CN109709103B - Orange early-stage rotting identification system and method based on annular stripe polishing imaging - Google Patents
Orange early-stage rotting identification system and method based on annular stripe polishing imaging Download PDFInfo
- Publication number
- CN109709103B CN109709103B CN201910044898.5A CN201910044898A CN109709103B CN 109709103 B CN109709103 B CN 109709103B CN 201910044898 A CN201910044898 A CN 201910044898A CN 109709103 B CN109709103 B CN 109709103B
- Authority
- CN
- China
- Prior art keywords
- image
- current component
- component image
- citrus
- annular
- 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.)
- Active
Links
Images
Abstract
The invention relates to the technical field of optics and agriculture, and discloses an early citrus fruit rot identification system and method based on annular stripe polishing imaging, which comprises the following steps: the detection method comprises a digital light projector, a visible near-infrared light source, a computer, a precision moving object stage and a near-infrared camera, wherein the phase image is demodulated through the computer to obtain a direct-current component image and an alternating-current component image, a binarization template is constructed based on the direct-current component image, the background of the alternating-current component image is removed, the alternating-current component image without the background is subjected to image decomposition and image reconstruction enhancement by adopting two-dimensional empirical mode decomposition to obtain a reconstructed image, and the early rotting area of the citrus to be detected is segmented based on the reconstructed image and by combining a segmentation algorithm. The early rotten fruit identification system of oranges and tangerines has the advantage that the identification efficiency is high and the position of rotten region can be accurately judged.
Description
Technical Field
The invention relates to the technical field of optics and agriculture, in particular to an early citrus fruit rot identification system and method based on annular stripe polishing imaging.
Background
The annual output of citrus in China exceeds 3700 ten thousand tons, and the citrus is the first in the world. Decay caused by fungal infection is the most easily occurring, most serious and most difficult defect to detect after the citrus is picked. At present, related detection technologies mainly include an RGB computer vision technology, a near infrared spectrum technology, a multispectral imaging technology, a hyperspectral imaging technology, a fluorescence imaging technology and the like. For the RGB computer vision technology, the color of the skin of the infected area is almost the same as that of the normal skin due to early rotten fruits caused by fungal infection, so that the defect is very difficult to detect by using a color camera. In the case of near infrared spectroscopy, although this technique can be used to specifically and effectively express different types of tissues, and to distinguish different characteristic information according to a model established by spectral information, thereby achieving classification. However, the near infrared spectrum technology belongs to a single-point detection technology, which can only express the regional information of the measuring points, and the comprehensive judgment by fully utilizing the spatial information of the fruit surface is difficult to realize accurate classification in the application of detecting the early decay defect caused by citrus fungal infection. For the multi/hyperspectral imaging technology, the technology is difficult to popularize and apply in practice at present due to the problems of large data information amount, high system cost, difficult real-time online detection and the like. In contrast to these techniques, although fluorescence techniques show a certain potential for rotten fruit detection, practically not all rotten areas emit fluorescence, which makes fluorescence techniques problematic for practical applications.
Disclosure of Invention
Technical problem to be solved
The invention aims to provide an early citrus rot recognition system and method based on annular stripe polishing imaging, and aims to solve the technical problems that early citrus rot due to fungal infection cannot be effectively recognized and recognition accuracy is low in the prior art.
(II) technical scheme
In order to solve the technical problem, according to a first aspect of the present invention, there is provided an early citrus fruit rot recognition system for ring stripe lighting imaging, comprising: a digital light projector; a visible near-infrared light source for providing near-infrared light to the digital light projector; a computer for generating and transmitting a two-dimensional annular fringe image to the digital light projector so that the digital light projector emits fringe light to illuminate the citrus fruit to be inspected; the precision moving object stage is used for receiving the citrus to be detected and driving the citrus to be detected to move left and right along the horizontal direction; and the near-infrared camera is used for acquiring a phase image of the citrus to be detected, the phase image is demodulated through the computer, so that a direct-current component image and an alternating-current component image are obtained, a binarization template is constructed on the basis of the direct-current component image, the background of the alternating-current component image is removed, the alternating-current component image without the background is subjected to image decomposition and image reconstruction enhancement by adopting two-dimensional empirical mode decomposition, so that a reconstructed image is obtained, and the early rotting region of the citrus to be detected is segmented on the basis of the reconstructed image and in combination with a segmentation algorithm.
Wherein the early fruit rot recognition system further includes a first polarizing plate disposed in front of the emission end of the digital light projector.
The early rotten fruit identification system further comprises a narrow-band filter arranged at the front end of the lens of the near-infrared camera and a second polaroid arranged at the front end of the narrow-band filter.
The precision moving objective table comprises a precision motor electrically connected with the computer, a rotating shaft connected with the output end of the precision motor, and an objective table sleeved on the periphery of the rotating shaft and capable of reciprocating along the axial direction of the rotating shaft.
