CN112163586A - Feature extraction method and device of target object and computer readable medium - Google Patents
Feature extraction method and device of target object and computer readable medium Download PDFInfo
- Publication number
- CN112163586A CN112163586A CN202011061687.1A CN202011061687A CN112163586A CN 112163586 A CN112163586 A CN 112163586A CN 202011061687 A CN202011061687 A CN 202011061687A CN 112163586 A CN112163586 A CN 112163586A
- Authority
- CN
- China
- Prior art keywords
- image
- polarization
- target object
- target
- gradient
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a method, a device and a computer readable medium for extracting the characteristics of a target object, wherein the method comprises the following steps: obtaining a Stokes parameter map according to a target radiation image of a target object; obtaining polarization information of the surface of the target object according to the Stokes parameter map, wherein the polarization information comprises a polarization degree and a polarization angle; obtaining a polarization operator according to the polarization information, wherein the polarization operator is used for calculating the gray level variation and the variation direction of the image; acquiring an unbiased image of a target object; and performing feature extraction on the unbiased image by using a polarization operator to obtain a target feature image, wherein the target feature image represents the outline of the target object. The method and the device can improve the accuracy of feature extraction of the target object.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a method and an apparatus for extracting features of a target object, and a computer readable medium.
Background
In the technical field of image information identification and extraction, detailed features such as edges and contours of images are often information which needs to be focused. Because the edge of the image often contains rich intrinsic information, the quality of the edge and the outline information of the image is the key for accurately identifying the image.
Although edge extraction is a very important subject in the field of image recognition at present, some problems that are not solved still exist in image information recognition and extraction, for example, although the extraction of image basic information can be realized in the prior art, information contained in a target object is complex, different types of objects often exist, when feature extraction is performed by using the prior image recognition technology, different types of target objects cannot be extracted differently, but contour features of each object contained in the target object are extracted, so that the accuracy of feature extraction of the target object is low.
Therefore, it is desirable to provide an image recognition and detection method that solves the above-mentioned problems.
Disclosure of Invention
The technical problem to be solved by the invention is that when the features of the target object are extracted, different types of target objects cannot be differentially extracted, but the contour features of all objects contained in the target object are extracted, so that the accuracy of extracting the features of the target object is low. Therefore, the present invention provides a method, an apparatus, and a computer-readable medium for extracting features of a target object, which can improve the accuracy of extracting features of the target object.
In a first aspect, an embodiment of the present invention provides a method for extracting features of a target object, including:
obtaining a Stokes parameter map according to a target radiation image of a target object;
obtaining polarization information of the surface of the target object according to the Stokes parameter map, wherein the polarization information comprises a polarization degree and a polarization angle;
obtaining a polarization operator according to the polarization information, wherein the polarization operator is used for calculating the gray level variation and the variation direction of the image;
acquiring an unbiased image of the target object;
and performing feature extraction on the unbiased image by using the polarization operator to obtain a target feature image, wherein the target feature image represents the outline of the target object.
Optionally, the obtaining a Stokes reference map according to a target radiation image of a target object includes:
respectively collecting target radiation images I of the target object0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight-hand circular polarization for characterizing the target radiation image, said ILCPFor characterizing the left-hand circular polarization of the target radiation image.
Optionally, the obtaining polarization information of the surface of the target object according to the Stokes parameter map includes:
calculating and obtaining a polarization information image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
Optionally, the performing feature extraction on the unbiased image by using the polarization operator to obtain a target feature image includes:
performing Gaussian filtering on the unbiased image to obtain a first characteristic image, wherein the first characteristic image is used for representing the image of the unbiased image after noise points are removed;
determining a first gradient amplitude and a gradient direction of the first characteristic image by utilizing finite difference calculation, wherein the first gradient amplitude is the amplitude of a partial derivative of the first characteristic image along the coordinate axis direction, the gradient direction is the direction of the first characteristic image with the fastest gradient change, and each first gradient amplitude corresponds to one pixel position;
acquiring a second gradient amplitude from the first gradient amplitude, wherein the second gradient amplitude is an amplitude image with non-edge pixels removed;
and determining a target characteristic image from the second gradient amplitude by using a preset double-threshold algorithm.
Optionally, the determining a first gradient magnitude and a gradient direction of the first feature image by using finite difference calculation includes:
determining a horizontal difference of the first characteristic image in the horizontal direction according to the following formula:
fx(x,y)＝S1＝I0-I90
wherein f isx(x, y) for characterizing the horizontal difference;
determining a vertical difference of the vertical direction of the first feature image according to the following formula:
fy(x,y)＝S2＝I45-I135
wherein f isy(x, y) for characterizing the vertical differential;
determining a first gradient magnitude and a gradient direction of the first feature image according to the following formulas by using the horizontal difference and the vertical difference:
wherein M (x, y) is used to characterize the first gradient magnitude and θ (x, y) is used to characterize the gradient direction.
In a second aspect, an embodiment of the present invention further provides a device for extracting features of a target object, including: the device comprises a first determining module, a second determining module, a third determining module, an obtaining module and a fourth determining module;
the first determination module is used for obtaining a Stokes parameter map according to a target radiation image of a target object;
the second determining module is configured to obtain polarization information of the surface of the target object according to the Stokes parameter map determined by the first determining module, where the polarization information includes a polarization degree and a polarization angle;
the third determining module is configured to obtain a polarization operator according to the polarization information determined by the second determining module, where the polarization operator is used to calculate a gray scale variation and a variation direction of an image;
the acquisition module is used for acquiring an unbiased image of the target object;
the fourth determining module is configured to perform feature extraction on the unbiased image acquired by the acquiring module by using the polarization operator determined by the third determining module to obtain a target feature image, where the target feature image represents a contour of the target object.
Optionally, the first determining module is configured to perform the following operations:
respectively collecting target radiation images I of the target object0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight-hand circular polarization for characterizing the target radiation image, said ILCPFor characterizing the left-hand circular polarization of the target radiation image.
Optionally, the second determining module is configured to perform the following operations:
calculating and obtaining a polarization information image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
In a third aspect, another embodiment of the present invention further provides a feature extraction apparatus for a target object, including at least one memory and at least one processor;
the at least one memory to store a machine readable program;
the at least one processor is configured to invoke the machine-readable program to perform the method of any of the above first aspects.
In a fourth aspect, the present invention also provides a computer-readable medium, on which computer instructions are stored, and when executed by a processor, the computer instructions cause the processor to execute the method of any one of the first aspect.
The method, the device and the computer readable medium for extracting the features of the target object have the following beneficial effects:
when the characteristics of the target object are extracted, firstly, a stokes parameter map is further obtained by obtaining a radiation image of the target object, then, the polarization information of the polarization degree and the polarization angle of the surface of the target object can be calculated and determined through the stokes parameter map, a polarization operator for extracting the characteristics of the image can be determined through the polarization information, and therefore, the characteristics of the unbiased image are extracted through the polarization operator to obtain the target characteristic image of the target object. Therefore, the polarization operator for feature extraction of the image is determined by using the polarization information. Since the common polarization operator has no specific physical meaning when extracting the contour features of the image, all contour information in the target image is extracted. The polarization operator is determined by the polarization information, the polarization degree in the polarization information is a value between 0 and 1, and different object materials correspond to different polarization degree values, so that when the polarization operator is used for feature extraction of a target, the contour features of the corresponding target object can be accurately obtained by determining different polarization degree values, and the accuracy of feature extraction of the target object can be improved.
Drawings
Fig. 1 is a flowchart of a feature extraction method for a target object according to an embodiment of the present invention;
fig. 2 is a flowchart of a feature extraction method for a target object according to another embodiment of the present invention;
fig. 3 is a schematic diagram of an apparatus where a feature extraction device of a target object is located according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a feature extraction apparatus for a target object according to an embodiment of the present invention;
fig. 5 is a schematic diagram of another feature extraction apparatus for a target object according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
As shown in fig. 1, an embodiment of the present invention provides a method for extracting features of a target object, where the method may include the following steps:
step 101: obtaining a Stokes parameter map according to a target radiation image of a target object;
step 102: obtaining polarization information of the surface of the target object according to the Stokes parameter map, wherein the polarization information comprises a polarization degree and a polarization angle;
step 103: obtaining a polarization operator according to the polarization information, wherein the polarization operator is used for calculating the gray level variation and the variation direction of the image;
step 104: acquiring an unbiased image of a target object;
step 105: and performing feature extraction on the unbiased image by using a polarization operator to obtain a target feature image, wherein the target feature image represents the outline of the target object.
In the embodiment of the invention, when the characteristic of the target object is extracted, a stokes parameter map is further obtained by obtaining a radiation image of the target object, the polarization information of the polarization degree and the polarization angle of the surface of the target object can be calculated and determined through the stokes parameter map, a polarization operator for extracting the characteristic of the image can be determined through the polarization information, and therefore, the characteristic of the unbiased image is extracted through the polarization operator to obtain the target characteristic image of the target object. Therefore, the polarization operator for feature extraction of the image is determined by using the polarization information. Since the common polarization operator has no specific physical meaning when extracting the contour features of the image, all contour information in the target image is extracted. The polarization operator is determined by the polarization information, the polarization degree in the polarization information is a value between 0 and 1, and different object materials correspond to different polarization degree values, so that when the polarization operator is used for feature extraction of a target, the contour features of the corresponding target object can be accurately obtained by determining different polarization degree values, and the accuracy of feature extraction of the target object can be improved.
Optionally, according to the feature extraction method of the target object shown in fig. 1, in the embodiment of the present invention, when the stokes parameter map is determined according to the target radiation image of the target object, the stokes parameter map may be obtained by:
respectively collecting target radiation images I of target objects0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight hand circular polarization, I, for characterizing a radiation image of an objectLCPFor characterizing the left-hand circular polarization of the target radiation image.
In an embodiment of the invention, a stokes parameter map is determined by using target radiation images acquired in different polarization directions. In the scheme, the light intensity information of each polarization direction is measured and acquired by rotating the polarizing film, and the Stokes parameter is represented by the light intensity, so that the polarization state of the target radiation image is described by adopting the Stokes parameter, and the method not only has the advantages of directly measuring and acquiring the light intensity and being simple and convenient to calculate, but also is the basis for acquiring the polarization information subsequently.
Optionally, as in the feature extraction method of the target object shown in fig. 1, in the embodiment of the present invention, according to the stokes parameter map, the polarization information of the surface of the target object is determined by using a polarization degree formula and a polarization angle formula, which may be specifically implemented as follows:
and (3) calculating to obtain a polarization image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
In the embodiment of the invention, the Stokes parameter map is utilized, and the polarization information of the surface of the target object can be determined according to the formula of the polarization degree and the polarization angle. According to the scheme, the polarization operator is used for carrying out feature extraction on the image, and the polarization operator is related to the polarization degree and the polarization angle, namely the polarization operator can be formed through the corresponding polarization degree and the corresponding polarization angle, so that the polarization information of the polarization degree and the polarization angle is calculated as the basis for forming the polarization operator, and therefore guarantee is provided for accurately obtaining the feature image of the target object.
Optionally, according to the feature extraction method of the target object shown in fig. 1, when determining the target feature image, the method may be mainly implemented through processes of filtering, gradient calculation, non-maximum suppression, dual-threshold detection, and the like, and the specific implementation manner is as follows:
performing Gaussian filtering on the unbiased image to obtain a first characteristic image, wherein the first characteristic image is used for representing the image of the unbiased image after noise points are removed;
determining a first gradient amplitude and a gradient direction of the first characteristic image by utilizing finite difference calculation, wherein the first gradient amplitude is the amplitude of a partial derivative of the first characteristic image along the coordinate axis direction, the gradient direction is the direction of the first characteristic image with the fastest gradient change, and each first gradient amplitude corresponds to a pixel position;
acquiring a second gradient amplitude from the first gradient amplitude, wherein the second gradient amplitude is an amplitude image with non-edge pixels removed;
and determining the target characteristic image from the second gradient amplitude by using a preset double-threshold algorithm.
In the embodiment of the invention, when a polarization operator is used for extracting the characteristics of an unbiased image, firstly, a proper filter needs to be selected, the unbiased image is filtered by using a filtering algorithm, some pixel points generated due to noise are removed, then, a first gradient amplitude of each pixel position in the filtered image is determined by using the polarization operator, a second gradient amplitude is further determined from the first gradient amplitude by using a non-maximum suppression method to remove non-edge pixel points in the image, and finally, a polarization characteristic image is determined by setting a dual-threshold algorithm of a high threshold and a low threshold. Therefore, the scheme not only removes noise influence in a filtering mode, but also sequentially judges whether the calculated pixel point amplitude corresponds to the edge pixel of the target object outline or not. Therefore, the purpose of improving the feature extraction precision and accuracy of the target object is achieved by continuously screening the edge pixels.
Optionally, according to the manner provided by the foregoing embodiment, when determining the first gradient magnitude and the gradient direction of the first feature image by using finite difference calculation, the calculation is performed based on polarization information, and the specific implementation manner is as follows:
determining a horizontal difference of the first feature image in the horizontal direction according to the following formula:
fx(x,y)＝S1＝I0-I90
wherein f isx(x, y) is used to characterize the horizontal differential;
determining a vertical difference of the vertical direction of the first feature image according to the following formula:
fy(x,y)＝S2＝I45-I135
wherein f isy(x, y) is used to characterize the vertical differential;
using the horizontal direction difference and the vertical direction difference, determining a first gradient magnitude and a gradient direction of the first feature image according to the following formulas:
where M (x, y) is used to characterize the first gradient magnitude and θ (x, y) is used to characterize the gradient direction.
When the feature contour extraction is performed on the image by using the common polarization operator, there is no difference extraction, that is, the target information with the contour is extracted, but in practical application, not every contour information needs to be extracted. In the embodiment of the invention, the gradient amplitude and the gradient direction are determined based on the calculation of the polarization operator, so that the profile of the target object can be accurately extracted by using the characteristics of different object materials corresponding to different polarization values, and the accuracy of the extraction of the profile of the target object is improved.
As shown in fig. 2, another embodiment of the present invention further provides a method for extracting features of a target object, which may include the following steps:
step 201: a target radiation image of a target object is acquired.
In the embodiment of the invention, the transmission light intensity I of the target object with the polarization directions of 0 degree, 45 degrees, 90 degrees and 135 degrees is obtained by rotating the polaroid by using the polarization imaging measuring device0°、I45°、I90°And I135°. In the infrared polarization imaging measurement process, polarization information is shifted on image representation due to factors such as device difference in use or moving polarization generated in the imaging process, and therefore the target radiation image needs to be registered. In view of the characteristic that the target radiation image can highlight the edge contour features of the target, the target radiation image is registered in an image feature anchoring mode.
The polarization imaging measurement device may include a polarizer, an imaging lens, a CCD imaging detector, a computer, and the like, and after acquiring the target radiation image, the image may be subjected to preliminary processing by software such as Matlab.
Step 202: and calculating Stokes parameters of the target radiation image.
In the embodiment of the invention, after the target radiation image is obtained by the polarization imaging measuring device, the Stokes parameter can be obtained by calculating according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight hand circular polarization, I, for characterizing a radiation image of an objectLCPFor characterizing the left-hand circular polarization of the target radiation image.
As known from the above calculation formula of Stokes, the Stokes parameter can be directly expressed by the collected light intensities with different polarization directions, wherein S0Can be approximated to represent the incident light intensity, S1、S2And S3Respectively, to represent the difference in light intensity in two mutually orthogonal directions.
The mueller matrix is a matrix for describing the action of the polarizing devices, and the mueller matrices corresponding to different polarizing devices are different. Therefore, in the embodiment of the present invention, the vector change process corresponding to the process of obtaining the emergent light after the incident light passes through the polarizer is that the incident stokes vector obtains the emergent stokes vector after passing through the miller matrix. For example, the change in the stokes vector for an incident light after passing through a polarizer can be represented by the following relationship:
therefore, in the embodiment of the present invention, the Mueller matrix M of the known polarizing plate and the Stokes vector S of the incident light can be usedinTo obtain the Stokes vector S of the emergent lightout。
Step 203: polarization information of the target object surface is determined.
In the embodiment of the invention, after the Stokes parameter map is obtained through the Stokes formula, the polarization degree and the polarization angle can be calculated by utilizing the Stokes parameter, so that the polarization information image of the surface of the target object is obtained. Specifically, calculating the degree of polarization and the polarization angle may be performed by the following equations:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
It is known from fresnel's law that when unpolarized light is incident on a target object medium surface and reflected, the polarization state of the incident light changes to generate partially polarized light, and radiation light generated by heat radiation of the object also exhibits a polarization effect, so that the polarization states of the radiation light waves reflected by different objects or different states of the same object have significant differences. Usually, the characteristic quantities characterizing the target polarization state information mainly include a polarization degree, a polarization angle and the like, wherein the polarization degree refers to a ratio of light intensity of a polarized part in a light beam to the whole light intensity, and is a dimensionless number with a value range within a [0,1] interval. Therefore, in the embodiment of the present invention, by representing the polarization degree image of the target object by using the information of the polarization degree and the polarization angle, the advantages of the polarization image in highlighting the target contour, improving the contrast of the disguised target, and the like can be utilized to extract the contour features of the target object, thereby improving the accuracy of feature extraction on the target object.
Here, since the circularly polarized component is very weak in the polarization effect of the natural atmospheric background and the target object on the solar incidence, the circularly polarized component S is not expressed in the above calculation formulas of the degree of polarization and the polarization angle3Among others, are considered.
Step 204: and determining a polarization operator according to the polarization information.
In the embodiment of the present invention, a polarization operator needs to be obtained according to the polarization degree and the polarization angle in the polarization information, so that the characteristic profile of the target object can be further extracted by using the polarization operator. Specifically, it can be obtained by the following correspondence relationship:
therefore, as can be seen from the correspondence between the polarization degree and the polarization angle and the polarization operator, G and θ of the polarization operator can be obtained by correspondence between the polarization degree P and the polarization angle α, so that the feature extraction of the unbiased image under the corresponding detection condition can be performed based on the polarization operator.
Step 205: an unbiased image of the target object is acquired.
In the embodiment of the invention, the unbiased image is any image obtained by detection with the polarization imaging measurement device without using a polarizing plate. The invention utilizes the polarization operator to extract the characteristics of the unbiased image, and the unbiased image is the most extensive image in the prior art, thereby reflecting the wide application scene and the practical application significance of utilizing the polarization operator to extract the characteristics of the unbiased image.
Step 206: and performing Gaussian filtering on the unbiased image.
Any edge detection algorithm cannot be well processed on unprocessed original data, so in the embodiment of the invention, a proper Gaussian filter is selected firstly, namely a filter function is determined, and filtering processing is carried out on an unbiased image, so that pixel points generated by noise in the unbiased image are removed. Specifically, the data of the unbiased image is first convolved with a gaussian smooth template, and the resulting image is slightly blurred compared to the original image. Thus, the noise of a single pixel becomes almost unaffected on the gaussian-smoothed image.
The image is subjected to Gaussian filtering, which can be realized by weighting two one-dimensional Gaussian kernels twice respectively, namely, one-dimensional convolution in the X direction is performed first, and the obtained result is subjected to one-dimensional convolution in the Y direction. Of course, it can also be realized by one convolution directly through a two-dimensional Gaussian kernel, that is, a two-dimensional convolution template. For example, in the process of implementing gaussian filtering by a two-dimensional convolution template, assume that the two-dimensional gaussian function is:
the gaussian coefficient of each point in the template can be calculated by the above formula, and it is necessary to normalize, that is, the coefficient of each point is divided by the sum of all coefficients, so that the final two-dimensional gaussian template is obtained. After the template is obtained through calculation, the unbiased image and the template are convolved, wherein the convolution means that the template size area near the pixel point in the image is multiplied by the Gaussian template area, and the obtained result is the result after the point is convolved. The core meaning of convolution is the property of acquiring the template features of the target image in the original image. For example, the expression of the gaussian-filtered image G (x, y) obtained by convolving the unbiased image F (x, y) with the gaussian template H (x, y) described above can be expressed as follows:
step 207: a first gradient magnitude and gradient direction are calculated.
In the embodiment of the present invention, a first gradient magnitude and a gradient direction of the first feature image need to be determined by using finite difference calculation, where the first gradient magnitude is a magnitude of a partial derivative of the first feature image along a coordinate axis direction, a gradient square is a direction in which a gradient of the first feature image changes fastest, and no first gradient magnitude corresponds to one pixel position. The first gradient magnitude and gradient direction may specifically be calculated by:
determining a horizontal difference of the first feature image in the horizontal direction according to the following formula:
fx(x,y)＝S1＝I0-I90
wherein f isx(x, y) is used to characterize the horizontal differential;
determining a vertical difference of the vertical direction of the first feature image according to the following formula:
fy(x,y)＝S2＝I45-I135
wherein the content of the first and second substances,fy(x, y) is used to characterize the vertical differential;
using the horizontal direction difference and the vertical direction difference, determining a first gradient magnitude and a gradient direction of the first feature image according to the following formulas:
where M (x, y) is used to characterize the first gradient magnitude and θ (x, y) is used to characterize the gradient direction.
According to the calculation process, the first gradient amplitude and the gradient direction of the first characteristic image are calculated by using the polarization operator formed by the polarization information, so that the characteristic that objects made of different materials correspond to different polarization degrees is used for accurately extracting the characteristic outline of the target object, and the purpose of improving the accuracy of extracting the characteristic of the target object is achieved.
Step 208: a second gradient magnitude is obtained from the first gradient magnitude.
In embodiments of the present invention, non-maxima suppression of gradient amplitudes is required. The larger the element in the image gradient magnitude matrix, the larger the gradient value of the point in the image, but it does not indicate that the point is an edge. The non-maximum suppression is to eliminate non-edge pixels, thin ridge bands in the amplitude image and only reserve points with the maximum local amplitude change, so that the characteristics of edges, contours and the like can be determined. In short, a local maximum of the pixel points is found, and then the gray value corresponding to the non-maximum point is set to 0, so that a majority of non-edge pixel points are eliminated.
Step 209: edge detection is performed using a dual threshold algorithm.
In the embodiment of the invention, the corresponding relation between the polarization degree characteristic of the surface of the target object and the material parameter of the surface of the target is considered, the edge contour characteristic extraction method based on the target polarization information has advantages for target extraction in a complex background, and a high threshold and a low threshold in a dual-threshold detection algorithm can be set according to the polarization characteristic of the surface material of the target object, so that the accuracy of target edge contour characteristic extraction is improved.
The method determines whether the pixel point corresponding to the second gradient amplitude is an edge pixel point by judging the size relationship between the second gradient amplitude and a preset double threshold value. Specifically, this can be achieved by:
determining a first threshold value and a second threshold value according to the polarization degree of the target object, wherein the second threshold value is larger than the first threshold value;
for each pixel in the bias-unbiased image, performing:
if the second gradient amplitude corresponding to the pixel is larger than or equal to a second threshold value, determining the pixel as an edge pixel of the target object;
if the second gradient amplitude corresponding to the pixel is larger than the first threshold and smaller than the second threshold, and the second gradient amplitude corresponding to at least one pixel adjacent to the pixel in the unbiased image is larger than or equal to the second threshold, determining the pixel as an edge pixel of the target object;
it can be seen that if the second gradient magnitude at a certain pixel location exceeds the second threshold, the pixel will be determined to be an edge pixel. If the second gradient magnitude at a pixel location is less than the first threshold, the pixel will be determined to be a non-edge pixel and thus excluded. If the second amplitude of a pixel location is between two thresholds, the pixel is determined to be an edge pixel only if it can be connected to a pixel for which the second gradient amplitude is greater than the second threshold. When the first threshold value and the second threshold value are determined, the value ranges of the double threshold values can be correspondingly set according to the polarization degree corresponding to the material of the target object, so that the contour information of the target object can be accurately extracted and the contour information of other objects can be excluded during feature extraction, and therefore the accuracy of feature extraction on the target object can be improved.
In the embodiment of the invention, when the edge is determined, the pixel points with larger gradient amplitude are likely to be edges, but a determined value is not available to determine which edge pixel points should be extracted. The scheme can distinguish different types by utilizing polarization informationThe characteristic of the material sets a high threshold (second threshold) and a low threshold (first threshold) for edge pixel detection. With these two thresholds, three edge images N can be obtainedH(x,y)、NL(x, y) and NM(x, y) wherein NH(x, y) is obtained by being larger than the high threshold, the noise is less interfered, and the second gradient amplitude is closer to the true edge, so that the second gradient amplitude can be directly larger than the high threshold NHAnd determining the pixel point of (x, y) as an edge pixel. In contrast, NL(x, y) is derived from less than the low threshold, which is clearly very unlikely to be an edge pixel, thereby excluding this pixel. N is a radical ofM(x, y) is a pixel point where the second gradient amplitude is between the high threshold and the low threshold, and it is obvious that the pixel point may be an edge pixel point or a non-edge pixel point, that is, only N is presentH(x,y)、NLThe (x, y) edge image is inevitably interrupted or not closed, and the dual-threshold algorithm judges NMWhether the (x, y) pixel points can find the contour edge which can be connected with the (x, y) pixel points in the eight surrounding pixel points or not is determined, so that whether the pixel points are edge pixels or not is determined, and when the contour feature of the target object is extracted, a target feature image with a complete contour can be obtained.
For example, for a vehicle which may contain a carving of a wooden material and a metal material in an image, in an actual requirement, when only the contour feature of the vehicle needs to be extracted, a high threshold and a low threshold may be set according to polarization information respectively corresponding to the metal material and the wooden material, and the wooden material is excluded in a manner of being smaller than the low threshold, so that the edge contour feature of the vehicle is accurately obtained.
As shown in fig. 3 and 4, the embodiment of the present invention provides an apparatus in which the feature extraction device of the target object is located and a feature extraction device of the target object. The device embodiments may be implemented by software, or by hardware, or by a combination of hardware and software. From a hardware level, as shown in fig. 3, a hardware structure diagram of a device in which a feature extraction apparatus of a target object is located is provided for an embodiment of the present invention, and in addition to the processor, the memory, the network interface, and the nonvolatile memory shown in fig. 3, the device in which the apparatus is located in the embodiment may also include other hardware, such as a forwarding chip responsible for processing a packet, in general. Taking a software implementation as an example, as shown in fig. 4, as a logical apparatus, the apparatus is formed by reading a corresponding computer program instruction in a non-volatile memory into a memory by a CPU of a device in which the apparatus is located and running the computer program instruction. As shown in fig. 4, an embodiment of the present invention provides a feature extraction apparatus for a target object, including: a first determining module 401, a second determining module 402, a third determining module 403, an obtaining module 404 and a fourth determining module 405;
a first determining module 401, configured to obtain a Stokes parameter map according to a target radiation image of a target object;
a second determining module 402, configured to obtain polarization information of the surface of the target object according to the Stokes parameter map determined by the first determining module 401, where the polarization information includes a polarization degree and a polarization angle;
a third determining module 403, configured to obtain a polarization operator according to the polarization information determined by the second determining module 402, where the polarization operator is used to calculate a gray scale variation and a variation direction of the image;
an obtaining module 404, configured to obtain an unbiased image of a target object;
a fourth determining module 405, configured to perform feature extraction on the unbiased image acquired by the acquiring module 404 by using the polarization operator determined by the third determining module 403, so as to obtain a target feature image, where the target feature image represents a contour of a target object.
Optionally, as shown in fig. 4, the first determining module 401 is configured to perform the following operations:
respectively collecting target radiation images I of target objects0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight hand circular polarization, I, for characterizing a radiation image of an objectLCPFor characterizing the left-hand circular polarization of the target radiation image.
Optionally, according to the above-described feature extraction apparatus of the target object, the second determining module 402 is configured to perform the following operations:
and (3) calculating and obtaining a polarization information image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
Alternatively, based on the feature extraction apparatus of the target object in fig. 4, as shown in fig. 5, the fourth determining module 405 includes: a filtering unit 4051, a calculating unit 4052, an obtaining unit 4053, and a determining unit 4054;
the filtering unit 4051 is configured to perform gaussian filtering on the unbiased image to obtain a first feature image, where the first feature image is used to represent an image of the unbiased image after removing the noise point;
a calculating unit 4052, configured to determine, by using finite difference calculation, a first gradient amplitude and a gradient direction of the first feature image determined by the filtering unit 4051, where the first gradient amplitude is an amplitude of a partial derivative of the first feature image along a coordinate axis direction, the gradient direction is a direction in which a gradient of the first feature image changes fastest, and each first gradient amplitude corresponds to a pixel position;
an obtaining unit 4053, configured to obtain a second gradient amplitude from the first gradient amplitude calculated by the calculating unit 4052, where the second gradient amplitude is an amplitude image from which non-edge pixels are removed;
the determining unit 4054 is configured to determine the target feature image from the second gradient amplitude acquired by the acquiring unit 4053 by using a preset double-threshold algorithm.
Alternatively, as shown in fig. 5, the computing unit 4052 is configured to perform the following operations:
determining a horizontal difference of the first feature image in the horizontal direction according to the following formula:
fx(x,y)＝S1＝I0-I90
wherein f isx(x, y) is used to characterize the horizontal differential;
determining a vertical difference of the vertical direction of the first feature image according to the following formula:
fy(x,y)＝S2＝I45-I135
wherein f isy(x, y) is used to characterize the vertical differential;
using the horizontal direction difference and the vertical direction difference, determining a first gradient magnitude and a gradient direction of the first feature image according to the following formulas:
where M (x, y) is used to characterize the first gradient magnitude and θ (x, y) is used to characterize the gradient direction.
The embodiment of the present invention further provides a device for extracting features of a target object, which is characterized by including: at least one memory and at least one processor;
the at least one memory to store a machine readable program;
the at least one processor is configured to invoke the machine-readable program to perform the feature extraction method for the target object according to any embodiment of the present invention.
Embodiments of the present invention further provide a computer-readable medium, where computer instructions are stored, and when executed by a processor, the computer instructions cause the processor to execute the feature extraction method for the target object in any embodiment of the present invention. Specifically, a method or an apparatus equipped with a storage medium on which a software program code that realizes the functions of any of the above-described embodiments is stored may be provided, and a computer (or a CPU or MPU) of the method or the apparatus is caused to read out and execute the program code stored in the storage medium.
In this case, the program code itself read from the storage medium can realize the functions of any of the above-described embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.
Examples of the storage medium for supplying the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (e.g., CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD + RW), a magnetic tape, a nonvolatile memory card, and a ROM. Alternatively, the program code may be downloaded from a server computer via a communications network.
Further, it should be clear that the functions of any one of the above-described embodiments can be implemented not only by executing the program code read out by the computer, but also by performing a part or all of the actual operations by an operation method or the like operating on the computer based on instructions of the program code.
In summary, the feature extraction method, apparatus and computer readable medium for a target object provided in the embodiments of the present invention at least have the following advantages:
1. in the embodiment of the invention, when the characteristic of the target object is extracted, a stokes parameter map is further obtained by obtaining a radiation image of the target object, the polarization information of the polarization degree and the polarization angle of the surface of the target object can be calculated and determined through the stokes parameter map, a polarization operator for extracting the characteristic of the image can be determined through the polarization information, and therefore, the characteristic of the unbiased image is extracted through the polarization operator to obtain the target characteristic image of the target object. Therefore, the polarization operator for feature extraction of the image is determined by using the polarization information. Since the common polarization operator has no specific physical meaning when extracting the contour features of the image, all contour information in the target image is extracted. The polarization operator is determined by the polarization information, the polarization degree in the polarization information is a value between 0 and 1, and different object materials correspond to different polarization degree values, so that when the polarization operator is used for feature extraction of a target, the contour features of the corresponding target object can be accurately obtained by determining different polarization degree values, and the accuracy of feature extraction of the target object can be improved.
2. In an embodiment of the invention, a stokes parameter map is determined by using target radiation images acquired in different polarization directions. In the scheme, the light intensity information of each polarization direction is measured and acquired by rotating the polarizing film, and the Stokes parameter is represented by the light intensity, so that the polarization state of the target radiation image is described by adopting the Stokes parameter, and the method not only has the advantages of directly measuring and acquiring the light intensity and being simple and convenient to calculate, but also is the basis for acquiring the polarization information subsequently.
3. In the embodiment of the invention, the Stokes parameter map is utilized, and the polarization information of the surface of the target object can be determined according to the formula of the polarization degree and the polarization angle. According to the scheme, the polarization operator is used for carrying out feature extraction on the image, and the polarization operator is related to the polarization degree and the polarization angle, namely the polarization operator can be formed through the corresponding polarization degree and the corresponding polarization angle, so that the polarization information of the polarization degree and the polarization angle is calculated as the basis for forming the polarization operator, and therefore guarantee is provided for accurately obtaining the feature image of the target object.
4. In the embodiment of the invention, when a polarization operator is used for extracting the characteristics of an unbiased image, firstly, a proper filter needs to be selected, the unbiased image is filtered by using a filtering algorithm, some pixel points generated due to noise are removed, then, a first gradient amplitude of each pixel position in the filtered image is determined by using the polarization operator, a second gradient amplitude is further determined from the first gradient amplitude by using a non-maximum suppression method to remove non-edge pixel points in the image, and finally, a polarization characteristic image is determined by setting a dual-threshold algorithm of a high threshold and a low threshold. Therefore, the scheme not only removes noise influence in a filtering mode, but also sequentially judges whether the calculated pixel point amplitude corresponds to the edge pixel of the target object outline or not. Therefore, the purpose of improving the feature extraction precision and accuracy of the target object is achieved by continuously screening the edge pixels.
5. When the feature contour extraction is performed on the image by using the common polarization operator, there is no difference extraction, that is, the target information with the contour is extracted, but in practical application, not every contour information needs to be extracted. In the embodiment of the invention, the gradient amplitude and the gradient direction are determined based on the calculation of the polarization operator, so that the profile of the target object can be accurately extracted by using the characteristics of different object materials corresponding to different polarization values, and the accuracy of the extraction of the profile of the target object is improved.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for extracting a feature of a target object, comprising:
obtaining a Stokes parameter map according to a target radiation image of a target object;
obtaining polarization information of the surface of the target object according to the Stokes parameter map, wherein the polarization information comprises a polarization degree and a polarization angle;
obtaining a polarization operator according to the polarization information, wherein the polarization operator is used for calculating the gray level variation and the variation direction of the image;
acquiring an unbiased image of the target object;
and performing feature extraction on the unbiased image by using the polarization operator to obtain a target feature image, wherein the target feature image represents the outline of the target object.
2. The method of claim 1, wherein obtaining a Stokes reference map from a target radiation image of a target object comprises:
respectively collecting target radiation images I of the target object0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight-hand circular polarization for characterizing the target radiation image, said ILCPFor characterizing the left-hand circular polarization of the target radiation image.
3. The method of claim 2, wherein obtaining polarization information of the target object surface from the Stokes parameter map comprises:
calculating and obtaining a polarization information image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
4. The method according to any one of claims 1 to 3, wherein the performing feature extraction on the unbiased image by using the polarization operator to obtain a target feature image comprises:
performing Gaussian filtering on the unbiased image to obtain a first characteristic image, wherein the first characteristic image is used for representing the image of the unbiased image after noise points are removed;
determining a first gradient amplitude and a gradient direction of the first characteristic image by utilizing finite difference calculation, wherein the first gradient amplitude is the amplitude of a partial derivative of the first characteristic image along the coordinate axis direction, the gradient direction is the direction of the first characteristic image with the fastest gradient change, and each first gradient amplitude corresponds to one pixel position;
acquiring a second gradient amplitude from the first gradient amplitude, wherein the second gradient amplitude is an amplitude image with non-edge pixels removed;
and determining a target characteristic image from the second gradient amplitude by using a preset double-threshold algorithm.
5. The method of claim 4, wherein determining the first gradient magnitude and gradient direction of the first feature image using finite difference computation comprises:
determining a horizontal difference of the first characteristic image in the horizontal direction according to the following formula:
fx(x,y)＝S1＝I0-I90
wherein f isx(x, y) for characterizing the horizontal difference;
determining a vertical difference of the vertical direction of the first feature image according to the following formula:
fy(x,y)＝S2＝I45-I135
wherein f isy(x, y) for characterizing the vertical differential;
determining a first gradient magnitude and a gradient direction of the first feature image according to the following formulas by using the horizontal difference and the vertical difference:
wherein M (x, y) is used to characterize the first gradient magnitude and θ (x, y) is used to characterize the gradient direction.
6. A feature extraction device of a target object, characterized by comprising: the device comprises a first determining module, a second determining module, a third determining module, an obtaining module and a fourth determining module;
the first determination module is used for obtaining a Stokes parameter map according to a target radiation image of a target object;
the second determining module is configured to obtain polarization information of the surface of the target object according to the Stokes parameter map determined by the first determining module, where the polarization information includes a polarization degree and a polarization angle;
the third determining module is configured to obtain a polarization operator according to the polarization information determined by the second determining module, where the polarization operator is used to calculate a gray scale variation and a variation direction of an image;
the acquisition module is used for acquiring an unbiased image of the target object;
the fourth determining module is configured to perform feature extraction on the unbiased image acquired by the acquiring module by using the polarization operator determined by the third determining module to obtain a target feature image, where the target feature image represents a contour of the target object.
7. The apparatus of claim 6,
the first determining module is configured to perform the following operations:
respectively collecting target radiation images I of the target object0°、I45°、I90°And I135°Wherein the target radiation image I0°、I45°、I90°And I135°Obtained at polarization directions of 0 °, 45 °, 90 ° and 135 °, respectively;
calculating the Stokes parameters of the target radiation image according to the following formula:
wherein S is used for representing Stokes parameters, S0、S1、S2And S3Are all components of the Stokes parameter S, IRCPRight-hand circular polarization for characterizing the target radiation image, said ILCPFor characterizing the left-hand circular polarization of the target radiation image.
8. The apparatus of claim 7, wherein the apparatus is a portable electronic device
The second determining module is configured to perform the following operations:
calculating and obtaining a polarization information image of the surface of the target object according to the following formula by using the Stokes parameter map:
wherein p is used for characterizing the polarization degree of the polarization information image, and α is used for characterizing the polarization angle of the polarization information image.
9. A feature extraction device of a target object, characterized by comprising: at least one memory and at least one processor;
the at least one memory to store a machine readable program;
the at least one processor, configured to invoke the machine readable program, to perform the method of any of claims 1 to 5.
10. A computer readable medium having stored thereon computer instructions which, when executed by a processor, cause the processor to perform the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011061687.1A CN112163586A (en) | 2020-09-30 | 2020-09-30 | Feature extraction method and device of target object and computer readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011061687.1A CN112163586A (en) | 2020-09-30 | 2020-09-30 | Feature extraction method and device of target object and computer readable medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112163586A true CN112163586A (en) | 2021-01-01 |
Family
ID=73860935
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011061687.1A Pending CN112163586A (en) | 2020-09-30 | 2020-09-30 | Feature extraction method and device of target object and computer readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112163586A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022236492A1 (en) * | 2021-05-08 | 2022-11-17 | 华为技术有限公司 | Image processing method and apparatus |
-
2020
- 2020-09-30 CN CN202011061687.1A patent/CN112163586A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022236492A1 (en) * | 2021-05-08 | 2022-11-17 | 华为技术有限公司 | Image processing method and apparatus |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022007431A1 (en) | Positioning method for micro qr two-dimensional code | |
| Schettini et al. | Underwater image processing: state of the art of restoration and image enhancement methods | |
| CN104143185B (en) | A kind of stain method for detecting area | |
| WO2020018359A1 (en) | Three-dimensional living-body face detection method, face authentication recognition method, and apparatuses | |
| JP2020536457A (en) | Image processing methods and devices, electronic devices, and computer-readable storage media | |
| Yu et al. | Efficient patch-wise non-uniform deblurring for a single image | |
| EP1958158B1 (en) | Method for detecting streaks in digital images | |
| WO2021217642A1 (en) | Infrared image processing method and apparatus, and movable platform | |
| FR2904716A1 (en) | APPARATUS AND METHOD FOR REMOVING LINE NOISE | |
| CN110660072B (en) | Method and device for identifying straight line edge, storage medium and electronic equipment | |
| CN113034525A (en) | Image edge detection method, device and equipment | |
| CN110378902B (en) | Scratch detection method under high noise background | |
| WO2011097936A1 (en) | Fingerprint image enhancement method | |
| CN112163587A (en) | Feature extraction method and device of target object and computer readable medium | |
| Deshpande et al. | A novel modified cepstral based technique for blind estimation of motion blur | |
| CN114332026A (en) | Visual detection method and device for scratch defects on surface of nameplate | |
| WO2021195873A1 (en) | Method and device for identifying region of interest in sfr test chart image, and medium | |
| CN112163586A (en) | Feature extraction method and device of target object and computer readable medium | |
| Wang et al. | Blob reconstruction using unilateral second order Gaussian kernels with application to high-ISO long-exposure image denoising | |
| JP2003016440A5 (en) | ||
| Shuchun et al. | Preprocessing for stereo vision based on LOG filter | |
| Liu et al. | Deblurring saturated night image with function-form kernel | |
| Reddy et al. | Canny scale edge detection | |
| Wang et al. | Digital image forgery detection based on the consistency of defocus blur | |
| Tiwari et al. | Certain investigations on motion blur detection and estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |