物温度检测系统及检测方法
CN115687928B (zh) * 2022-11-11 2023-06-27 广芯微电子（广州）股份有限公司 一种异物识别模型的构建方法、装置及异物识别方法
CN115687928A (zh) * 2022-11-11 2023-02-03 广芯微电子（广州）股份有限公司 一种异物识别模型的构建方法、装置及异物识别方法
CN116665051B (zh) * 2023-05-29 2024-03-08 谢馥聪 基于rgb图像重建高光谱图像的垃圾中金属再筛选方法
CN116665051A (zh) * 2023-05-29 2023-08-29 山东睿政信息科技有限公司 基于rgb图像重建高光谱图像的垃圾中金属再筛选方法
CN116993717B (zh) * 2023-09-25 2024-02-13 深圳市五轮科技股份有限公司 一种电子烟发热丝生产质量视觉检测方法
CN116993717A (zh) * 2023-09-25 2023-11-03 深圳市五轮科技股份有限公司 一种电子烟发热丝生产质量视觉检测方法
CN119150028A (zh) * 2024-11-14 2024-12-17 江西升泰克电子科技有限公司 一种基于金属传感器的金属智能识别方法及系统
CN120181113A (zh) * 2025-05-21 2025-06-20 北京众驰伟业科技发展有限公司 一种全自动凝血测试仪的样本条码扫描系统

Also Published As

Publication number Publication date
CN113260882B (zh) 2024-03-12
CN113260882A (zh) 2021-08-13

Similar Documents

Publication Publication Date Title
WO2022016328A1 (zh) 金属异物检测方法、装置及终端设备
US10062008B2 (en) Image based object classification
CN106469301B (zh) 自适应可调节的人脸识别方法和装置
US9519942B2 (en) Methods and a computing device for determining whether a mark is genuine
US7734097B1 (en) Detecting objects in images with covariance matrices
US9466000B2 (en) Dynamic Bayesian Networks for vehicle classification in video
CN108765465B (zh) 一种无监督sar图像变化检测方法
US20180089505A1 (en) System and method for deep network fusion for fast and robust object detection
US10796144B2 (en) Method and device for classifying scanned documents
CN109993201A (zh) 一种图像处理方法、装置和可读存储介质
US20140198968A1 (en) Duplicate check image resolution
Guofeng et al. Traffic sign recognition based on SVM and convolutional neural network
CN111814562A (zh) 车辆的识别方法、车辆识别模型的训练方法及相关装置
CN117292179B (zh) 基于质检场景的难例筛选方法、系统、设备及存储介质
Tian et al. License plate detection in an open environment by density-based boundary clustering
CN115343774B (zh) 一种检测金属物体的方法、装置、终端设备以及存储介质
Promsuk et al. Numerical Reader System for Digital Measurement Instruments Embedded Industrial Internet of Things.
Albatal et al. Empirical exploration of extreme SVM-RBF parameter values for visual object classification
CN119169634A (zh) 一种基于深度学习的文字识别方法、装置及存储介质
CN116451112B (zh) 一种电磁信号识别方法、系统、电子设备及介质
Ye et al. BIVS: block image and voting strategy for weather image classification
CN110097135B (zh) 一种基于双张量的全息衍射标签图像识别算法
Tripathi et al. An Effective Machine Learning Approach to Detect Satellite Signals in Passive RF Space Domain Awareness Data
CN113640750A (zh) 基于图像的雷达信号波形调制方式识别方法、平台及介质
CN116758314A (zh) 一种基于SIFT和K-means算法的学术图像匹配查重系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20946129

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 25/04/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20946129

Country of ref document: EP

Kind code of ref document: A1