Plankton detection method and system based on image processing and biological information correction

Abstract

The present invention relates to a plankton detection method and system based on image processing and biological information correction, and the method includes: performing multi-view three-dimensional scanning on a glass slide to obtain a multi-view original image sequence; performing detection on the original image sequence to obtain a detection result image sequence; splicing the detection result image sequence to obtain a final detection result image; splicing the final detection result images of all fields of view to obtain a panoramic detection result image; judging whether there are incomplete plankton in the panoramic detection result image, and if so, correcting the panoramic detection result image using a secondary detection result image; performing detection on the glass slide based on a biological detection method to obtain a standard detection result, and using the standard detection result to perform biological correction on the panoramic detection result image. The present invention is based on computer vision image detection, first performs computer vision correction, and then performs biological correction, so as to obtain the final accurate detection result of plankton.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a plankton detection method and system based on image processing and biological information correction. Background Art

[0002] With the development and progress of society, water ecology detection is becoming more and more important. At present, some scientific research institutions and related enterprises have completed the sequencing of some cyanobacteria and can detect cyanobacteria in water bodies, but the statistical data are often biased, and due to the huge workload, the sequencing of other algae, such as green algae and diatoms, has not been completed, and the counting ability of large algae is weak. For example, for large algae, multiple probes often point to the same algae individual, which will cause non-negligible errors in the statistics of algae abundance.

[0003] With the development of GPU hardware technology, a new generation of computer vision technology based on deep neural networks began to be applied to various industries around 2010. In 2017, computer vision technology began to be combined with algae morphological recognition, and a few technology companies and research institutions launched this generation of technology into the market in 2021. At present, computer vision technology is limited by the current status of hardware and theory, and has gradually approached the bottleneck of the current scientific and technological level. The accuracy of morphological recognition using the most advanced computer vision technology is also limited by this. The current products can accurately detect common algae to the genus, and there is still room for optimization in meeting the requirements of ordinary first-line detection sites. For example, in the recognition of computer images, there are many areas worth optimizing and upgrading.

[0004] Therefore, existing simple biotechnology or computer technology is not capable of measuring algae. Summary of the invention

[0005] The present invention provides a plankton detection method and system based on image processing and biological information correction to solve at least one of the above technical problems.

[0006] The technical solution of the present invention to solve the above technical problems is as follows: a plankton detection method based on image processing and biological information correction comprises the following steps:

[0007] S1, making a glass slide of a water sample to be tested containing plankton, and performing a multi-view three-dimensional scanning on the glass slide to obtain a multi-view original image sequence;

[0008] S2, using a deep learning method to perform plankton detection based on image recognition on each frame of the original image sequence of each field of view, to obtain a detection result image sequence of each field of view;

[0009] S3, performing image stitching on each visual field detection result image sequence by pyramid fusion to obtain the final detection result image of each visual field;

[0010] S4, performing image stitching on the final detection result images of all fields of view using pyramid fusion to obtain a panoramic detection result image based on computer vision;

[0011] S5, determining whether there are incomplete plankton at the image stitching edge of each field of view in the panoramic detection result image, if yes, executing S6 and S7, if no, executing S7;

[0012] S6, performing image stitching on each field of view original image sequence by pyramid fusion to obtain each field of view synthetic image, and performing image stitching on all field of view synthetic images by pyramid fusion to obtain a panoramic synthetic image; based on the field of view where incomplete plankton exists at the edge of the image in the panoramic detection result image, performing plankton detection based on image recognition on a preset area of ​​the panoramic synthetic image by using a deep learning method to obtain a secondary detection result image; and correcting the panoramic detection result image by using the secondary detection result image to obtain a corrected panoramic detection result image;

[0013] S7, performing plankton detection on the glass slide based on a biological detection method to obtain a standard detection result, and using the standard detection result to perform biological correction on the panoramic detection result image to obtain a final plankton detection result.

[0014] Based on the above technical solution, the present invention can also be improved as follows.

[0015] Further, in S3, pyramid fusion is used to perform image splicing on any visual field detection result image sequence, specifically,

[0016] S31, calculating the weight of each frame image in any visual field detection result image sequence, and selecting one frame image from the visual field detection result image sequence as a base image, and the remaining images as non-base images;

[0017] S32, setting the confidence of the area where plankton exists in the bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the bottom plate image to 1, to obtain an initial weighted image of the bottom plate;

[0018] S33, setting the confidence of the area where plankton exists in the non-bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the non-bottom plate image to 0, to obtain a non-bottom plate initial weight image;

[0019] S34, performing Gaussian filtering on the bottom plate initial weight image and the non-bottom plate initial weight image respectively, to obtain a bottom plate weight image and a non-bottom plate weight image respectively;

[0020] S35, taking the bottom plate weighted image as a reference, using a pyramid fusion method to fuse all non-bottom plate weighted images located on different focal planes into the bottom plate weighted image, to obtain a final detection result image of the field of view.

[0021] Furthermore, in S3, each field of view final detection result image includes a plankton final detection result, and the plankton final detection result is queried in the corresponding field of view final detection result image by full-text searching the bottom plate weight image.

[0022] Furthermore, before S4, the method further includes performing white balance processing on the final detection result image of each field of view respectively;

[0023] In S4, the final detection result images of all visual fields after white balance processing are spliced ​​by pyramid fusion according to the XY axis position information.

[0024] Furthermore, in S4, if the sizes of the two frames of images to be stitched are inconsistent during the image stitching process, the overlapping area of ​​the two frames of images to be stitched is first calculated, and then the image canvas is expanded according to the overlapping area, and finally the image stitching is performed using pyramid fusion.

[0025] Further, in S6, the preset area corresponds to the middle area of ​​the stitching unit formed by the fields of view on both sides of the stitching edge of the image where incomplete plankton exists in the panoramic detection result image;

[0026] The panoramic detection result image is corrected using the secondary detection result image, specifically,

[0027] Extracting a detection frame from the secondary detection result image;

[0028] Comparing the detection frame in the secondary detection result image with the panoramic detection result image;

[0029] If the detection frame in the secondary detection result image has an overlapping area with the panoramic detection result image, the panoramic detection result image is corrected based on the secondary detection result image.

[0030] Further, in S1, the water sample to be tested containing plankton is made into a glass slide by,

[0031] Adding a water sample to be tested containing plankton into a gene probe chip made of a transparent medium;

[0032] After adding internal reference nucleic acid to the gene probe chip to which the water sample to be tested containing plankton is added, DNA fluorescence labeling is performed;

[0033] The gene probe chip after DNA fluorescence labeling is used as the glass slide.

[0034] Further, in S7, plankton detection is performed on the glass slide based on a biological detection method, specifically,

[0035] Performing DNA chip hybridization on the gene probe chip after DNA fluorescence labeling to obtain a DNA hybridization chip;

[0036] Scanning the DNA hybridization chip to obtain chip imaging results;

[0037] The standard detection result is obtained by calculating the fluorescence signal intensity and position information in the chip imaging result.

[0038] Further, in S7, the panoramic detection result image is biologically corrected using the standard detection result, specifically,

[0039] Comparing the standard detection result with the detection result included in the panoramic detection result image based on the same position;

[0040] When the standard detection result conflicts with the detection result included in the panoramic detection result image at the same position, the detection result in the panoramic detection result image is corrected based on the standard detection result at that position.

[0041] On the basis of the above-mentioned plankton detection method based on image processing and biological information correction, the present invention also provides a plankton detection system based on image processing and biological information correction.

[0042] A plankton detection system based on image processing and biological information correction includes a processor, a computer-readable storage medium, and a computer program or instruction stored in the computer-readable storage medium. When the computer program or instruction is executed by the processor, the plankton detection method based on image processing and biological information correction as described above is implemented.

[0043] The beneficial effects of the present invention are as follows: the plankton detection method and system based on image processing and biological information correction of the present invention are based on computer vision detection, which can avoid the limitation of excessive workload and statistical data bias caused by biotechnology detection; at the same time, in the process of computer vision detection, the two-dimensional image is displayed in a full-view three-dimensional microscopic world, giving users a more straightforward presentation of individual plankton; and in the process of full-view three-dimensional microscopic world display, the detection results of the panoramic synthetic image are used to correct the single-view detection results, avoiding the problem of misdetection of cut plankton due to incomplete morphology, thereby effectively increasing the accuracy of plankton image recognition; finally, the panoramic detection result image is biologically corrected using the standard detection results obtained by the biological detection method, so as to obtain the final accurate plankton detection result. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the plankton detection method based on image processing and biological information correction of the present invention;

[0045] Figure 2 It is a structural block diagram of the plankton detection system based on image processing and biological information correction of the present invention. DETAILED DESCRIPTION

[0046] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0047] Embodiment 1:

[0048] like Figure 1 As shown, the plankton detection method based on image processing and biological information correction includes the following steps:

[0049] S1, making a glass slide of a water sample to be tested containing plankton, and performing a multi-view three-dimensional scanning on the glass slide to obtain a multi-view original image sequence;

[0050] S2, using a deep learning method to perform plankton detection based on image recognition on each frame of the original image sequence of each field of view, to obtain a detection result image sequence of each field of view;

[0051] S3, performing image stitching on each visual field detection result image sequence by pyramid fusion to obtain the final detection result image of each visual field;

[0052] S4, performing image stitching on the final detection result images of all fields of view using pyramid fusion to obtain a panoramic detection result image based on computer vision;

[0053] S5, determining whether there are incomplete plankton at the image stitching edge of each field of view in the panoramic detection result image, if yes, executing S6 and S7, if no, executing S7;

[0054] S6, performing image stitching on each field of view original image sequence by pyramid fusion to obtain each field of view synthetic image, and performing image stitching on all field of view synthetic images by pyramid fusion to obtain a panoramic synthetic image; based on the field of view where incomplete plankton exists at the edge of the image in the panoramic detection result image, performing plankton detection based on image recognition on a preset area of ​​the panoramic synthetic image by using a deep learning method to obtain a secondary detection result image; and correcting the panoramic detection result image by using the secondary detection result image to obtain a corrected panoramic detection result image;

[0055] S7, performing plankton detection on the glass slide based on a biological detection method to obtain a standard detection result, and using the standard detection result to perform biological correction on the panoramic detection result image to obtain a final plankton detection result.

[0056] Each step is described in detail below.

[0057] In said S1:

[0058] The specific steps of making a glass slide of a water sample containing plankton to be tested are as follows:

[0059] Adding a water sample to be tested containing plankton into a gene probe chip made of a transparent medium;

[0060] After adding internal reference nucleic acid to the gene probe chip to which the water sample to be tested containing plankton is added, DNA fluorescence labeling is performed;

[0061] The gene probe chip after DNA fluorescence labeling is used as the glass slide.

[0062] The method of performing multi-view three-dimensional scanning on the glass slide is scanning from the bottom layer to the top layer. Compared with the traditional scanning from top to bottom, the three-dimensional scanning method of scanning from the bottom layer to the top layer will make the image clearer, and the results obtained by subsequent image recognition based on deep learning will be faster and more accurate. In the scanning process from the bottom layer to the top layer, the Canny operator is used to perform edge detection on each frame of the acquired image. If the detected result meets the preset conditions, the image that meets the preset conditions will be retained; the purpose of this is to retain objects with clear targets or larger targets and filter out impurity data that interferes with the calculation.

[0063] In said S2:

[0064] The deep learning method can be an existing method, such as the method disclosed in patent application number CN202211393317.7.

[0065] Before using the deep learning method to detect plankton in each frame of the original image sequence of each field of view, it is necessary to build a neural network detection model based on deep learning. In building the neural network detection model, it is necessary to use samples in a pre-constructed image database to train the neural network. The pre-constructed image database of the present invention includes negative samples. The method for making negative samples is: the gene probe chip is used as a glass slide in the traditional sense, and pure water or water mist or impurities without algae are added thereto, and the image is scanned and taken, and the image is retained, and is added to the image database as a negative sample for training. The advantage of this is that the amount of negative samples can be extremely conveniently increased, and the probability of false detection in actual scanning detection can be reduced.

[0066] Of course, the deep learning-based neural network used in the present invention can also be a target detection neural network such as the YOLO series and SSD series for target detection, or an image segmentation neural network such as the UNET series, etc.

[0067] It should also be noted that each frame of each visual field detection result image sequence includes a plankton detection result; wherein the plankton detection result includes the location information, quantity information, DNA information and biological classification information of the plankton.

[0068] In the S3:

[0069] For any visual field detection result image sequence, pyramid fusion is used to perform image stitching, specifically,

[0070] S31, calculating the weight of each frame image in any visual field detection result image sequence (the calculation method of the weight of the image can be referred to the patent application number CN202311014429.1, which will not be repeated here), and selecting a frame image from the visual field detection result image sequence as the base image, and the remaining images as non-base images;

[0071] S32, setting the confidence of the area where plankton exists in the bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the bottom plate image to 1, to obtain an initial weighted image of the bottom plate;

[0072] S33, setting the confidence of the area where plankton exists in the non-bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the non-bottom plate image to 0, to obtain a non-bottom plate initial weight image;

[0073] S34, performing Gaussian filtering on the bottom plate initial weight image and the non-bottom plate initial weight image respectively, to obtain a bottom plate weight image and a non-bottom plate weight image respectively;

[0074] S35, taking the bottom plate weighted image as a reference, using a pyramid fusion method to fuse all non-bottom plate weighted images located on different focal planes into the bottom plate weighted image, to obtain a final detection result image of the field of view.

[0075] Preferably, in S3, each field of view final detection result image includes a plankton final detection result, and the plankton final detection result is queried in the corresponding field of view final detection result image by a full-text search of the bottom plate weight image. The plankton final detection result includes the location information, quantity information, DNA information, and biological classification information of the plankton.

[0076] In said S4:

[0077] Before S4, the method further includes performing white balance processing on the final detection result image of each field of view respectively;

[0078] In S4, the final detection result images of all visual fields after white balance processing are spliced ​​by pyramid fusion according to the XY axis position information.

[0079] Specifically, each field of view is based on the XY axis position information (the XY axis position information is obtained from the controller), and pyramid fusion is performed based on this XY axis position information. Two frames are combined into one frame. Take the minimum splicing unit of two frames of images (1000*1000) as an example: a pyramid weight image is established for each frame of image, and the overlapped position of the two frames of image is set to 0.5, the non-overlapped position is set to 1, and the other positions are set to 0. It is worth noting that the weight image is the actual size of the area composed of the two frames of image, that is, if the overlapping pixels of the two frames of image are 100, the weight image is (1900*1900). It is worth noting that not every two frames of image are the same size. For example, a 1900*1900 image needs to be fused with a 1000*1000 image. At this time, it is necessary to calculate the overlapping area of ​​the two, and then expand and fuse the image canvas.

[0080] In said S5:

[0081] If there are plankton in the panoramic detection result image at the edge of the image and it appears incomplete, it will cause false detection, so it is necessary to perform secondary detection and correction as described in S6.

[0082] In said S6:

[0083] The preset area corresponds to the middle area of ​​a stitching unit formed by fields of view on both sides of the stitching edge of the image where incomplete plankton exists in the panoramic detection result image.

[0084] For example, if there are incomplete plankton at the image stitching edge of the nth field of view (image size 1000*1000) and the n+1th field of view (image size 1000*1000) in the panoramic detection result image, then an area with an image size of 1000*1000 is intercepted in the middle of the stitching unit composed of the nth field of view and the n+1th field of view, and the intercepted area is mapped to the panoramic composite image to obtain the preset area.

[0085] The panoramic detection result image is corrected using the secondary detection result image, specifically,

[0086] Extracting a detection frame from the secondary detection result image;

[0087] Comparing the detection frame in the secondary detection result image with the panoramic detection result image;

[0088] If the detection frame in the secondary detection result image has an overlapping area with the panoramic detection result image, the panoramic detection result image is corrected based on the secondary detection result image.

[0089] Since image stitching can easily cause plankton to be cut in the image, the cut plankton may cause misdetection due to its incomplete morphology. Therefore, the present invention utilizes the detection results of the panoramic composite image to correct the detection results of a single field of view, which can effectively increase the accuracy of plankton image recognition.

[0090] Since computer vision technology is limited by the current hardware and theoretical status, it has gradually approached the bottleneck of the current scientific and technological level. The accuracy of morphological recognition using the most advanced computer vision technology is also limited by this. Therefore, relying solely on computer vision to detect plankton may result in misdetection. Therefore, the present invention also needs to perform three detections and corrections as described in S7.

[0091] In the S7:

[0092] The slide is subjected to plankton detection based on a biological detection method, specifically,

[0093] Performing DNA chip hybridization on the gene probe chip after DNA fluorescence labeling to obtain a DNA hybridization chip;

[0094] Scanning the DNA hybridization chip to obtain chip imaging results;

[0095] The standard test result is obtained by calculating the fluorescence signal intensity and position information in the chip imaging result; wherein the standard test result includes the position information, quantity information, DNA information and biological classification information of the plankton.

[0096] The specific steps of DNA chip hybridization on the gene probe chip after DNA fluorescence labeling are as follows: the DNA sequence is fixed on a solid surface to form a chip, and then the target DNA sample is hybridized with the DNA sequence on the chip, and the signal intensity of the hybridization reaction is detected to understand the presence and expression of the labeled DNA in the sample.

[0097] Scanning the DNA hybridization chip specifically involves scanning the chip with a high-resolution fluorescence microscope so that each fluorescence signal corresponds to a specific position on the chip. The intensity and color of the fluorescence signal are measured and recorded by a corresponding instrument to obtain the chip imaging result.

[0098] In the process of calculating the fluorescence signal intensity and position information in the chip imaging results, in order to avoid the interference of some non-specific probe signals in the hybridization experiment, the following steps are used to interpret the probe signal of the gene probe core: First, all probe signals are screened for hybridization confidence, and the probe hybridization fluorescence signals with lower confidence are deleted. Whether the signal of a hybridization site is qualified is mainly determined by statistically analyzing the fluorescence values ​​of all pixels contained in the hybridization site image. Subsequently, the detection standard for a certain strain is that the proportion of qualified signal values ​​in all specific probes contained in the strain is greater than 70% (PDR, Probe Detection Rate). The probe signal analysis steps mainly include:

[0099] 1) Screening of signal-to-noise ratio of probe signals. If the signal-to-noise ratio is greater than 2, the single probe signal value minus the background value is divided by the standard deviation of the background value. If the result is greater than 2, the probe signal is judged to be qualified.

[0100] 2) Screening of probe signal signal ratio: If the signal-to-noise ratio is greater than 2, the single probe signal value is divided by the background signal value, and the probe signal with a result greater than 2 is judged as a qualified signal.

[0101] 3) Probe Detection Rate (PDR) screening: The proportion of a certain strain with qualified probe signals is counted. When the PDR is greater than 70%, the strain is considered to be detected in the sample.

[0102] It should be noted that any one of the above-mentioned method 1) and method 2) for the probe signal analysis can be selected according to actual needs.

[0103] The specific method of obtaining the standard test result is that when the gene probe chip presents a fluorescent effect, that is, the plankton to be detected is detected in the water sample, the DNA information detected by the gene probe chip can be queried in the pre-established DNA database, thereby obtaining the standard test result, that is, the position information of the plankton to be detected in the gene probe chip, the DNA information of the plankton, and the biological classification of the plankton detected in DNA probe biology. It should be noted that the position information of the plankton to be detected in the gene probe chip can refer to the physical position information that can distinguish the position by formulating a coordinate system based on the gene probe chip, or according to other needs. The position information of the plankton to be detected in the gene probe chip mentioned below is the same as that here, because the two methods need to rely on the biological classification and DNA information under the same position information for comparison, so the position information of the plankton to be detected in the gene probe chip is consistent with the coordinate system or physical position information set by the position information of the plankton to be detected in the gene probe chip, or can be converted into the same coordinate system or physical position information through a certain calculation method.

[0104] In addition, the process of establishing a DNA database can be: sorting out common plankton DNA data, determining the types of gene sequencing, in other words, investigating common algae, and then sequencing them, and establishing the data as a DNA database of common plankton (algae), and then searching and searching according to the DNA database to obtain the corresponding biological information of plankton. In the process of determining the types of gene sequencing, the DNA samples of plankton are downsampled, and a large amount of gene sequencing work is replaced by common DNA sequence determination. There are about 2,000 genera of algae, and the common ones generally do not exceed 100 genera, so the DNA information of common algae is used to replace the DNA sequence of the spectrum, which will greatly reduce the workload. Uncommon plankton (such as Echinococcus, etc.) or plankton that can be well solved by computer identification are not included in this list. Uncommon plankton is the plankton that is not very common in water bodies.

[0105] The standard detection result is used to perform biological correction on the panoramic detection result image, specifically,

[0106] Comparing the standard detection result with the detection result included in the panoramic detection result image based on the same position;

[0107] When the standard detection result conflicts with the detection result included in the panoramic detection result image at the same position, the detection result in the panoramic detection result image is corrected based on the standard detection result at that position.

[0108] The same gene probe chip can be used to obtain the position of the plankton to be detected on the chip, and the data at the same position can be compared for calibration, so as to reduce the impact of human errors on subsequent data calibration. In addition, the computer technology for image detection is relatively low-cost and reliable to a certain extent. The results of computer vision measurement are corrected based on the gene probe chip, which is lower in cost and the same result is accurate. At the same time, the volume information, quantity information and distribution of the plankton to be detected can also be collected and counted in computer vision measurement. The classification information and quantity volume information of algae are given by computer vision measurement, and the classification information of algae is obtained by DNA gene probe chip measurement. The classification information of algae in the DNA gene probe chip measurement is used to calibrate the errors of algae classification information that may occur in computer vision measurement, and obtain the correct algae classification and quantity information and volume information that cannot be obtained by the gene probe chip. In addition, the volume can be estimated by computer frame, because the density of plankton is consistent with the density of water, and the three-dimensional cells of a certain species can be roughly assigned to a certain three-dimensional model, and then combined with the frame size information detected by the computer, it can be roughly estimated, which is convenient and fast.

[0109] In general, the present invention is based on computer vision image detection, first performs computer vision correction, and then performs biological correction, so as to obtain the final accurate detection result of plankton. In the present invention, the image splicing plankton detection and correction method is the foundation, and the biological gene chip detection data is used as a supplementary means.

[0110] Embodiment 2:

[0111] On the basis of the above-mentioned plankton detection method based on image processing and biological information correction, the present invention also provides a plankton detection system based on image processing and biological information correction.

[0112] like Figure 2 As shown, a plankton detection system based on image processing and biological information correction includes a processor, a computer-readable storage medium, and a computer program or instruction stored in the computer-readable storage medium. When the computer program or instruction is executed by the processor, the plankton detection method based on image processing and biological information correction as described above is implemented.

[0113] The plankton detection method and system based on image processing and biological information correction of the present invention are based on computer vision detection, which can avoid the limitation of excessive workload and statistical data deviation caused by biotechnology detection; at the same time, in the process of computer vision detection, the two-dimensional image is displayed in a full-view three-dimensional microscopic world to present a more straightforward presentation of individual plankton to the user; and in the process of displaying the full-view three-dimensional microscopic world, the detection results of the panoramic composite image are used to correct the single-view detection results, avoiding the problem of misdetection of cut plankton due to incomplete morphology, thereby effectively increasing the accuracy of plankton image recognition; finally, the panoramic detection result image is biologically corrected using the standard detection results obtained by the biological detection method, so as to obtain the final accurate plankton detection result.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A plankton detection method based on image processing and biological information correction, characterized in that: The following steps are included: S1, making a glass slide of a water sample to be tested containing plankton, and performing a multi-view three-dimensional scanning on the glass slide to obtain a multi-view original image sequence; S2, using a deep learning method to perform plankton detection based on image recognition on each frame of the original image sequence of each field of view, to obtain a detection result image sequence of each field of view; S3, performing image stitching on each visual field detection result image sequence by pyramid fusion to obtain the final detection result image of each visual field; S4, performing image stitching on the final detection result images of all fields of view using pyramid fusion to obtain a panoramic detection result image based on computer vision; S5, determining whether there are incomplete plankton at the image stitching edge of each field of view in the panoramic detection result image, if yes, executing S6 and S7, if no, executing S7; S6, performing image stitching on each field of view original image sequence by using pyramid fusion to obtain a composite image for each field of view, and performing image stitching on all field of view composite images by using pyramid fusion to obtain a panoramic composite image; Based on the presence of incomplete plankton vision at the edge of the panoramic detection result image, a plankton detection based on image recognition is performed on a preset area of ​​the panoramic composite image using a deep learning method to obtain a secondary detection result image; Correcting the panoramic detection result image using the secondary detection result image to obtain a corrected panoramic detection result image; S7, performing plankton detection on the glass slide based on a biological detection method to obtain a standard detection result, and using the standard detection result to perform biological correction on the panoramic detection result image to obtain a final plankton detection result.

2. The method for detecting plankton based on image processing and biological information correction according to claim 1, characterized in that: In S3, pyramid fusion is used to perform image stitching on any visual field detection result image sequence, specifically, S31, calculating the weight of each frame image in any visual field detection result image sequence, and selecting one frame image from the visual field detection result image sequence as a base image, and the remaining images as non-base images; S32, setting the confidence of the area where plankton exists in the bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the bottom plate image to 1, to obtain an initial weighted image of the bottom plate; S33, setting the confidence of the area where plankton exists in the non-bottom plate image to the weight, and setting the confidence of the area where plankton does not exist in the non-bottom plate image to 0, to obtain a non-bottom plate initial weight image; S34, performing Gaussian filtering on the bottom plate initial weight image and the non-bottom plate initial weight image respectively, to obtain a bottom plate weight image and a non-bottom plate weight image respectively; S35, taking the bottom plate weighted image as a reference, using a pyramid fusion method to fuse all non-bottom plate weighted images located on different focal planes into the bottom plate weighted image, to obtain a final detection result image of the field of view.

3. The method for detecting plankton based on image processing and biological information correction according to claim 2, characterized in that: In S3, each field of view final detection result image includes a plankton final detection result, and the plankton final detection result is queried in the corresponding field of view final detection result image by full-text searching the bottom plate weight image.

4. The method for detecting plankton based on image processing and biological information correction according to claim 1, characterized in that: Before S4, the method further includes performing white balance processing on the final detection result image of each field of view respectively; In S4, the final detection result images of all visual fields after white balance processing are spliced ​​by pyramid fusion according to the XY axis position information.

5. The method for detecting plankton based on image processing and biological information correction according to claim 4, characterized in that: In S4, if the sizes of the two frames of images to be stitched are inconsistent during the image stitching process, the overlapping area of ​​the two frames of images to be stitched is first calculated, and then the image canvas is expanded according to the overlapping area, and finally the image stitching is performed using pyramid fusion.

6. The method for detecting plankton based on image processing and biological information correction according to claim 1, characterized in that: In S6, the preset area corresponds to the middle area of ​​a stitching unit formed by the fields of view on both sides of the stitching edge of the image where incomplete plankton exists in the panoramic detection result image; The panoramic detection result image is corrected using the secondary detection result image, specifically, Extracting a detection frame from the secondary detection result image; Comparing the detection frame in the secondary detection result image with the panoramic detection result image; If the detection frame in the secondary detection result image has an overlapping area with the panoramic detection result image, the panoramic detection result image is corrected based on the secondary detection result image.

7. The method for detecting plankton based on image processing and biological information correction according to claim 1, characterized in that: In said S1, the water sample to be tested containing plankton is made into a glass slide by, Adding a water sample to be tested containing plankton into a gene probe chip made of a transparent medium; After adding internal reference nucleic acid to the gene probe chip to which the water sample to be tested containing plankton is added, DNA fluorescence labeling is performed; The gene probe chip after DNA fluorescence labeling is used as the glass slide.

8. The method for detecting plankton based on image processing and biological information correction according to claim 7, characterized in that: In S7, plankton detection is performed on the glass slide based on a biological detection method, specifically, Performing DNA chip hybridization on the gene probe chip after DNA fluorescence labeling to obtain a DNA hybridization chip; Scanning the DNA hybridization chip to obtain chip imaging results; The standard detection result is obtained by calculating the fluorescence signal intensity and position information in the chip imaging result.

9. The method for detecting plankton based on image processing and biological information correction according to claim 1, characterized in that: In S7, biological correction is performed on the panoramic detection result image using the standard detection result, specifically, Comparing the standard detection result with the detection result included in the panoramic detection result image based on the same position; When the standard detection result conflicts with the detection result included in the panoramic detection result image at the same position, the detection result in the panoramic detection result image is corrected based on the standard detection result at that position.

10. A plankton detection system based on image processing and biological information correction, characterized in that: It includes a processor, a computer-readable storage medium, and a computer program or instruction stored in the computer-readable storage medium. When the computer program or instruction is executed by the processor, the plankton detection method based on image processing and biological information correction as described in any one of claims 1 to 9 is implemented.

Images