WO2020118538A1

WO2020118538A1 - Image collection method and device

Info

Publication number: WO2020118538A1
Application number: PCT/CN2018/120445
Authority: WO
Inventors: 王星泽; 闫静; 舒远
Original assignee: 合刃科技（深圳）有限公司
Priority date: 2018-12-12
Filing date: 2018-12-12
Publication date: 2020-06-18
Also published as: CN111344711A

Abstract

An image collection method comprising: collecting two or more material images with one or more optical parameters, wherein one of the material images corresponds to a parameter interval of the optical parameters, and the optical parameter at least comprises at least one of a polarization direction and a spectrum (S102); and carrying out image fusion on the collected material images to obtain a target image for feature recognition (S104). Further disclosed is an image collection device. With this method, the amount of information in the collected image can be maximized, thereby improving the accuracy of image recognition.

Description

Image acquisition method and device

Technical field

The invention relates to the field of image processing, in particular to an image acquisition method and device.

Background technique

In the traditional technology, bar code and two-dimensional code recognition technology has the functions of data acquisition, automatic recognition, fast response, etc. It has the characteristics of convenient production, low cost, convenient use, perfect technology, etc., identity verification in daily life, online payment, Information identification and other scenarios provide users with a very convenient human-computer interaction experience.

The barcode and two-dimensional code recognition technology in the traditional technology includes two links, one is the collection and acquisition of the image information of the barcode and the two-dimensional code, and the second is the recognition of the image information of the barcode and the two-dimensional code. However, if the barcode or QR code image is unclear during the collection and acquisition of barcode and QR code image information, it will greatly affect the accuracy of barcode and QR code image recognition.

For example, the bar code or two-dimensional code pattern printed on the high reflective arc surface or transparent/translucent material, the brightness of the bar code or two-dimensional code image on the high reflective arc surface is very uneven due to the problem of the light source angle, and may even appear Half bright and half dark yin and yang codes. In this case, the bar code or two-dimensional code image collected by the ordinary barcode reader is very unclear, and the amount of information is seriously lost, so that the subsequent recognition process will either fail to recognize specific information, or will Identify the wrong information.

At present, in view of the problem of noise caused by uneven illumination of the bar code or two-dimensional code image during shooting, although the median filter principle is used in the traditional technology, the bar code and the two-dimensional code image are subjected to noise reduction processing, combined with adaptive The brightness binary processing method solves the problem of uneven illumination of bar code and two-dimensional code images, but for the detection of two-dimensional codes of highly reflective arc surfaces or transparent/translucent materials, the two-dimensional codes are partially caused by perforation and staining. In the case of damage, the identification still has the problem of low accuracy or unrecognizable.

Summary of the invention

Based on this, in order to increase the information amount of the feature information contained in the image to be recognized obtained by image collection, and to solve the problem of image recognition in the prior art due to the blurring of the illumination angle and the image to be recognized in the image collection process, the recognition accuracy rate is relatively high Low technical problems, especially proposed an image acquisition method.

An image acquisition method, including:

Acquiring two or more material images under one or more optical parameters, one material image corresponding to a parameter interval of the optical parameters, the optical parameters including at least one of polarization direction and spectrum;

Perform image fusion on the collected material images to obtain a target image for feature recognition.

In one of the embodiments, the acquisition of two or more material images under one or more optical parameters includes:

Acquiring two or more material images through a hyperspectral image acquisition element and/or a polarized light image acquisition element, the optical parameter corresponding to the hyperspectral image acquisition element is a spectrum, and the optical parameter corresponding to the polarized light image acquisition element Is the polarization direction.

In one of the embodiments, the image fusion of the collected material images to obtain a target image for feature recognition includes:

Performing wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum;

Merge sub-band images belonging to the same sub-band spectrum into sub-band target images according to a preset fusion strategy;

The sub-band target image corresponding to each sub-band spectrum is subjected to inverse wavelet transform to obtain a target image for feature recognition.

In one of the embodiments, the fusion of sub-band images belonging to the same sub-band spectrum into a sub-band target image according to a preset fusion strategy includes:

Divide subband images belonging to the same subband spectrum into two or more regions;

The area images of the subband images of the same subband spectrum in each of the areas are fused according to a preset fusion strategy to obtain a subband target image.

In one of the embodiments, the fusion of the sub-band images of the same sub-band spectrum in each of the regions according to a preset fusion strategy includes:

Calculating the area variance of the area images corresponding to the sub-band images in each of the areas;

Calculate the similarity of images in each area according to the regional variance;

According to the size of the similarity, the region images are fused by selection or weighted merging.

In one of the embodiments, after obtaining the target image for feature recognition, the method further includes:

Perform two-dimensional code/barcode identification on the target image.

In addition, in order to improve the information amount of the feature information contained in the image to be recognized obtained by image collection, and to solve the problem of image recognition in the prior art due to the blurring of the illumination angle and the image to be recognized in the image collection process, the recognition accuracy is low The technical problem, especially proposed an image acquisition device.

An image acquisition device, including:

The material image acquisition module collects two or more material images under one or more optical parameters, one material image corresponds to a parameter interval of the optical parameters, and the optical parameters include at least the polarization direction and the spectrum At least one of

The image fusion module performs image fusion on the collected material images to obtain a target image for feature recognition.

In one of the embodiments, the material image acquisition module is used to acquire two or more material images through a hyperspectral image acquisition element and/or a polarized light image acquisition element, the optical corresponding to the hyperspectral image acquisition element The parameter is the spectrum, and the optical parameter corresponding to the polarized light image acquisition element is the polarization direction.

In one of the embodiments, the image fusion module is further used to perform wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum; sub-bands belonging to the same subband spectrum The band images are fused into sub-band target images according to a preset fusion strategy; the sub-band target images corresponding to each sub-band spectrum are subjected to inverse wavelet transform to obtain target images for feature recognition.

In one of the embodiments, the image fusion module is further used to divide the sub-band images belonging to the same sub-band spectrum into two or more regions; the sub-band images of the same sub-band spectrum are in each of the regions The images of the region are merged according to the preset fusion strategy to obtain the sub-band target image.

In one of the embodiments, the image fusion module is further used to calculate the regional variance of the regional images corresponding to the sub-band images in each of the regions; calculate the similarity of each regional image according to the regional variance; The size of the similarity is fused with the area images by means of selection or weighted merging. The implementation of the embodiments of the present invention will have the following beneficial effects:

After using the above image acquisition method and device, for the glare interference caused by the illumination direction, the material images in different polarization directions can be collected and then fused to avoid the reflection of polarized light generated in the specific illumination direction to the captured image. Interference of feature information, and the glare covered feature information collected in other polarization directions is used as a supplement to the final collected image to be recognized; at the same time, for the transparent or semi-transparent image to be recognized, or the background color fades or Contaminated and unclear pictures to be recognized can be obtained by collecting material images under different spectral bands, and extracting the feature information contained in the material images under each spectrum band, and adding to the final collected image to be recognized through image fusion . This makes the collected image to be recognized contain the feature information in the picture to be recognized to the greatest extent, which further improves the accuracy of subsequent image recognition.

BRIEF DESCRIPTION

The drawings required in the embodiments or the description of the prior art will be briefly introduced below.

In order to more clearly explain the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only These are some embodiments of the present application. For those of ordinary skill in the art, without paying any creative work, other drawings can be obtained based on these drawings.

FIG. 1 is an architecture diagram of an image acquisition system in an embodiment;

2 is a flowchart of an image acquisition method in another embodiment;

3 is a schematic diagram of fusing material images with multiple polarization directions in an embodiment;

FIG. 4 is a schematic diagram of fusing material images of multiple spectral bands in an embodiment;

5 is a flowchart of an image fusion process based on wavelet transform in an embodiment;

6 is a schematic diagram of the process of regional image fusion in an embodiment;

7 is a schematic diagram of an image acquisition device according to an embodiment;

FIG. 8 is a schematic diagram of the composition of a computer system running the foregoing image acquisition method in one embodiment.

detailed description

The technical solutions in the embodiments of the present invention will be described clearly and completely in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by a person of ordinary skill in the art without making creative efforts fall within the protection scope of the present invention.

In order to solve the image recognition technology in the prior art, especially barcodes and two-dimensional codes, such as QR codes (English: Quick Response Code, Chinese: quick response code) in the recognition process, due to the uneven illumination direction, or The effect of the transparency of the bar code/two-dimensional code picture itself, which leads to the low quality and unobvious characteristics of the image to be recognized, which in turn leads to the technical problem of low image recognition accuracy. The present invention specifically proposes an image acquisition method and device And an image acquisition system for implementing the image acquisition method.

In one embodiment, the implementation of the image acquisition method proposed by the present invention is based on the image acquisition system shown in FIG. 1. Compared with the conventional two-dimensional code or barcode reader, the selection of the photosensitive element is different. In the technology, a camera or other photoelectric sensors such as CMOS and CCD are used to capture a two-dimensional code or barcode image to complete the collection of the image to be recognized. In the image acquisition system of this embodiment, a hyperspectral image acquisition element and/or The polarized light image acquisition element is used as a sensor device for image acquisition, in which the hyperspectral image acquisition element can acquire multi-spectral images under multiple spectral bands, and can extract corresponding spectral band images under a specific spectral band; while polarized light image acquisition The element can acquire multiple images in multiple polarization directions, or one image per specific polarization direction.

In this embodiment, the image acquisition system may further include an image processing chip, which may merge material images of multiple spectral intervals collected by the hyperspectral image acquisition element into images to be recognized, or may integrate multiple polarized light image acquisition elements. The material images in each irradiation direction are fused into the image to be recognized, or the material images collected by the hyperspectral image collection element and the polarized light image collection element can be fused into the image to be recognized.

In other embodiments, the material images collected by the hyperspectral image collection element and/or the material images collected by the polarized light image collection element can also be directly sent to an external computer device, and the computer device can process and fuse these material images To get the image to be recognized.

Specifically, as shown in FIG. 2, the image acquisition method based on the image acquisition system includes:

Step S102: Acquire two or more material images under one or more optical parameters, one material image corresponding to a parameter interval of the optical parameters, the optical parameters at least including at least one of the polarization direction and the spectrum One kind.

In one embodiment, the image acquisition system running this method only includes a polarized light image acquisition element. The optical parameter on which image acquisition depends is a single optical parameter of the polarization direction, and the angle of different polarization directions corresponds to the polarization direction. The parameter interval, the images at multiple polarization angles collected by the polarized light image collection element are the material images used for later image fusion. For example, in the coordinate system of the polarized image acquisition element itself, with an angle range of 45 degrees as a parameter interval, 8 pictures can be acquired as material images for subsequent image fusion.

The significance of collecting material pictures in multiple polarization directions through the polarized light image acquisition element is that when the image to be recognized is attached to a smooth medium surface, under a certain illumination angle (called Brewster angle, which is related to the refractive index of the substance) Relevant), the glare formed by reflection is polarized light, and the strong distinction between light and dark caused by glare is the biggest interference in the acquisition process of the identification image. In this case, by collecting material pictures in multiple polarization directions, the polarized light causing glare can be avoided, and material images in other polarization directions are fused to obtain an image to be recognized without glare interference.

In another embodiment, the image acquisition system running this method only includes a hyperspectral image acquisition element. The optical parameter on which image acquisition depends is a single optical parameter of frequency spectrum, and the intervals of different wavelength lengths correspond to the parameters of the frequency spectrum. Intervals, images under multiple spectral bands (wavelength intervals) collected by hyperspectral image acquisition elements are material images used for later image fusion. For example, a specific wavelength length can be used as a parameter interval, a predetermined number of pictures can be collected as a material image for subsequent image fusion, or a non-specific length wavelength interval such as ultraviolet, visible, near-infrared, and other light can be collected as a parameter interval. The image collected by the hyperspectral image acquisition element is a hyperspectral image, and images under corresponding parameter intervals can be extracted as material images, respectively.

The significance of collecting material pictures in multiple polarization directions through the hyperspectral image acquisition element is that when the color and the background color of the image to be recognized are similar and unclear, for example, the white background material of the QR code/barcode changes due to long-term use. Yellow, or the black background due to the attachment of dirt makes it difficult to distinguish the original white background and the characteristic lines representing the features under visible light, the imaging situation of the image to be recognized at each wavelength interval can be collected by the hyperspectral image acquisition element. The contrast between the background color of the recognition image and the material images of the characteristic part in different wavelength ranges (that is, the images corresponding to the spectral bands of the corresponding wavelength range) are different, so using a certain image fusion method, the material image with better contrast can be used as the basic fusion Obtain a high-quality image to be recognized.

In another embodiment, the image acquisition system running this method may include both polarized light image acquisition elements and hyperspectral image acquisition elements, which means that when the image acquisition system acquires multiple material pictures, not only different polarizations are considered The direction also considers different spectral bands. For example, material image 1 can be collected in polarization direction 1 and band 1, material image 2 can be collected in polarization direction 1 and band 2, and material image 3 can be collected in polarization direction 2 and band 1. When there are m parameter sections in the polarization direction and n parameter sections in the wavelength section, theoretically m×n material images can be collected, but in practical applications, they can be screened according to the actual situation and the collected material images Filtering is performed, and when the computing power is adapted, a suitable material image is selected from the set of m×n material images for image fusion.

Further, in this embodiment, the image acquisition system may also integrate a microstructure array for image acquisition, that is, a chip combined with a deep learning artificial intelligence algorithm. The microarray structure may be a polarized microarray, a spectral microarray, or both. The combined microarray can effectively collect optical information in multiple dimensions including light intensity, phase, spectrum, incidence, and polarization direction. This type of optical chip is highly integrated, and it is also very small and light.

Step S104: Perform image fusion on the collected material images to obtain a target image for feature recognition.

Image fusion is the process of extracting the image data collected by multiple source channels on the same target through image processing and computer technology to maximize the beneficial information in the respective channels and finally synthesizing high-quality images. In this embodiment, it is the process of extracting information such as edges, contours, textures and the like in multiple material images and integrating them into a high-quality image that can reflect the characteristics of the two-dimensional code/barcode to the maximum extent.

Referring to FIG. 3, FIG. 3 shows the process of fusing material images collected by the polarized light image collecting element when the optical parameter is the polarization direction. For two-dimensional code/barcode images that are attached to non-horizontal planes, due to the problem of the illumination angle, the Brewster effect may be generated on the surface of the two-dimensional code/barcode image, thereby generating polarized light that causes glare effects. As a result, a large light spot appears in the picture to be recognized collected by the conventional image collection device, and the characteristic part in the original picture is obscured. However, by collecting multiple material pictures in multiple polarization directions through a polarized light image acquisition element, a partially clear picture can be obtained in some polarization directions, while in other polarization directions, clear pictures of other parts can be collected . These partially clear material pictures are stitched together through image fusion to obtain an image to be recognized with good illumination.

Therefore, the polarized light image collection element collects the material images at multiple polarization angles and then fuses them to avoid the glare generated by the illumination angle from interfering with the collected images with identification, and to make the fused image to be identified more clear.

Referring again to FIG. 4, FIG. 4 shows the process of fusing the material images collected by the hyperspectral image collection element when the optical parameter is the spectrum. For the background color or the characteristic part of the picture itself due to fading, pollution, etc., or the blurry two-dimensional code/barcode image caused by transparency or translucency, through the traditional image acquisition device (only the visible light is collected (Spectral optical signals such as optical signals of the frequency spectrum or optical signals of the spectrum where infrared rays are located) collected images to be identified, in the visible light spectrum range, blurred images to be identified, as shown by the naked eye, will be collected. This poses difficulties for subsequent image recognition.

By collecting multiple material pictures in multiple spectral bands, ie multiple wavelength intervals, through hyperspectral image acquisition elements, pictures with relatively high contrast can be obtained in certain wavelength intervals, or pictures with high contrast in some areas, while At other wavelength intervals, images with relatively high contrast in other parts are obtained. By fusing these high-contrast material pictures through image fusion, an image with good contrast can be obtained, which is convenient for subsequent image recognition.

Therefore, the hyperspectral image acquisition component collects the material images in the multi-spectral spectrum band and then fuses it to avoid the blurry situation caused by the transparency or translucency of the picture to be recognized, fading, pollution and other reasons, making the fusion The image to be recognized is clearer.

In this embodiment, the image fusion of the material images is mainly based on the wavelet transform method. The inherent characteristics of wavelet transform make it have the following advantages in image processing: perfect reconstruction ability to ensure that the signal has no information loss and redundant information during the decomposition process; the image is decomposed into a combination of average image and detail image, which respectively represent The different structure of the image, so it is easy to extract the structural information and detailed information of the original image; wavelet analysis provides a selective image that matches the direction of the human visual system. The frequency domain analysis of an image through wavelet transform can decompose a material image into multiple images in the frequency domain, each image corresponds to the corresponding subband spectrum, and the decomposed image corresponding to the subband spectrum is Subband image.

Specifically, as shown in FIG. 5, the image fusion method by wavelet decomposition includes the following steps:

Step S202: Perform wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum.

The wavelet decomposition of the material image is to select an appropriate wavelet base to decompose the material image into multiple sub-images in the frequency domain. The frequency of each wavelet base is the corresponding subband spectrum, and each sub-image is the corresponding subband spectrum. Subband image. In this embodiment, the selection of the wavelet basis function is not limited to a specific method, and the scale and decomposition coefficient of wavelet decomposition are also not limited, and can be selected according to actual conditions.

In wavelet decomposition, the low-band subband image corresponds to the background and background of the image, while the high-band subband image corresponds to the edge and texture features of the image.

For example, if the scale is selected as 1, the material image is decomposed using two orthogonal wavelet bases each time, then the first decomposition obtains two subband images of L1 and H1, and then continues to decompose H1 to obtain two subbands of L2 and H2 Image, and then continue to decompose H2 to get two subband images of L3 and H3. The sub-band images of the material image finally obtained are L1, L2, L3 and H3.

Step S204: Fusion sub-band images belonging to the same sub-band spectrum into a sub-band target image according to a preset fusion strategy.

Step S206: Inverse wavelet transform is performed on the sub-band target images corresponding to each sub-band spectrum to obtain a target image for feature recognition.

As in the above example, for the material images A and B, the decomposed subband images of the material image A are LA1, LA2, LA3, and HA3, and the decomposed subband images of the material image B are LB1, LB2, LB3, and HB3. The fusion of sub-band images is to fuse LA1 and LB1 into LF1, LA2 and LB2 into LF2, LA2 and LB2 into LF2, LA3 and LB3 into LF3, and HA1 and HB1 into HF1.

After LF1, LF2, LF3 and HF1 are obtained, the inverse transform of the frequency domain is inversely transformed into the image F to be recognized in the time domain by wavelet inverse transform.

Further, fusing sub-band images belonging to the same sub-band spectrum into a sub-band target image according to a preset fusion strategy includes:

Divide the sub-band images belonging to the same sub-band spectrum into two or more regions; merge the sub-band images of the same sub-band spectrum in the region images of each of the regions according to a preset fusion strategy to obtain the sub-band target image.

Taking the fusion of sub-band images LA1 and LB1 as an example, LA1 and LB1 can be divided into N×N regions, as shown in FIG. 6, which can be divided into 3×3 regions, and the position sequence numbers are from 1 to 9 in sequence. The band image LA1 contains nine sub-regions A1 to A9, and LB1 contains nine corresponding sub-regions B1 to B9.

The fusion strategy may include two methods of selection and weighted merging, for example, for the two areas of position numbers 1 and 2 (the area image corresponding to the position in the subband image LA1 is A1 and A2, and the area corresponding to the position in the subband image LB1 The images are B1 and B2). The features of LA1 at these two positions are not obvious, and the features of LB1 at these two positions are obvious. Therefore, for the two areas of position numbers 1 and 2, you can choose both B1 and B2 of LB1. The area images are taken as the area images F1 and F2 in the two areas of the position numbers 1 and 2 of the fused sub-band image LF1.

Correspondingly, for position numbers 6, 7 and 9, the characteristics of LA1 at these three positions are obvious, and the characteristics of LB1 at these three positions are obvious. Therefore, for the three areas of position numbers 6, 7 and 9, select LA1 The three area images of A6, A7 and A9 are used as the area images F6, F7 and F9 of the three areas of the fused subband image LF1 at the position numbers 6, 7 and 9.

For the four positions of position numbers 3, 4, 5 and 8, LA1 and LB1 include features that are not prominent but have differences in these four positions, then A3, A4, A5 and A8 can be combined with B3, B4, B5 And B8 are weighted, merged and fused according to position, respectively, to obtain the area images F3, F4, F5 and F8 of the subband image LF1 in the four areas of position numbers 3, 4, 5 and 8.

After the above fusion, it can be seen that the area images F1 to F9 at the position numbers 1 to 9 of the subband image LF1 of the final image to be recognized all contain better image features.

Further, the fusion of the sub-band images of the same sub-band spectrum in each of the regions according to a preset fusion strategy includes:

Calculating the regional variance of the regional images corresponding to the sub-band images in each of the regions; calculating the similarity of each regional image according to the regional variance; selecting or weighting and combining the similarities according to the similarity Regional image fusion.

In one embodiment, let R(x) be the subband coefficient matrix of the subband image after the material image x is subjected to wavelet decomposition, l is the position or position number of the subband image, and R(x, p) represents the position l The values of the decomposition coefficients at X, x (x, l) and u (x, l) respectively represent the variance and average value of the area size Q of the area of each subband matrix l in the image x, and:

X(x,l)=∑|R(x,q)-u(x,l)| ²

Where q represents a point within the area Q.

The similarity of the material images A and B in the region Q can be calculated by the following formula:

The similarity M _A,B (1) reflects the similarity of the variances of the two image regions. When the area images of the sub-band images of the material images A and B at l are more similar, the similarity M _A,B (l) is closer to 1; when the area images of the sub-band images of the material images A and B are at l The more different the images, the closer the similarity MA _{, B} (l) approaches zero.

In this embodiment, when the regional images are fused, whether to use the selected strategy or the weighted merge strategy depends on the size of the similarity M _A,B (l), and a similarity threshold T can be introduced. By comparing the similarity M _A,B (1) and the size of the similarity threshold T to determine whether to use the selected strategy or the weighted merge strategy. When the similarity M _{A, B} (l) is large, a weighted merge fusion strategy can be used, and when M _{A, B} (l) is small, the selected strategy is adopted, namely:

When M _{A, B} (l) <T, you can use:

That is, the sub-band image of the material image with a large regional variance is selected in the region of the region. This is because the region with the larger regional variance is usually a region with more obvious features, including differences in edges, textures, and information. Areas with large regional variances are usually backgrounds, solid colors, etc. Select the sub-band image of the material image with a large regional variance in the region image of the region, then select the region image with more feature information such as edges and textures as the sub-band image of the fusion image in the region, so that it contains More feature information, which makes the recognition accuracy higher.

When M _{A, B} (l)>T, a weighted fusion strategy can be used:

That is to say, when the similarity is large, the sub-band image of the material image with a large regional variance is more selected in the region image of the region, and the sub-band image of the material image with a large regional variance is less selected in the region The regional image of the region, and then weighted and fused by corresponding weighting coefficients W _max and W _min , where W _max is the weighting coefficient of the regional image of the sub-band image of the material image with a large regional variance in the region, and W _min is the region The weighting coefficient of the area image of the sub-band image of the material image with a small variance in this area, and the sum of W _max and W _min is 1.

In this embodiment, W _max and W _min are set as:

It should be noted that, in other embodiments, W _max and W _min may also be set according to actual scenarios. Choosing appropriate W _max and W _min can make the fused image retain more feature information.

The weighted fusion of the sub-band images of each material image in a certain area by this strategy not only retains the feature information contained in the sub-band image of the material image with relatively large regional variance, but also takes into account the regional variance. The feature information contained in the sub-band image of the relatively small material image ensures that the feature information is not omitted, so that the sub-band image in the region after the fusion image can contain more feature information, so that the identifiable accuracy Degree is higher.

The above image acquisition method is mainly applicable to the application scenarios of two-dimensional code/barcode recognition. After the target image for feature recognition is obtained, the two-dimensional code/barcode recognition can be performed on the target image. The above image acquisition method can also be applied to other image recognition fields, such as face recognition, vehicle detection, security inspection, and other application scenarios that require image acquisition and identification of features. The image acquisition method of the present invention is not limited to the foregoing The two-dimensional code/barcode identification process.

In one embodiment, for the above image acquisition method, an image acquisition device is also provided correspondingly. Specifically, as shown in FIG. 7, it includes a material image acquisition module 102 and an image fusion module 104, where:

The material image acquisition module 102 collects two or more material images under one or more optical parameters, one material image corresponds to a parameter interval of the optical parameters, and the optical parameters include at least the polarization direction and the spectrum At least one of them.

The image fusion module 104 performs image fusion on the collected material images to obtain a target image for feature recognition.

In one embodiment, the material image acquisition module 102 is used to acquire two or more material images through a hyperspectral image acquisition element and/or a polarized light image acquisition element, and the optical parameters corresponding to the hyperspectral image acquisition element are Spectrum, the optical parameter corresponding to the polarized light image acquisition element is the polarization direction.

In one embodiment, the image fusion module 104 is further used to perform wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum; subband images belonging to the same subband spectrum According to the preset fusion strategy, the sub-band target image is fused; the sub-band target image corresponding to each sub-band spectrum is subjected to inverse wavelet transform to obtain the target image for feature recognition.

In one embodiment, the image fusion module 104 is further used to divide the sub-band images belonging to the same sub-band spectrum into two or more regions; the sub-band images of the same sub-band spectrum are in the regions of each of the regions The images are fused according to a preset fusion strategy to obtain sub-band target images.

In one embodiment, the image fusion module 104 is further used to calculate the regional variance of the regional images corresponding to the sub-band images in each of the regions; calculate the similarity of each regional image according to the regional variance; according to the similarity The size of the degrees is selected or weighted to merge the area images.

The implementation of the embodiments of the present invention will have the following beneficial effects:

After the above image acquisition method and device are adopted, for the glare interference caused by the illumination direction, the material images in different polarization directions can be collected and then fused to avoid the reflection of polarized light generated in the specific illumination direction to the captured image. Interference of feature information, and the glare covered feature information collected in other polarization directions is used as a supplement to the final collected image to be recognized; at the same time, for the transparent or semi-transparent image to be recognized, or the background color fades or Contaminated and unclear pictures to be recognized can be obtained by collecting material images under different spectral bands, and extracting the feature information contained in the material images under each spectrum band, and adding to the final collected image to be recognized through image fusion . This makes the collected image to be recognized contain the feature information in the picture to be recognized to the greatest extent, which further improves the accuracy of subsequent image recognition.

In one embodiment, as shown in FIG. 8, FIG. 8 shows a computer system based on the von Neumann system that runs the above image acquisition method. Specifically, it may include an external input interface 1001 connected to the system bus, a processor 1002, a memory 1003, and an output interface 1004. The external input interface 1001 may optionally include at least a network interface 10012 and a USB interface 10014. The memory 1003 may include an external memory 10032 (for example, a hard disk, an optical disk, or a floppy disk, etc.) and an internal memory 10034. The output interface 1004 may include at least a display screen 10042 and other devices.

In this embodiment, the operation of the method is based on a computer program whose program files are stored in the external memory 10032 of the aforementioned computer system 10 based on the von Neumann system and are loaded into the internal memory 10034 during operation. It is then compiled into machine code and passed to the processor 1002 for execution, so that a logical material image acquisition module 102 and an image fusion module 104 are formed in the computer system 10 based on the von Neumann system. In addition, during the execution of the above image acquisition method, the input parameters are received through the external input interface 1001 and passed to the memory 1003 for buffering, and then input to the processor 1002 for processing. The processed result data may be cached in the memory 1003 for processing Subsequent processing, or passed to the output interface 1004 for output.

The above disclosure is only preferred embodiments of the present invention, and of course it cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope of the present invention.

Claims

An image acquisition method, characterized in that it includes:

Acquiring two or more material images under one or more optical parameters, one material image corresponding to a parameter interval of the optical parameters, the optical parameters including at least one of polarization direction and spectrum;

Perform image fusion on the collected material images to obtain a target image for feature recognition.
The image acquisition method according to claim 1, wherein the acquisition of two or more material images under one or more optical parameters includes:

Acquiring two or more material images through a hyperspectral image acquisition element and/or a polarized light image acquisition element, the optical parameter corresponding to the hyperspectral image acquisition element is a spectrum, and the optical parameter corresponding to the polarized light image acquisition element Is the polarization direction.
The image collection method according to claim 1, wherein the image fusion of the collected material images to obtain a target image for feature recognition includes:

Performing wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum;

Merge sub-band images belonging to the same sub-band spectrum into sub-band target images according to a preset fusion strategy;

The sub-band target image corresponding to each sub-band spectrum is subjected to inverse wavelet transform to obtain a target image for feature recognition.
The image acquisition method according to claim 3, wherein the fusing the subband images belonging to the same subband spectrum into a subband target image according to a preset fusion strategy includes:

Divide subband images belonging to the same subband spectrum into two or more regions;

The area images of the subband images of the same subband spectrum in each of the areas are fused according to a preset fusion strategy to obtain a subband target image.
The image acquisition method according to claim 4, wherein the fusion of the area images of the sub-band images of the same sub-band spectrum in each of the areas according to a preset fusion strategy includes:

Calculating the area variance of the area images corresponding to the sub-band images in each of the areas;

Calculate the similarity of images in each area according to the regional variance;

According to the size of the similarity, the region images are fused by selection or weighted merging.
The image acquisition method according to any one of claims 1 to 5, wherein after obtaining the target image for feature recognition, the method further includes:

Perform two-dimensional code/barcode identification on the target image.
An image acquisition device is characterized by comprising:

The material image acquisition module collects two or more material images under one or more optical parameters, one material image corresponds to a parameter interval of the optical parameters, and the optical parameters include at least the polarization direction and the spectrum At least one of

The image fusion module performs image fusion on the collected material images to obtain a target image for feature recognition.
The image acquisition device according to claim 7, wherein the material image acquisition module is used to acquire two or more material images through a hyperspectral image acquisition element and/or a polarized light image acquisition element, the The optical parameter corresponding to the hyperspectral image acquisition element is the spectrum, and the optical parameter corresponding to the polarized light image acquisition element is the polarization direction.
The image acquisition device according to claim 8, wherein the image fusion module is further used to perform wavelet decomposition on the material image to obtain two or more subband images corresponding to the subband spectrum; The sub-band images belonging to the same sub-band spectrum are fused into sub-band target images according to a preset fusion strategy; the sub-band target images corresponding to each sub-band spectrum are subjected to inverse wavelet transformation to obtain a target image for feature recognition.
The image acquisition device according to claim 9, wherein the image fusion module is further used to divide the subband images belonging to the same subband spectrum into two or more regions; The region images of the sub-band images in each of the regions are fused according to a preset fusion strategy to obtain a sub-band target image.
The image acquisition device according to claim 10, wherein the image fusion module is further used to calculate the regional variance of the regional images corresponding to the subband images in each of the regions; The similarity of the regional images; according to the size of the similarity, the regional images are fused by selection or weighted merging.