The computer comprises a motor control module, an image acquisition module, a projection control module and an annular fringe image generation module, wherein the motor control module is electrically connected with the precision motor and used for controlling the rotation and stop of the precision motor; the image acquisition module is electrically connected with the near-infrared camera and is used for acquiring a phase image of the citrus to be detected; the projection control module is electrically connected with the digital light projector and is used for controlling the digital light projector to emit annular light; the annular stripe image generation module is electrically connected with the projection control module and is used for generating a two-dimensional annular stripe image and loading the two-dimensional annular stripe image to the projection control module.
Wherein a first installation included angle is formed between the central line of the digital light projector and the central line of the near-infrared camera, and the size range of the first installation included angle is greater than or equal to 30 degrees and less than or equal to 45 degrees.
According to the second aspect of the application, a citrus early rotting identification method based on annular stripe polishing imaging is further provided, and comprises the following steps: generating a two-dimensional annular stripe image by adopting an annular stripe image generation module; loading the generated two-dimensional annular fringe image into a digital light projector; acquiring three phase images of the citrus to be detected by a near-infrared camera by means of fringe light emitted by a digital light projector and a precision moving objective table; demodulating the three phase images respectively to obtain a direct current component image and an alternating current component image; constructing a binarization template based on the direct current component image and removing the background of the alternating current component image; performing image decomposition and image reconstruction enhancement on the alternating current component image without the background by adopting two-dimensional empirical mode decomposition so as to obtain a reconstructed image; and segmenting early rotten regions of the citrus to be detected, which are formed by fungal infection, based on the obtained reconstructed image and by combining a traditional segmentation algorithm.
Wherein the conventional segmentation algorithm comprises one of a watershed algorithm and a global threshold method.
Wherein the method further comprises: the rotation of the output end of the precision motor is controlled through a computer, so that the rotation shaft is driven to rotate, and the object stage is driven to move along the axial direction of the rotation shaft according to the 2 pi/3 phase offset through the rotation of the rotation shaft.
Wherein the generation formula of the two-dimensional annular stripe image is
Here, I denotes a two-dimensional annular fringe image, f is a spatial frequency, x and y each denote a coordinate point on an annular ring in the two-dimensional annular fringe image, and DC denotes a direct-current component image; AC denotes an alternating current component image.
When three phase images are acquired, the initial position of the precision moving object stage is firstly set to be phase 0, at the moment, a first phase image is acquired, then the moving phase offset of the precision moving object stage is 2 pi/3 to acquire a second phase image, the moving phase offset of the precision moving object stage is 2 pi/3 again to acquire a third phase image, and then the precision moving object stage returns to the initial position with the phase 0.
Wherein, the formula of the DC component image obtained after demodulation is as follows
The formula for obtaining the alternating component image AC is
The image reconstruction and enhancement method is to perform image reconstruction and enhancement on an alternating current component image by adopting a formula of AC1 ═ AC-IMF1+ IMF2+ IMF3)/R after two-dimensional empirical mode decomposition, wherein,
the AC1 is an image of the alternating current component image after decomposition and reconstruction enhancement, the AC is an alternating current component image, and the IMF1, the IMF2, the IMF3 and R represent the 1 st, 2 nd and 3 rd intrinsic mode function images and the residual image generated after the alternating current component image is decomposed by the two-dimensional empirical mode respectively.
(III) advantageous effects
Compared with the prior art, the orange early-stage rotting identification system based on annular stripe polishing imaging provided by the invention has the following advantages:
the method comprises the steps of emitting stripe light through a digital light projector and irradiating the stripe light on the to-be-detected citrus, collecting three phase images of the to-be-detected citrus under specific frequency through a near-infrared camera, demodulating the three phase images of the to-be-detected citrus by adopting a proper demodulation technology to obtain a key alternating current component image, wherein the alternating current component image can highlight subcutaneous rot characteristics of the to-be-detected citrus, and further decomposing and reconstructing and enhancing the alternating current component image without background by adopting advanced two-dimensional empirical mode decomposition to further achieve the purpose of enhancing the contrast of a normal region and an early infection and rot region in the alternating current component image, and finally realizing the identification of the infection and rot region of the to-be-detected citrus.
In addition, the early rotten fruit identification system of oranges and tangerines of this application has the advantage that realizes simple relatively and recognition efficiency is high, has great application prospect in the detection of automatic oranges and tangerines quality. The method has important significance for researching and developing comprehensive quality grading equipment of high-end citrus fruits, reducing damage of the picked citrus fruits and increasing income of farmers.
Drawings
Fig. 1 is a schematic diagram of the overall structure of an annular streak illuminated imaging citrus early sapping identification system according to an embodiment of the present application;
fig. 2 is a schematic flow chart illustrating steps of a citrus early rot identification method by ring stripe polishing imaging according to an embodiment of the present application;
FIG. 3 is a two-dimensional annular fringe image generated by the annular fringe image generating module;
FIG. 4a is a first phase image of a sample acquired by a near infrared camera;
FIG. 4b is a second phase image of the sample acquired by the near infrared camera;
FIG. 4c is a third phase image of the sample acquired by the near infrared camera;
FIG. 5a is an image of a citrus fruit to be inspected;
fig. 5b is a direct current component image DC;
fig. 5c is an alternating current component image AC;
fig. 6 is a binarized image obtained based on the direct current component image DC in fig. 5 b;
FIG. 7 is a process image of an example of image decomposition and reconstruction enhancement of an alternating current component image AC based on a two-dimensional empirical mode decomposition;
FIG. 8a is a first process image of watershed defect segmentation for a reconstructed enhanced image;
FIG. 8b is a second process image of watershed defect segmentation for a reconstructed enhanced image;
FIG. 8c is a third process image of watershed defect segmentation for a reconstructed enhanced image;
FIG. 9a is an image of an example citrus fruit with a smaller infected area and with the infected area at the edge of the fruit to be inspected;
fig. 9b is an AC1 image of an example citrus fruit after image decomposition and reconstruction enhancement with a smaller infected area and with the infected area at the edge of the fruit to be inspected;
fig. 9c is an image of an infected rot area after an example citrus fruit has been segmented with smaller infected areas and the infected area at the edge of the fruit to be inspected.
In the figure, 1: a near-infrared camera; 2: a lens; 3: a narrow band filter; 4: a second polarizing plate; 5: a computer; 6: a digital light projector; 7: a visible near-infrared light source; 8: a cut-off filter; 9: a first polarizing plate; 20: a precision moving stage; 10: a precision motor; 11: a rotating shaft; 12: an object stage; 13: an optical fiber; 14: and (5) detecting the citrus.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
As shown in fig. 1, the citrus early-rotting identification system is schematically shown to include a near-infrared camera 1, a digital light projector 6, a visible near-infrared light source 7, a computer 5, a precision moving stage 20, and citrus fruit 14 to be inspected.
In the embodiment of the present application, the visible near-infrared light source 7 is used to provide near-infrared light to the digital light projector 6.
The computer 5 is configured to generate and transmit a two-dimensional annular fringe image to the digital light projector 6 such that the digital light projector 6 emits a fringe light that impinges upon the citrus fruit 14 to be inspected.
The precision moving stage 20 is used for receiving the citrus fruit 14 to be detected and driving the citrus fruit 14 to be detected to move left and right along the horizontal direction. Like this, through the motion that makes accurate removal objective table 20 to drive this to detect oranges and tangerines 14 and carry out the motion, through this motion of detecting oranges and tangerines 14, can make things convenient for near-infrared camera 1 to shoot the phase image that detects oranges and tangerines 14 and be in different phases.
The near-infrared camera 1 is used for acquiring a phase image of the citrus fruit 14 to be detected, wherein the phase image is demodulated by the computer 5, so that a direct-current component image DC and an alternating-current component image AC are obtained, a binarization template is constructed based on the direct-current component image DC and the background of the alternating-current component image AC is removed, image decomposition and image reconstruction enhancement are performed on the background-removed alternating-current component image AC by adopting two-dimensional empirical mode decomposition, so that a reconstructed image is obtained, and the early rotting region of the citrus fruit 14 to be detected is segmented based on the reconstructed image and by combining a segmentation algorithm. Specifically, streak light is emitted by the digital light projector 6 and is irradiated on the citrus fruit 14 to be detected, three phase images of the citrus fruit 14 to be detected at a specific frequency are acquired by the near-infrared camera 1, then, the three phase images of the citrus fruit 14 to be detected are demodulated by adopting a proper demodulation technology to obtain a key alternating current component image AC, the key alternating current component image AC can highlight the subcutaneous rotting characteristics of the citrus fruit 14 to be detected, and further, the background-removed alternating current component image AC is decomposed and reconstructed and enhanced by adopting advanced two-dimensional empirical mode decomposition to further enhance the contrast of a normal region and an early infected rotting region in the alternating current component image AC, so that the identification of the infected rotting region of the citrus fruit 14 to be detected is finally realized.
In addition, the early rotten fruit identification system of oranges and tangerines of this application has the advantage that realizes simple relatively and recognition efficiency is high, has great application prospect in the detection of automatic oranges and tangerines quality. The method has important significance for researching and developing comprehensive quality grading equipment of high-end citrus fruits, reducing damage of the picked citrus fruits and increasing income of farmers.
As shown in fig. 1, in a more preferred embodiment of the present application, the early-rot identification system further comprises a first polarizer 9 disposed in front of the emission end of the digital light projector 6. It should be noted that, since the "polarizer" is an optical element capable of converting natural light into polarized light, it has functions of shielding and transmitting incident light, and has two types of black and white and color, and can be divided into three types of transmission, transflective and anti-transmissive according to the application.
As shown in fig. 1, in order to further optimize the early rotten fruit identification system in the above technical solution, on the basis of the above technical solution, the early rotten fruit identification system further includes a narrow band filter 3 disposed at a front end of the lens 2 of the near infrared camera 1 and a second polarizer 4 disposed at a front end of the narrow band filter 3. Note that, by adding the narrow band filter 3 to the front end of the lens of the near-infrared camera 1, it is possible to allow only light in a specific wavelength range to pass therethrough. By "specific wavelength" is meant a wavelength between 805 nanometers and 815 nanometers in size.
The center wavelength of the narrow band filter 3 is 810nm (nanometers), and the wave width is 10nm (nanometers), so that only an image imaged by light passing through the narrow band filter 3 can be obtained on the near-infrared camera 1 with the help of the narrow band filter 3, that is, a near-infrared image having a center wavelength of 810nm (nanometers) is obtained. Here, a near infrared image with a central wavelength of 810nm (nanometers) is chosen, on the one hand, because the near infrared image is insensitive to the color of the fruit skin, and on the other hand, 810nm (nanometers) is the sensitive wavelength of the rotten region of the citrus fruit. By installing the second polarizing plate 4 in front of the narrow band filter 3, the second polarizing plate 4 is effective for light in the near infrared region, and the near infrared camera 1 is connected to the computer 5 through a data line.
Meanwhile, in the citrus early rotting fruit identification system shown in fig. 1, a cut-off filter 8 is installed in front of the visible near-infrared light source 7, the cut-off filter 8 has a cut-off effect on light below 780nm (nanometers) to ensure that light larger than 780nm (nanometers) can smoothly enter the digital light projector 6 through the optical fiber 13, the light entering the digital light projector 6 is only near-infrared light, the front end of the digital light projector 6 is provided with a first polarizer 9 matched with a second polarizer 4, the second polarizer 4 and the first polarizer 9 together form a polarizer group, and by matching the second polarizer 4 and the first polarizer 9, highlight (also called mirror reflection spot) formed by the near-infrared light on the surface of a spherical-like fruit can be effectively eliminated, and the digital light projector 6 is connected with the computer 5 through a data line.
As shown in fig. 1, in order to further optimize the precision moving object stage 20 in the above technical solution, on the basis of the above technical solution, the precision moving object stage 20 includes a precision motor 10 electrically connected to the computer 5, a rotating shaft 11 connected to an output end of the precision motor 10, and an object stage 12 sleeved on a periphery of the rotating shaft 11 and capable of reciprocating along an axial direction of the rotating shaft 11. It should be noted that the output end of the precision motor 10 is coaxially connected with the rotating shaft 11, and preferably, a shaft sleeve can be used to realize a firm connection between the two. Specifically, through starting this precision electric machine 10 and making this precision electric machine 10's output carry out axial rotation, axial rotation through this output just can drive this axis of rotation 11's axial rotation, and axial rotation through this axis of rotation 11 just can make this objective table 12 carry this to detect oranges and tangerines 14 and carry out reciprocating motion along the axial of this axis of rotation 11, and further, ensure that near-infrared camera 1 can shoot out the phase image that detects oranges and tangerines 14 and be in different phases.
In a preferred embodiment, the computer 5 comprises a motor control module, an image acquisition module, a projection control module and an annular fringe image generation module, wherein the motor control module is electrically connected with the precision motor 10 and used for controlling the rotation and stop of the precision motor 10.
The image acquisition module is electrically connected with the near-infrared camera 1 and is used for acquiring the phase image of the citrus fruit 14 to be detected.
The projection control module is electrically connected to the digital light projector 6 for controlling the digital light projector 6 to emit ring light.
The annular fringe image generation module is electrically connected with the projection control module and is used for generating a two-dimensional annular fringe image and loading the two-dimensional annular fringe image to the projection control module.
In another preferred embodiment, a first installation angle is configured between the center line of the digital light projector 6 and the center line of the near-infrared camera 1, the first installation angle having a magnitude in a range of 30 ° or more and 45 ° or less. Wherein the first mounting angle is preferably 33 °, 36 °, 39 ° or 42 °.
It should be noted that by controlling the size of the first installation angle to be in the range of 30 ° to 45 °, the streak light emitted by the digital light projector 6 can be accurately irradiated on the outer surface of the citrus fruit 14 to be detected and an optimal phase image can be obtained.
As shown in fig. 2, 3, 4a, 4b, 4c, 5a, 5b, 5c, 6, 7, 8a, 8b, 8c, 9a, 9b and 9c, according to the second aspect of the present application, there is also provided a method for identifying early rotting fruits of citrus using ring-shaped stripe polishing imaging, the method comprising: and step S1, generating a two-dimensional annular fringe image by adopting an annular fringe image generating module.
In step S2, the generated two-dimensional annular stripe image is loaded into the digital light projector 6.
In step S3, the near-infrared camera 1 acquires three phase images of the citrus fruit 14 to be inspected by means of the streak light emitted from the digital light projector 6 and the precision moving stage 20.
In step S4, the three phase images are demodulated to obtain a direct-current component image DC and an alternating-current component image AC, respectively.
In step S5, a binarization template is constructed based on the direct current component image DC and the background of the alternating current component image AC is removed.
And step S6, performing image decomposition and image reconstruction enhancement on the background-removed alternating current component image AC by adopting two-dimensional empirical mode decomposition, thereby obtaining a reconstructed image.
In step S7, based on the reconstructed image obtained and in combination with a conventional segmentation algorithm, the early rotting region of the citrus fruit 14 to be detected, which is formed due to fungal infection, is segmented.
In a preferred embodiment, the conventional segmentation algorithm comprises one of a watershed algorithm and a global threshold method. It should be noted that, since the watershed algorithm and the global threshold method are well known to those skilled in the art, they are not described in detail herein for the sake of brevity.
In a specific embodiment, the annular fringe image generating module may be Matlab software.
In a specific example, orange is a citrus fruit, which has a very important economic value in our country. Therefore, this embodiment will be described with respect to oranges.
Firstly, 100 orange samples are prepared, wherein the samples comprise 50 normal fruits and 50 early rotten fruits formed after fungal infection, wherein the fungal infection fruits are obtained by adopting an artificial inoculation method, and the specific method comprises the following steps: pericarp tissues infected with fungi are cut from the diseased part of orange fruits which are naturally diseased, spores are washed by sterile water, the spores are dissolved in the sterile water to form a spore solution, then a disposable syringe is used for inoculating the spore solution to 50 normal fruits, the inoculation depth is about 5mm (millimeter) below the skin, and the solution amount of each orange inoculated spore is 0.05ml (milliliter). And then, placing the inoculated sample in an incubator (the ambient temperature is 25-27 ℃, and the relative humidity is 96% -98%) for 2 days to form an infected area with the diameter of 5-15 mm (millimeters), wherein the infected area is difficult to identify by naked eyes.
Then, the image acquisition of all samples is carried out by using the citrus early rot identification system (shown in fig. 1) provided by the invention.
When collecting a sample image of a citrus fruit 14 to be detected, first, Matlab software is used to generate a two-dimensional annular stripe image as shown in fig. 3, where the two-dimensional annular stripe image is generated according to the formula
Here, I denotes a two-dimensional annular fringe image, f is a spatial frequency, x and y each denote a coordinate point on an annular ring in the two-dimensional annular fringe image, and DC denotes a direct-current component image; AC denotes an alternating current component image. Wherein f is equal to 0.15mm-1。
The generated two-dimensional annular stripe image is then loaded into digital light projector 6, and two-dimensional annular stripe lighting can be achieved by starting digital light projector 6.
Next, the citrus fruit 14 to be detected is placed on the stage 12 of the precision moving stage 20, and the initial position of the precision moving stage 20 is set to be phase 0, at this time, the sample image collected by the near-infrared camera 1 is the first phase image (see fig. 4 a).
Then, the stage 12 is moved along the rotation axis 11 by a phase shift of 2 π/3 under the control of the precision motor 10 of the precision moving stage 20, and a second phase image is obtained by the near-infrared camera 1 (see FIG. 4 b).
Again, the stage 12 is caused to move by a phase shift of 2 pi/3 along the axis of rotation 11 to obtain a third phase image (see fig. 4 c).
Subsequently, the precision movement stage 20 returns to the initial position of phase 0.
After three phase images are obtained, demodulating the three phase images of the citrus 14 to be detected to obtain a direct current component image DC and an alternating current component image AC, wherein the formula for obtaining the direct current component image DC by phase demodulation is
The formula for obtaining the alternating component image AC is
Wherein P1, P2, and P3 represent three phase images, respectively. Fig. 5a shows the citrus fruit 14 to be detected in the present embodiment, in which the circled area is an infected area (the original image is colored, and the fungal infected area is slightly yellow, and there is no obvious contrast with the citrus peel), and a rotten area that is difficult to identify on the citrus fruit 14 to be detected in fig. 5a is represented in the alternating current component image AC (as shown in fig. 5 c).
To further enhance the control of the normal peel and infected areas of the citrus fruit 14 to be tested, the alternating current component image AC is further processed here.
First, a binarized template image is obtained by using the direct current component image DC of fig. 5b in combination with a single threshold theory (the threshold may be 20), and then the background of the alternating current component image AC is removed by multiplying the binarized template image by the alternating current component image AC. And (3) performing image decomposition by adopting two-dimensional empirical mode decomposition based on the alternating current component image AC without the background, and taking the first 3 intrinsic mode function images and a residual image R after decomposition, wherein the 1 st intrinsic mode function image represents noise, the 2 nd and 3 rd intrinsic mode function images represent details, and the residual image R represents surface brightness information of the citrus to be detected 14.
Then, the formula is adopted as
AC1＝(AC-IMF1+IMF2+IMF3)/R，
And performing image reconstruction enhancement, wherein AC1 is an image obtained by decomposing and reconstructing an alternating current component image, AC is an alternating current component image, and IMF1, IMF2, IMF3 and R respectively represent the 1 st, 2 nd and 3 rd intrinsic mode function images and residual images generated after the alternating current component image is decomposed by a two-dimensional empirical mode.
Fig. 7 shows that the alternating current component image AC is subjected to image decomposition and reconstruction enhancement based on two-dimensional empirical mode decomposition, and AC1 is an image of the alternating current component image AC after decomposition and reconstruction enhancement. Comparing the image AC1 with the image AC, it is easy to find that the brightness of the image after two-dimensional empirical mode decomposition and reconstruction is more uniform and the contrast between the normal area and the rotten area of the sample is more obvious.
Based on the image AC1, an early stage infection region formed by orange fungal infection is segmented by using a watershed segmentation algorithm, and the segmentation result is shown in fig. 8a, 8b and 8c, wherein fig. 8a shows a watershed segmentation line formed after the watershed segmentation algorithm is used, it can be clearly seen from fig. 8a that a rotten region has been effectively segmented, fig. 8b is an image in which only the rotten region segmentation line is reserved after the orange boundary in fig. 8a is removed, and fig. 8c is an infection rotten region of an orange which is actually segmented.
To further verify the performance of the present application, as a specific example, the infected area on the surface of the orange is very small (less than 5 mm), and the infected area is located at the edge of the fruit, as shown in fig. 9a, where the circular area is the infected area (because the fruit is quasi-spherical, the illumination on the surface is not uniform, usually the middle is bright, and the edge is dark, so that when the infected area is located at the edge of the fruit, the defect segmentation is most difficult), and at this time, it is most difficult to perform the image defect segmentation detection, according to the same method as the foregoing embodiment, the detection result is shown in fig. 9b and 9c, fig. 9b is an AC1 image enhanced by image decomposition and reconstruction, and fig. 9c is an image of the infected rotten area obtained after segmentation.
As can be seen from the detection results, the technology provided by the invention is not influenced by the position of the infected area, and satisfactory results can be obtained even if the infected area is positioned at the most edge of the orange in the image.
The identification results of 100 samples (50 of each of the normal fruits and the fungal infection fruits) by using the technology provided by the invention are shown in the table 1, and the identification accuracy is 100%, thereby further proving the effectiveness of the invention.
TABLE 1
In summary, the digital light projector 6 emits the streak light to irradiate the citrus fruit 14 to be detected, the near-infrared camera 1 acquires three phase images of the citrus fruit 14 to be detected at a specific frequency, and then the three phase images of the citrus fruit 14 to be detected are demodulated by using a proper demodulation technique to obtain a key AC component image AC, which can highlight the subcutaneous rot characteristics of the citrus fruit 14 to be detected, and further, the background-removed AC component image AC is decomposed and reconstructed and enhanced by using advanced two-dimensional empirical mode decomposition to further enhance the contrast between the normal region and the early stage infection and rot region in the AC component image AC, so as to finally realize the identification of the infection and rot region of the citrus fruit 14 to be detected.
In addition, the early rotten fruit identification system of oranges and tangerines of this application has the advantage that realizes simple relatively and recognition efficiency is high, has great application prospect in the detection of automatic oranges and tangerines quality. The method has important significance for researching and developing comprehensive quality grading equipment of high-end citrus fruits, reducing damage of the picked citrus fruits and increasing income of farmers.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (12)
1. An early citrus fruit rot identification system with annular stripes for polishing and imaging is characterized by comprising:
a digital light projector;
a visible near-infrared light source for providing near-infrared light to the digital light projector;
a computer for generating and transmitting a two-dimensional annular fringe image to the digital light projector so that the digital light projector emits fringe light to illuminate the citrus fruit to be inspected;
the precision moving object stage is used for receiving the citrus to be detected and driving the citrus to be detected to move left and right along the horizontal direction; and
the near-infrared camera is used for acquiring a phase image of the citrus to be detected, the phase image is demodulated through the computer, so that a direct-current component image and an alternating-current component image are obtained, a binarization template is constructed on the basis of the direct-current component image, the background of the alternating-current component image is removed, the alternating-current component image without the background is subjected to image decomposition and image reconstruction enhancement by adopting two-dimensional empirical mode decomposition, so that a reconstructed image is obtained, and the early rotting region of the citrus to be detected is segmented on the basis of the reconstructed image and in combination with a segmentation algorithm;
the image reconstruction and enhancement method is that the alternating current component image adopts a formula after being decomposed by a two-dimensional empirical mode
AC1＝(AC-IMF1+IMF2+IMF3)/R
And performing image reconstruction enhancement, wherein AC1 is an image obtained by decomposing and reconstructing an alternating current component image, AC is an alternating current component image, and IMF1, IMF2, IMF3 and R respectively represent the 1 st, 2 nd and 3 rd intrinsic mode function images and residual images generated after the alternating current component image is decomposed by a two-dimensional empirical mode.
2. An annular striped illuminated imaged citrus early rot identification system according to claim 1, further comprising a first polarizer disposed in front of the emission end of the digital light projector.
3. An annular streak illuminated imaged citrus early rot identification system according to claim 1, further comprising a narrow band filter disposed in front of the lens of the near infrared camera and a second polarizer disposed in front of the narrow band filter.
4. An early citrus rot recognition system according to claim 1, wherein the precision moving stage comprises a precision motor electrically connected to the computer, a rotating shaft connected to an output end of the precision motor, and a stage that is disposed around the rotating shaft and can reciprocate in an axial direction of the rotating shaft.
5. The system for identifying early rotted citrus fruits according to claim 4, wherein the computer comprises a motor control module, an image acquisition module, a projection control module and an annular stripe image generation module, wherein the motor control module is electrically connected with the precision motor and used for controlling the rotation and stop of the precision motor;
the image acquisition module is electrically connected with the near-infrared camera and is used for acquiring a phase image of the citrus to be detected;
the projection control module is electrically connected with the digital light projector and is used for controlling the digital light projector to emit annular light;
the annular stripe image generation module is electrically connected with the projection control module and is used for generating a two-dimensional annular stripe image and loading the two-dimensional annular stripe image to the projection control module.
6. An annular streak illuminated imaged citrus early rotting identification system according to any of the claims 1 to 5, wherein a first installation angle is configured between the center line of the digital light projector and the center line of the near infrared camera, said first installation angle having a size ranging from 30 ° or more and 45 ° or less.
7. An early citrus fruit rot identification method based on annular stripe polishing imaging is characterized by comprising the following steps: generating a two-dimensional annular stripe image by adopting an annular stripe image generation module;
loading the generated two-dimensional annular fringe image into a digital light projector;
acquiring three phase images of the citrus to be detected by a near-infrared camera by means of fringe light emitted by a digital light projector and a precision moving objective table;
demodulating the three phase images respectively to obtain a direct current component image and an alternating current component image;
constructing a binarization template based on the direct current component image and removing the background of the alternating current component image;
performing image decomposition and image reconstruction enhancement on the alternating current component image without the background by adopting two-dimensional empirical mode decomposition so as to obtain a reconstructed image;
segmenting an early rotting area of the citrus to be detected, which is formed by fungal infection, based on the obtained reconstructed image and by combining a traditional segmentation algorithm;
the image reconstruction and enhancement method is that the alternating current component image adopts a formula after being decomposed by a two-dimensional empirical mode
AC1＝(AC-IMF1+IMF2+IMF3)/R
And performing image reconstruction enhancement, wherein AC1 is an image obtained by decomposing and reconstructing an alternating current component image, AC is an alternating current component image, and IMF1, IMF2, IMF3 and R respectively represent the 1 st, 2 nd and 3 rd intrinsic mode function images and residual images generated after the alternating current component image is decomposed by a two-dimensional empirical mode.
8. The method of claim 7, wherein the conventional segmentation algorithm comprises one of a watershed algorithm and a global thresholding method.
9. The method of claim 7, further comprising: the rotation of the output end of the precision motor is controlled through a computer, so that the rotation of a rotating shaft connected with the output end of the precision motor is driven, and the object stage is driven to move along the axial direction of the rotating shaft according to the phase shift of 2 pi/3 through the rotation of the rotating shaft.
10. The method of claim 7, wherein the two-dimensional annular fringe image is generated by the formula
Here, I denotes a two-dimensional annular fringe image, f is a spatial frequency, x and y each denote a coordinate point on an annular ring in the two-dimensional annular fringe image, and DC denotes a direct-current component image; AC denotes an alternating current component image.
11. The method of claim 7, wherein in acquiring the three phase images, the initial position of the fine moving stage is first set to phase 0, at which time a first phase image is acquired, then the fine moving stage is moved by a phase offset of 2 pi/3 to acquire a second phase image, and the fine moving stage is moved again by a phase offset of 2 pi/3 to acquire a third phase image, and then the fine moving stage is returned to the initial position of phase 0.
12. The method according to claim 7, wherein the dc component image obtained after demodulation is formulated as
The formula for obtaining the alternating component image AC is
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910044898.5A CN109709103B (en) | 2019-01-17 | 2019-01-17 | Orange early-stage rotting identification system and method based on annular stripe polishing imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910044898.5A CN109709103B (en) | 2019-01-17 | 2019-01-17 | Orange early-stage rotting identification system and method based on annular stripe polishing imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109709103A CN109709103A (en) | 2019-05-03 |
| CN109709103B true CN109709103B (en) | 2020-08-11 |
Family
ID=66262414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910044898.5A Active CN109709103B (en) | 2019-01-17 | 2019-01-17 | Orange early-stage rotting identification system and method based on annular stripe polishing imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109709103B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113128249B (en) * | 2019-12-26 | 2022-08-23 | 合肥美亚光电技术股份有限公司 | Garlic scar identification method and system and sorting equipment |
| US20230236117A1 (en) * | 2020-06-24 | 2023-07-27 | Rubens IP Pty Ltd | Detecting plant product properties |
| CN113256575B (en) * | 2021-05-13 | 2023-10-27 | 南京农业大学 | Fruit defect detection method based on structured hyperspectral system |
| CN115620299B (en) * | 2022-12-14 | 2023-03-21 | 深圳思谋信息科技有限公司 | Image recognition method and device, computer equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102854192A (en) * | 2012-08-22 | 2013-01-02 | 北京农业智能装备技术研究中心 | System and method for detecting apple surface defect |
| CN103698009A (en) * | 2013-12-25 | 2014-04-02 | 北京农业智能装备技术研究中心 | Multispectral image acquiring method and system on basis of line scanning hyperspectral imaging |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2985025B1 (en) * | 2011-12-23 | 2014-12-12 | Maf Agrobotic | DEVICE AND METHOD FOR NON-DESTRUCTIVE DETECTION OF DEFECTS IN FRUIT AND VEGETABLES |
| CN105427274B (en) * | 2015-10-28 | 2017-12-26 | 北京农业智能装备技术研究中心 | It is a kind of that rotten citrus image detecting method is caused by mould infection |
| CN106442561B (en) * | 2016-08-29 | 2023-10-31 | 北京农业智能装备技术研究中心 | Online image acquisition system and method for detecting citrus peel defects |
| JP6203923B1 (en) * | 2016-10-04 | 2017-09-27 | 三井金属計測機工株式会社 | Fruit and vegetable inspection equipment |
-
2019
- 2019-01-17 CN CN201910044898.5A patent/CN109709103B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102854192A (en) * | 2012-08-22 | 2013-01-02 | 北京农业智能装备技术研究中心 | System and method for detecting apple surface defect |
| CN103698009A (en) * | 2013-12-25 | 2014-04-02 | 北京农业智能装备技术研究中心 | Multispectral image acquiring method and system on basis of line scanning hyperspectral imaging |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109709103A (en) | 2019-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109709103B (en) | Orange early-stage rotting identification system and method based on annular stripe polishing imaging | |
| JP5443435B2 (en) | Tire defect detection method | |
| CN101832941B (en) | Fruit quality evaluation device based on multispectral image | |
| CN108765402B (en) | Non-woven fabric defect detection and classification method | |
| Regeling et al. | Development of an image pre‐processor for operational hyperspectral laryngeal cancer detection | |
| CN108490000A (en) | A kind of Bar Wire Product surface defect on-line measuring device and method | |
| CN108961251A (en) | A kind of raw cotton fault and defects inspecting and recognition methods and system | |
| JP6318755B2 (en) | Filamentous bacteria detection apparatus and filamentous bacteria detection method | |
| JP6454923B2 (en) | Food inspection equipment | |
| CN106296670A (en) | A kind of Edge detection of infrared image based on Retinex watershed Canny operator | |
| US20140334711A1 (en) | Image processing | |
| CN111122590A (en) | Ceramic surface defect detection device and detection method | |
| Al-Kadi | A gabor filter texture analysis approach for histopathological brain tumor subtype discrimination | |
| CN112304876A (en) | Cotton color degree detection device and detection method based on machine vision | |
| Takahama Junior et al. | Usefulness of tissue autofluorescence imaging in actinic cheilitis diagnosis | |
| CN103091330A (en) | Copper strip surface defect identifying device for simulating human visual perception mechanism and method thereof | |
| CN113256575B (en) | Fruit defect detection method based on structured hyperspectral system | |
| Defrianto et al. | Fluorescence spectrum analysis on leaf and fruit using the ImageJ software application | |
| CN102184389B (en) | Automatic annual ring identification method used for tree species with high chromatic aberration of heartwood and sapwood | |
| CN109990742A (en) | Chinese chestnut detection method based on image processing techniques | |
| CN110987957A (en) | Intelligent defect removing method based on machine vision and laser processing | |
| CN105954295A (en) | Helmet top facing defect detection system and method | |
| CN109506900B (en) | Imaging frame rate detection system and detection method for near infrared camera | |
| Zhang et al. | Identification of cotton contaminants using neighborhood gradient based on YCbCr color space | |
| Wang et al. | Segmentation of pathological features of rat bile duct carcinoma from hyperspectral images |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |