CN115861308B - Acer truncatum disease detection method - Google Patents
Acer truncatum disease detection method Download PDFInfo
- Publication number
- CN115861308B CN115861308B CN202310145539.5A CN202310145539A CN115861308B CN 115861308 B CN115861308 B CN 115861308B CN 202310145539 A CN202310145539 A CN 202310145539A CN 115861308 B CN115861308 B CN 115861308B
- Authority
- CN
- China
- Prior art keywords
- plaque
- area
- acer truncatum
- value
- edge
- 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
- 241000219226 Acer truncatum Species 0.000 title claims abstract description 103
- 201000010099 disease Diseases 0.000 title claims abstract description 33
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 33
- 238000001514 detection method Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 118
- 241000221785 Erysiphales Species 0.000 claims abstract description 79
- 238000000034 method Methods 0.000 claims abstract description 46
- 238000012545 processing Methods 0.000 claims abstract description 20
- 230000009467 reduction Effects 0.000 claims abstract description 11
- 230000002159 abnormal effect Effects 0.000 claims description 34
- 239000011159 matrix material Substances 0.000 claims description 29
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000003708 edge detection Methods 0.000 claims description 6
- 238000005286 illumination Methods 0.000 abstract description 8
- 230000009286 beneficial effect Effects 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 5
- 230000002411 adverse Effects 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 14
- 238000004458 analytical method Methods 0.000 description 7
- 238000012937 correction Methods 0.000 description 5
- 239000003086 colorant Substances 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 230000012010 growth Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000011946 reduction process Methods 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 241000221696 Sclerotinia sclerotiorum Species 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000005080 plant death Effects 0.000 description 1
- 230000008635 plant growth Effects 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image data processing, in particular to a method for detecting diseases of acer truncatum, which comprises the following steps: acquiring each plaque area in the acer truncatum leaves to be detected, and determining the fluctuation degree and the filling degree of each plaque area according to the image characteristic information of each plaque area; correcting weight coefficients corresponding to any two initial feature vectors obtained in advance by using fluctuation degree and filling degree to obtain corrected weight coefficients, and further realizing feature vector dimension reduction processing by using the corrected weight coefficients to obtain each first feature vector; and classifying each first feature vector by using a nearest neighbor classifier to obtain a classification result, and further judging whether powdery mildew exists in the acer truncatum leaves to be detected. The invention increases the characteristic difference between the powdery mildew area and the facula area, is beneficial to reducing the adverse effect of illumination factors on the detection of the Acer truncatum diseases and improves the accuracy of the detection of the Acer truncatum diseases.
Description
Technical Field
The invention relates to the technical field of image data processing, in particular to a method for detecting diseases of acer truncatum.
Background
Acer truncatum, also known as Acer truncatum, is an excellent landscaping tree species. In the growth period, powdery mildew often appears on Acer truncatum, so that big spots with inconspicuous edges appear on the back of leaves in the growth process, and white mould layers appear in severe cases, so that the Acer truncatum leaves earlier. The powdery mildew on Acer truncatum plants is harmful to the plant when the plant growth is poor, and the plant death is caused when the damage is serious.
For powdery mildew detection of Acer truncatum plants in Acer truncatum forest, the acquired leaf images are converted into HIS space from RGB space, an OSTU (maximum inter-class variance threshold segmentation algorithm) threshold segmentation method is utilized to calculate an H channel histogram to obtain a local optimal threshold, the local optimal threshold is utilized to carry out binarization and normalization processing on the leaf images to obtain feature vectors, and a classifier is utilized to classify the feature vectors to realize disease identification. According to the method, the characteristic vector of the blade image is extracted through the color information of the blade, but the blade image is influenced by natural conditions such as illumination and the like when the image is acquired, white light spots appear in the blade image, and the white light spots are quite similar to the image characteristics of powdery mildew, so that the disease is identified independently from the color information in the blade image, the situation of false identification is easy to appear, and the detection accuracy of the Acer truncatum disease is low.
Disclosure of Invention
In order to solve the technical problem of low detection accuracy of Acer truncatum diseases, the invention aims to provide an Acer truncatum disease detection method, which adopts the following technical scheme:
the embodiment of the invention provides a method for detecting diseases of Acer truncatum, which comprises the following steps:
obtaining a surface image of a acer truncatum leaf to be detected, and carrying out plaque detection on the surface image to obtain each plaque area in the acer truncatum leaf to be detected, wherein the plaque areas are areas containing powdery mildew or light spots;
obtaining edge gradient sequences corresponding to the plaque areas according to the plaque areas, and determining fluctuation degrees corresponding to the plaque areas according to the edge gradient sequences; determining the filling degree corresponding to each plaque area according to the pixel points in each plaque area;
determining initial feature vectors of all the plaque areas according to pixel values of all the pixel points in all the plaque areas; correcting weight coefficients corresponding to the initial feature vectors of any two plaque areas obtained in advance according to the fluctuation degree and the filling degree corresponding to each plaque area, and obtaining corrected weight coefficients;
and carrying out feature vector dimension reduction processing on each initial feature vector according to the corrected weight coefficient to obtain each first feature vector, classifying each first feature vector by utilizing a nearest neighbor classifier to obtain a classification result, and judging whether powdery mildew exists in the acer truncatum leaves to be detected according to the classification result.
Further, determining the fluctuation degree corresponding to each plaque region according to the edge gradient sequence comprises the following steps:
according to the target gradient value of each pair of adjacent edge pixel points in the edge gradient sequence corresponding to each plaque area, calculating a target gradient difference value corresponding to each pair of adjacent edge pixel points, and obtaining a differential sequence corresponding to each plaque area according to the target gradient difference value;
calculating the sum of the absolute value of each differential value and the value 1 in the differential sequence, taking each differential value as a molecule of a ratio, taking the sum of the absolute value of each differential value and the value 1 as a denominator of the ratio, and determining the ratio as an edge fluctuation index of the corresponding differential value;
counting the number of edge fluctuation indexes of adjacent differential values in the differential sequence corresponding to each plaque region, and determining the number as the fluctuation number of the corresponding plaque region, wherein the preset numerical condition is that one edge fluctuation index is positive and the other edge fluctuation index is negative in the edge fluctuation indexes of the adjacent differential values;
and calculating the ratio of the fluctuation quantity of each plaque area to the quantity of the difference values in the corresponding difference sequence, and determining the ratio as the fluctuation degree corresponding to the corresponding plaque area.
Further, plaque detection is carried out on the surface image, so that each plaque area in the acer truncatum leaves to be detected is obtained, and the method comprises the following steps:
clustering the surface images to obtain two cluster types, calculating average pixel values corresponding to the two cluster types according to pixel values of all pixel points in the two cluster types, determining the pixel points in the cluster type corresponding to the larger average pixel value as abnormal pixel points, and determining the pixel points in the cluster type corresponding to the smaller average pixel value as normal pixel points;
and carrying out connected domain processing on each abnormal pixel point in the surface image to obtain each connected domain in the surface image, and taking a closed region formed by the connected domains as a plaque region to obtain each plaque region in the acer truncatum leaves to be detected.
Further, determining, according to the pixel points in each plaque area, a filling degree corresponding to each plaque area includes:
counting the number of normal pixels and the number of abnormal pixels in each patch area, calculating the number difference between the number of abnormal pixels and the number of normal pixels in each patch area, taking the number difference as a ratio molecule, taking the number of abnormal pixels in each patch area as a denominator of the ratio, and taking the ratio as the filling degree corresponding to each patch area.
Further, the calculation formula of the corrected weight coefficient is as follows:
wherein ,initial feature vector for the ith patch area and initial feature vector for the jth patch areaThe corrected weight coefficient corresponding to the symptom vector,for the degree of fluctuation corresponding to the i-th plaque region,for the extent of fluctuation corresponding to the jth plaque region,to pair(s)The absolute value is obtained and the absolute value is calculated,for the filling level corresponding to the i-th plaque region,for the filling level corresponding to the jth patch area,to pair(s)The absolute value is obtained and the absolute value is calculated,the initial feature vector of the ith patch area and the weight coefficient corresponding to the initial feature vector of the jth patch area are obtained.
Further, obtaining an edge gradient sequence corresponding to each plaque region according to each plaque region, including:
performing edge detection on each plaque area to obtain each edge pixel point corresponding to each plaque area, calculating the gradient value of each edge pixel point in the direction outside the plaque area, taking the gradient value in the direction outside the plaque area as a target gradient value, obtaining the target gradient value of each edge pixel point corresponding to each plaque area, and taking a sequence formed by the target gradient values of each edge pixel point as an edge gradient sequence corresponding to the corresponding plaque area.
Further, judging whether powdery mildew exists on the acer truncatum leaves to be detected according to the classification result, wherein the method comprises the following steps:
if the first characteristic vector of the powdery mildew exists in the classification result, judging that the powdery mildew exists in the acer truncatum leaves to be detected, otherwise, judging that the powdery mildew does not exist in the acer truncatum leaves to be detected.
Further, determining an initial feature vector of each patch area according to the pixel value of each pixel point in each patch area includes:
constructing a minimum circumscribed rectangle of each plaque area, and determining a feature matrix of each plaque area according to the minimum circumscribed rectangle, wherein each numerical value in the feature matrix is a pixel value of a pixel point at a corresponding position;
and extracting a feature vector of the feature matrix according to each numerical value in the feature matrix, and determining the feature vector as an initial feature vector of the corresponding plaque area.
The invention has the following beneficial effects:
the invention provides a method for detecting diseases of acer truncatum, which comprises the steps of carrying out plaque detection on surface images of acer truncatum leaves to obtain plaque areas, wherein the plaque areas in the acer truncatum leaves are beneficial to improving the efficiency and accuracy of extracting powdery mildew image features. Since the image representation characteristic of powdery mildew is a plaque area with an unobvious edge, in order to analyze the fitting degree of the image characteristic of each plaque area and the powdery mildew image characteristic, the corresponding fluctuation degree of each plaque area, namely the uniformity degree of the plaque area edge gradient, is determined based on the edge gradient sequence corresponding to the plaque area; in order to further analyze the powdery mildew degree of each plaque area, the filling degree capable of representing the powdery mildew is obtained according to the distribution condition of abnormal pixel points in the plaque area, and the filling degree of each plaque area can reduce the adverse effect of unstable illumination on the surface image of the Acer truncatum leaves; and the powdery mildew degree of each plaque area is analyzed from two angles, so that the detection accuracy of the follow-up powdery mildew defect is improved. The powdery mildew areas and the facula areas in the surface images of the acer truncatum leaves have similar pixel colors, in order to enhance the differentiation characteristics between the initial characteristic vectors corresponding to the powdery mildew areas and the initial characteristic vectors corresponding to the facula areas, the accuracy of detecting the acer truncatum diseases is improved, the powdery mildew degrees of all plaque areas are utilized to correct the weight coefficients corresponding to the initial characteristic vectors corresponding to any two plaque areas, the weight coefficients are fully combined with the actual powdery mildew degrees of all the plaques, and the association degree of the weight coefficients relative to the detection of the acer truncatum diseases is improved. The corrected weight coefficient is utilized to realize feature vector dimension reduction processing, each first feature vector is obtained, the calculated amount of disease detection is effectively reduced, each first feature vector is classified by a nearest neighbor classifier, and a powdery mildew area and a facula area can be accurately distinguished; and judging whether powdery mildew exists on the acer truncatum leaves to be detected according to an accurate classification result, so that misjudgment is reduced, and the detection precision and accuracy of the acer truncatum diseases are improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for detecting Acer truncatum diseases according to the present invention;
fig. 2 is a surface image of a acer truncatum leaf in an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purpose, the following detailed description is given below of the specific implementation, structure, features and effects of the technical solution according to the present invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
In the process of detecting powdery mildew of acer truncatum leaves, the acer truncatum leaves are affected by sunlight, spot areas can appear in the acer truncatum leaves, and pixel points in the spot areas and pixel points in the powdery mildew areas have the same color. The traditional method can realize powdery mildew detection based on the feature vector extracted from the Acer truncatum leaf image, but the method can lead powdery mildew areas and facula areas with similar pixel values not to be accurately distinguished, and seriously reduces the accuracy rate of Acer truncatum disease detection. In order to improve the accuracy of detecting Acer truncatum diseases, the difference between the stretching facula area and the powdery mildew area is stretched, and the embodiment provides an Acer truncatum disease detection method, as shown in fig. 1, comprising the following steps:
s1, obtaining a surface image of a acer truncatum leaf to be detected, carrying out plaque detection on the surface image to obtain each plaque area in the acer truncatum leaf to be detected, and the method comprises the following steps:
first, obtaining a surface image of a acer truncatum leaf to be detected.
According to the embodiment, when the surface image of the Acer truncatum leaves to be detected is acquired, the Acer truncatum leaves in the Acer truncatum forest are acquired through overlooking angles by using the high-definition camera of the unmanned aerial vehicle, and the flight route of the unmanned aerial vehicle is required to be planned in advance in the acquisition process. According to the flight route of the unmanned aerial vehicle, the surface image of the Acer truncatum leaves is obtained in real time, the surface image at the current moment is used as the surface image of the Acer truncatum leaves to be detected, and the surface image of the Acer truncatum leaves is shown in figure 2.
So far, the reference image for detecting powdery mildew is obtained by acquiring the images of the Acer truncatum leaves in the Acer truncatum forest by the unmanned aerial vehicle.
And secondly, carrying out plaque detection on the surface image of the acer truncatum leaves to be detected to obtain each plaque area in the acer truncatum leaves to be detected.
Clustering the surface images to obtain two cluster types, calculating average pixel values corresponding to the two cluster types according to pixel values of all pixel points in the two cluster types, determining the pixel points in the cluster type corresponding to the larger average pixel value as abnormal pixel points, and determining the pixel points in the cluster type corresponding to the smaller average pixel value as normal pixel points; and carrying out connected domain processing on each abnormal pixel point in the surface image to obtain each connected domain in the surface image, and taking a closed region formed by the connected domains as a plaque region to obtain each plaque region in the acer truncatum leaves to be detected.
First, clustering is performed on the surface image to obtain average pixel values corresponding to two clusters.
In this embodiment, clustering analysis is performed on the surface images of the acer truncatum leaves to be detected by using an FCMA (fuzzy C-means algorithm based on fuzzy local information C-means clustering method) clustering method with the number of clusters k=2, so as to obtain two clusters after segmentation, wherein the implementation process of the FCMA clustering method is the prior art and is not described in detail herein. Meanwhile, the surface image of the acer truncatum leaves to be detected is subjected to graying treatment by using a weighted average method, so that the pixel value of each pixel point in the gray image of the acer truncatum leaves to be detected is obtained, wherein the pixel value is the gray value, and the implementation process of the weighted average method is the prior art and is not explained in detail herein. And mapping all pixel points in the two clusters into the gray level image, so that pixel values of all pixel points in the two clusters can be obtained, further calculating average pixel values corresponding to the two clusters, and calculating the average pixel values to facilitate the subsequent obtaining of each abnormal pixel point in the acer truncatum leaves to be detected.
It should be noted that, when dividing all the pixels in the surface image of the acer truncatum leaves to be detected, the FCMA clustering method can determine the cluster membership degree corresponding to each central pixel according to the cluster membership degree corresponding to the adjacent pixels of each central pixel, that is, if a plurality of adjacent pixels of a certain central pixel are abnormal pixels, then the central pixel should be also the abnormal pixel. Compared with other existing clustering methods, the FCMA clustering method can retain local information, is beneficial to improving the dividing accuracy of two divided clusters, and is more suitable for clustering of the surface images of the Acer truncatum leaves.
And then, according to the average pixel values corresponding to the two clusters, obtaining each abnormal pixel point in the acer truncatum leaves to be detected.
It should be noted that, the colors of the normal areas of the acer truncatum leaves in the same growth cycle are the same, and the appearance characteristic of the powdery mildew of the acer truncatum leaves is that a large number of powdery mildew defective pixels exist on the acer truncatum leaves, the powdery mildew defective pixels are greatly different from the pixels in the normal areas, in order to analyze whether the powdery mildew defective pixels exist in the acer truncatum leaves to be detected, the pixels with similar colors are clustered by means of clustering the pixels on the surface images, and the pixels are divided into normal pixels and abnormal pixels.
In this embodiment, average pixel values corresponding to two clusters are compared, and in combination with the image characteristics of powdery mildew, the pixel points in the cluster corresponding to the larger average pixel value are determined to be abnormal pixel points, the pixel points in the cluster corresponding to the smaller average pixel value are determined to be normal pixel points, wherein the abnormal pixel points refer to powdery mildew pixel points or facula pixel points, the pixel values of the powdery mildew pixel points and the facula pixel points are both larger, and the normal pixel points refer to pixel points in a normal area which are not affected by powdery mildew and illumination.
And finally, according to each abnormal pixel point in the acer truncatum leaves to be detected, each plaque area in the acer truncatum leaves to be detected is obtained.
In this embodiment, the connected regions of each abnormal pixel point in the acer truncatum leaves to be detected are processed to obtain each connected region in the surface image, and the closed regions formed by the connected regions are used as plaque regions, so that each plaque region in the acer truncatum leaves to be detected can be obtained, wherein the plaque regions are regions containing powdery mildew or light spots, and the plaque regions refer to multiple connected regions in which part surrounded by a closed curve does not all belong to the abnormal pixel points. For the ith plaque area in the acer truncatum leaves to be detected,the number of all abnormal pixel points in the ith patch area is recorded asThe number of all normal pixel points in the ith patch area is recorded asThe implementation process of the connected domain processing is the prior art, and is not in the scope of the present invention, and will not be described in detail here.
S2, obtaining edge gradient sequences corresponding to the plaque areas according to the plaque areas, and determining fluctuation degrees corresponding to the plaque areas according to the edge gradient sequences; according to the pixel points in each plaque area, determining the filling degree corresponding to each plaque area, wherein the method comprises the following steps:
and a first step of obtaining an edge gradient sequence corresponding to each plaque region according to each plaque region.
Performing edge detection on each plaque area to obtain each edge pixel point corresponding to each plaque area, calculating the gradient value of each edge pixel point in the direction outside the plaque area, taking the gradient value in the direction outside the plaque area as a target gradient value, obtaining the target gradient value of each edge pixel point corresponding to each plaque area, and taking a sequence formed by the target gradient values of each edge pixel point as an edge gradient sequence corresponding to the corresponding plaque area.
In this embodiment, the Sobel edge detection operator is used to perform edge detection on each patch area, so as to obtain each edge pixel point corresponding to each patch area, and the implementation process of the Sobel edge detection operator is in the prior art and will not be described in detail herein. Since the edge pixel points are similar to the pixel points in the plaque area, the gradient values of the edge pixel points in the direction of the interior of the plaque area are calculated, the gradient values of the edge pixel points are lower and have no reference value, so the gradient values of the edge pixel points in the direction of the exterior of the plaque area are extracted in a clockwise direction, the gradient values in the direction of the exterior of the plaque area are taken as target gradient values, the target gradient values of the edge pixel points corresponding to the plaque areas are obtained, and the plaque areas are pairedThe target gradient value of the corresponding jth edge pixel point is recorded as. The target gradient values of the edge pixel points corresponding to each plaque area can form an edge gradient sequence, so that the edge gradient sequence corresponding to each plaque area is obtained, and the edge gradient sequence is beneficial to the follow-up analysis of the fluctuation degree corresponding to each plaque area.
It should be noted that, the calculating process of the gradient value of the edge pixel point is the prior art, and is not in the scope of the present invention, and the gradient value is extracted in the clockwise direction for facilitating the subsequent analysis and calculation, however, the practitioner may also extract the gradient value in the counterclockwise direction.
And secondly, determining the fluctuation degree corresponding to each plaque area according to the edge gradient sequence.
It should be noted that powdery mildew can cause the acer truncatum leaves to have plaques with insignificant edges, so the image characteristics based on powdery mildew defects can show that if a certain plaque area is a powdery mildew area, the gradient value fluctuation of the edges of the plaque area can be larger; based on the image features of the spot areas, if a certain spot area is a spot area, the gradient value fluctuation of the edge of the spot area is smaller, and the gradient value distribution is more uniform. Therefore, by analyzing the fluctuation degree corresponding to each plaque region through the uniformity degree of the edge gradient sequence corresponding to each plaque region, the fluctuation degree can represent the powdery mildew degree of the plaque region, and the method comprises the following steps:
firstly, calculating a target gradient difference value corresponding to each pair of adjacent edge pixel points in an edge gradient sequence corresponding to each plaque region according to the target gradient value of each pair of adjacent edge pixel points in the edge gradient sequence corresponding to each plaque region, and obtaining a differential sequence corresponding to each plaque region according to the target gradient difference value.
In this embodiment, the calculation formula of the target gradient difference value corresponding to each pair of adjacent edge pixel points may be:
wherein ,for the target gradient difference value corresponding to the n-th edge pixel point and the n+1-th edge pixel point in the edge gradient sequence corresponding to each patch area, the n-th edge pixel point and the n+1-th edge pixel point are a pair of adjacent edge pixel points,the target gradient value of the (n+1) th edge pixel point in the edge gradient sequence corresponding to each plaque region,and the target gradient value of the nth edge pixel point in the edge gradient sequence corresponding to each plaque area.
In the calculation formula of the target gradient difference value,the difference condition of the target gradient values corresponding to the adjacent edge pixel points can be represented, and each target gradient difference value corresponding to each plaque areaThe composed sequence may be a differential sequence, with each plaque region having its corresponding differential sequence. The values in the differential sequence are differential values, and the target gradient difference value is the differential value, wherein the differential value can represent the change condition of the target gradient values of the front and rear adjacent edge pixel points.
Then, the sum of the absolute value of each differential value and the value 1 in the differential sequence is calculated, each differential value is taken as a molecule of the ratio, the sum of the absolute value of each differential value and the value 1 is taken as a denominator of the ratio, and the ratio is determined as an edge fluctuation index of the corresponding differential value.
In this embodiment, when the number of times of variation of the sign value in the differential sequence is relatively large, it is indicated that the edge of the plaque area corresponding to the differential sequence has relatively large fluctuation, and the plaque area corresponding to the differential sequence has relatively high possibility of powdery mildew defect. For the positive and negative relationship of each differential value in the differential sequence, the positive and negative relationship is obtained through 1 norm of each differential value, namely the differential value is divided by the sum of its own absolute value and the value 1, wherein in order to avoid the special case that the score denominator is 1, the representation form of the edge fluctuation index can be "+" or "-", and the calculation formula of the edge fluctuation index can be:
wherein ,for the mth edge fluctuation index in the differential sequence corresponding to each plaque region,for the mth differential value in the differential sequence corresponding to each plaque region,for a pair ofThe absolute value is determined.
And then, counting the number of edge fluctuation indexes of adjacent differential values in the differential sequence corresponding to each plaque region, which meet the preset value condition, and determining the number as the fluctuation number of the corresponding plaque region.
In this embodiment, according to the edge fluctuation index of the adjacent difference value in the difference sequence corresponding to each plaque region, the number of the adjacent edge fluctuation indexes satisfying the preset value condition is counted, where one edge fluctuation index is positive and the other edge fluctuation index is negative in the edge fluctuation indexes of the adjacent difference value. It should be noted that, when the pair of adjacent edge fluctuation indexes meets the preset numerical condition, the positive-negative relationship of the pair of adjacent edge fluctuation indexes is not divided, that is, the first edge fluctuation index of the pair of adjacent edge fluctuation indexes may be positive, the second edge fluctuation index may be negative, and likewise, the first edge fluctuation index of the pair of adjacent edge fluctuation indexes may be negative, and the second edge fluctuation index may be positive.
For example, establish a size ofAnd sliding the sliding window in the differential sequence corresponding to each patch area, wherein the sliding step length is 1, and based on the edge fluctuation index of each differential value in the differential sequence, the accumulator is added with 1 each time when the edge fluctuation index of a negative differential value and a positive differential value appears in the sliding window, until the sliding window is completely slid, the accumulator value of the differential sequence corresponding to each patch area can be obtained, and the accumulator value is used as the fluctuation quantity of the corresponding patch area.
And finally, calculating the ratio of the fluctuation quantity of each plaque area to the quantity of the difference values in the corresponding difference sequence, and determining the ratio as the fluctuation degree corresponding to the corresponding plaque area.
In this embodiment, each two adjacent differential values in the differential sequence have only one fluctuation state, so the fluctuation number of each plaque area is necessarily smaller than the number of the differential values in the differential sequence corresponding to the corresponding plaque area, and the fluctuation degree of the current plaque area can be represented by the ratio of the fluctuation number to the number of the differential values, and the calculation formula can be as follows:
wherein ,to the extent of the fluctuation of the a-th plaque area,for the number of undulations of the a-th plaque area,for the difference corresponding to the a-th plaque regionNumber of differential values in the sequence.
It should be noted that, the fluctuation degree of the plaque area may represent the fluctuation condition of the edge pixel points of the plaque area, and in the subsequent powdery mildew detection and analysis process, the fluctuation condition of the edge pixel points of each plaque area may measure whether each plaque area is caused by powdery mildew or by illumination spots.
And thirdly, determining the filling degree corresponding to each plaque area according to the pixel points in each plaque area.
Counting the number of normal pixels and the number of abnormal pixels in each patch area, calculating the number difference between the number of abnormal pixels and the number of normal pixels in each patch area, taking the number difference as a ratio molecule, taking the number of abnormal pixels in each patch area as a denominator of the ratio, and taking the ratio as the filling degree corresponding to each patch area.
First, when the Acer truncatum leaves are damaged by insect damage, the Acer truncatum leaves are affected by unstable illumination characteristics, and when illumination passes through the damaged areas of the leaves, the leaves below the damaged areas of the leaves may also be affected by light spots with strong edge fluctuation. Therefore, the fluctuation degree of each plaque area is determined, and the filling degree of abnormal pixel points in each plaque area is considered, so that the forming reason of each plaque area is analyzed from the two angles of the fluctuation degree and the filling degree, and the powdery mildew degree of each plaque area is obtained.
In this embodiment, the filling degree of the patch area is determined by the number of normal pixel points and the number of abnormal pixel points in the patch area. When the number of normal pixels in the closed area formed by the plaque area is larger, the more uneven the abnormal pixel distribution in the plaque area is, the more uniform the abnormal pixel distribution in the plaque area needs to be analyzed on the basis of analyzing powdery mildew by fluctuation degree, and the calculation formula of the filling degree corresponding to each plaque area can be as follows:
wherein ,for the filling level corresponding to the i-th plaque region,for the number of outlier pixels in the ith patch area,for the number of normal pixels in the ith patch area,the difference value between the number of abnormal pixels and the number of normal pixels in the ith patch area is obtained.
In the calculation formula of the filling degree, the number of normal pixel points in the ith patch areaThe larger the number difference between the number of abnormal pixels and the number of normal pixels in the ith patch areaThe smaller the i-th plaque region will be, the corresponding filling levelThe smaller will be.
Thus, the fluctuation degree corresponding to each plaque area can be obtained through analysis of the fluctuation condition of the pixel points at the upper edge of each plaque area, and the filling degree corresponding to each plaque area can be obtained through analysis of the filling degree of the abnormal pixel points in each plaque area. The higher the fluctuation degree and the filling degree corresponding to a certain plaque area, the more accords with the image characteristics of powdery mildew, namely the fluctuation degree and the filling degree can be used as powdery mildew degree indexes of acer truncatum leaves, and can be used for correcting the characteristic vectors extracted subsequently, so that the distance between the leaf area with light spots and the leaf area with powdery mildew is increased, and the accuracy of detecting the acer truncatum diseases is improved.
S3, determining initial feature vectors of all the plaque areas according to pixel values of all the pixel points in all the plaque areas; correcting the weight coefficient corresponding to the initial feature vector of any two plaque areas obtained in advance according to the corresponding fluctuation degree and filling degree of each plaque area to obtain the corrected weight coefficient, wherein the method comprises the following steps:
first, determining initial feature vectors of each plaque area according to pixel values of each pixel point in each plaque area.
Constructing a minimum circumscribed rectangle of each plaque area, and determining a feature matrix of each plaque area according to the minimum circumscribed rectangle, wherein each numerical value in the feature matrix is a pixel value of a pixel point at a corresponding position; and extracting a feature vector of the feature matrix according to each numerical value in the feature matrix, and determining the feature vector as an initial feature vector of the corresponding plaque area.
In this embodiment, the process of constructing the minimum bounding rectangle of each plaque area is the prior art, and will not be described in detail here. Because the feature matrix is a square matrix, in order to facilitate the subsequent determination of the feature matrix corresponding to each plaque area, the minimum circumscribed rectangle is expanded to the size of the length value based on the length value of the rectangle, and the widths of the two sides are simultaneously expanded during expansion, so that the circumscribed square of each plaque area is obtained. According to the circumscribed square, the characteristic matrix can be obtained by analogy, each plaque area is provided with the corresponding characteristic matrix, each numerical value in each characteristic matrix is the pixel value of the pixel point at the corresponding position, and if the numerical value which does not belong to the corresponding plaque area exists in the characteristic matrix, the numerical value is assigned to be 0. Through the feature matrix corresponding to each plaque area, the feature vector corresponding to the feature matrix can be obtained by using the calculation method of the feature vector of the existing matrix, the number of the feature vector of each feature matrix is possibly 1 or 2, and one feature vector is randomly selected as the initial feature vector of the corresponding feature matrix. It should be noted that the eigenvectors of the eigenvector matrix are determined according to the dimension of the matrix, and the gray image matrix is a two-dimensional matrix, and the number of eigenvectors of each eigenvector matrix may be 1 or 2 because there is only one real solution.
Thus, initial feature vectors of the plaque areas are obtained, and the initial feature vectors can characterize image feature information of the plaque areas.
And secondly, correcting the weight coefficients corresponding to the initial feature vectors of any two plaque areas according to the corresponding fluctuation degree and filling degree of each plaque area, and obtaining corrected weight coefficients.
First, in the LPP (Locality Preserving Projections, partial reservation projection algorithm) projection algorithm, the objective function is to analyze the weight coefficient between one feature vector and another feature vector by the distance between the two feature vectors, and the two feature vectors are not identical but very similar. In order to facilitate the subsequent correction of the weight coefficient based on the powdery mildew degree corresponding to each plaque area, the distance between the feature vector corresponding to the powdery mildew area and the feature vector corresponding to the facula area is stretched, and the weight coefficient corresponding to the initial feature vector of any two plaque areas needs to be obtained, wherein the calculation formula can be as follows:
wherein ,the weight coefficient corresponding to the initial feature vector of the ith patch area and the initial feature vector of the jth patch area, e is a natural constant,for the initial feature vector of the i-th plaque region,for the initial feature vector of the jth plaque region, t is a super parameter, in this embodiment t=5,is thatAnd (3) withThe distance between the two plates is set to be equal,is based on natural constant eIs a negative correlation index of (a).
It should be noted that, the calculation formula of the weight coefficients corresponding to the two initial feature vectors is the existing formula, the super parameter t is set by human, and represents the bandwidth of the gaussian kernel function, and when the weight coefficient corresponding to the two initial feature vectors is 1, it is described that the two initial feature vectors are the same feature vector. For any two initial feature vectors in feature spaceAndprojection may be achieved by one initial feature vector between two initial feature vectors.
Then, according to the corresponding fluctuation degree and filling degree of each plaque region, the weight coefficient corresponding to the initial feature vector of any two plaque regions obtained in advance is corrected, and the corrected weight coefficient is obtained.
In this embodiment, the degree of fluctuation and the degree of filling may be used as correction factors in the feature dimension reduction process. In the process of performing feature dimension reduction processing on each initial feature vector by using an LPP projection algorithm, when the distance between the initial feature vectors of the powdery mildew area and the facula area is calculated, since any two initial feature vectors are quite similar, the powdery mildew area and the facula area cannot be effectively distinguished, and the powdery mildew area and the facula area are plaque areas. For distinguishing powdery mildew areas from facula areas, the weight coefficient of the distance between the powdery mildew areas and the facula areas in the projection process needs to be increased, namely, in order to stretch the distance between the powdery mildew areas and the initial feature vectors of the facula areas, local image information of the plaque areas is reserved, the weight coefficient needs to be corrected through two correction factors of fluctuation degree and filling degree, the corrected weight coefficient is obtained, and the calculation formula can be as follows:
wherein ,as the corrected weight coefficient corresponding to the initial feature vector of the ith patch area and the initial feature vector of the jth patch area,for the degree of fluctuation corresponding to the i-th plaque region,for the extent of fluctuation corresponding to the jth plaque region,to pair(s)The absolute value is obtained and the absolute value is calculated,for the filling level corresponding to the i-th plaque region,for the filling level corresponding to the jth patch area,to pair(s)The absolute value is obtained and the absolute value is calculated,the initial feature vector of the ith patch area and the weight coefficient corresponding to the initial feature vector of the jth patch area are obtained.
In the calculation formula of the corrected weight coefficient,the difference condition of the fluctuation degrees corresponding to the two plaque areas can be represented, and when the fluctuation degree difference corresponding to the two plaque areas is larger, the dimension-reduction weight coefficient corresponding to the initial feature vector of the two plaque areas is required to be increased, so that the two initial feature vectors are differentiated through powdery mildew image features of the plaque areas;the difference in the corresponding filling level of the two plaque areas can be characterized,influence on the initial feature vectorThe influence on the initial feature vector is the same, and the filling degree is further corrected on the basis of the fluctuation degree, so that the accuracy of the corrected weight coefficient is improved, and the adverse effect of unstable illumination on the extracted powdery mildew image features is reduced; the 1 in the calculation formula is to avoid the special case where the weight coefficient after correction is 0.
And S4, performing feature vector dimension reduction processing on each initial feature vector according to the corrected weight coefficient to obtain each first feature vector, classifying each first feature vector by using a nearest neighbor classifier to obtain a classification result, and judging whether powdery mildew exists in the acer truncatum leaves to be detected according to the classification result.
And performing feature vector dimension reduction processing on each initial feature vector according to the corrected weight coefficient to obtain each first feature vector, and classifying each first feature vector by using a nearest neighbor classifier to obtain a classification result.
In this embodiment, first, the corrected weight coefficient is applied to the LPP projection algorithm to obtain a new LPP projection algorithm, and the new LPP projection algorithm is used to perform feature vector dimension reduction processing on the initial feature vector corresponding to each plaque area to obtain a feature vector after dimension reduction processing, where the feature vector after dimension reduction processing is used as a first feature vector. The implementation process of the LPP projection algorithm is prior art and is not within the scope of the present invention, and is not described in detail herein. And classifying each first feature vector by utilizing a nearest neighbor classifier, and labeling corresponding labels for each first feature vector to obtain a final classification result of the acer truncatum leaves to be detected. The process of implementing the classification processing by the nearest neighbor classifier is the prior art, and is not in the scope of the present invention, and will not be described in detail here.
It should be noted that, any two initial feature vectors have their corresponding corrected weight coefficients, and the LPP projection algorithm can ensure that adjacent samples in each initial feature vector are still adjacent in each first feature vector, that is, the feature vector which is conducive to dimension reduction retains original adjacent information, and the retaining of the original adjacent information is conducive to the nearest neighbor classifier to accurately classify the leaf disease information of each first feature vector, and the leaf disease information is divided into a first feature vector with powdery mildew and a first feature vector without powdery mildew, that is, a plaque area with powdery mildew and a plaque area without powdery mildew.
And secondly, judging whether powdery mildew exists in the acer truncatum leaves to be detected according to the classification result.
In this embodiment, if the first feature vector whose label is powdery mildew exists in the final classification result, it is determined that powdery mildew exists in the acer truncatum leaves to be detected, otherwise, it is determined that powdery mildew does not exist in the acer truncatum leaves to be detected.
So far, the embodiment realizes the detection of the Acer truncatum diseases by processing the image data of the Acer truncatum leaf surface images.
The invention provides a method for detecting Acer truncatum diseases, which is characterized in that the method obtains the fluctuation degree and the filling degree of a weight coefficient which can be used for correcting a local maintenance projection algorithm by carrying out image data processing and analysis on the surface image of Acer truncatum leaves, and stretches powdery mildew areas and facula areas in the dimension reduction process based on two correction factors, namely changes initial feature vectors of corresponding plaque areas according to the disease degree of each plaque area, which is beneficial to updating the initial feature vectors of each plaque area based on the disease degree of the corresponding plaque area, thereby further improving the detection precision of the Acer truncatum diseases.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the scope of the embodiments of the present application, and are intended to be included within the scope of the present application.
Claims (5)
1. The method for detecting the Acer truncatum diseases is characterized by comprising the following steps of:
obtaining a surface image of a acer truncatum leaf to be detected, and carrying out plaque detection on the surface image to obtain each plaque area in the acer truncatum leaf to be detected, wherein the plaque areas are areas containing powdery mildew or light spots;
obtaining edge gradient sequences corresponding to the plaque areas according to the plaque areas, and determining fluctuation degrees corresponding to the plaque areas according to the edge gradient sequences; determining the filling degree corresponding to each plaque area according to the pixel points in each plaque area;
determining initial feature vectors of all the plaque areas according to pixel values of all the pixel points in all the plaque areas; correcting weight coefficients corresponding to the initial feature vectors of any two plaque areas obtained in advance according to the fluctuation degree and the filling degree corresponding to each plaque area, and obtaining corrected weight coefficients;
performing feature vector dimension reduction processing on each initial feature vector according to the corrected weight coefficient to obtain each first feature vector, classifying each first feature vector by using a nearest neighbor classifier to obtain a classification result, and judging whether powdery mildew exists on the acer truncatum leaves to be detected according to the classification result;
determining the fluctuation degree corresponding to each plaque area according to the edge gradient sequence, wherein the method comprises the following steps:
according to the target gradient value of each pair of adjacent edge pixel points in the edge gradient sequence corresponding to each plaque area, calculating a target gradient difference value corresponding to each pair of adjacent edge pixel points, and obtaining a differential sequence corresponding to each plaque area according to the target gradient difference value;
calculating the sum of the absolute value of each differential value and the value 1 in the differential sequence, taking each differential value as a molecule of a ratio, taking the sum of the absolute value of each differential value and the value 1 as a denominator of the ratio, and determining the ratio as an edge fluctuation index of the corresponding differential value;
counting the number of edge fluctuation indexes of adjacent differential values in the differential sequence corresponding to each plaque region, and determining the number as the fluctuation number of the corresponding plaque region, wherein the preset numerical condition is that one edge fluctuation index is positive and the other edge fluctuation index is negative in the edge fluctuation indexes of the adjacent differential values;
calculating the ratio of the fluctuation quantity of each plaque area to the quantity of the difference values in the corresponding difference sequence, and determining the ratio as the fluctuation degree corresponding to the corresponding plaque area;
determining initial feature vectors of each patch area according to pixel values of each pixel point in each patch area, wherein the initial feature vectors comprise:
constructing a minimum circumscribed rectangle of each plaque area, and determining a feature matrix of each plaque area according to the minimum circumscribed rectangle, wherein each numerical value in the feature matrix is a pixel value of a pixel point at a corresponding position;
extracting a feature vector of the feature matrix according to each numerical value in the feature matrix, and determining the feature vector as an initial feature vector of a corresponding plaque area;
the calculation formula of the corrected weight coefficient is as follows:
wherein ,for the initial feature vector of the ith patch area and the corrected weight coefficient corresponding to the initial feature vector of the jth patch area,/for the initial feature vector of the jth patch area>For the extent of fluctuation corresponding to the ith plaque area, < >>For the extent of fluctuation corresponding to the jth plaque region, < >>For->Absolute value is determined for->Filling degree corresponding to ith plaque area, < >>Filling degree corresponding to jth plaque area, < >>For->Absolute value is determined for->For the ith plaque regionThe initial feature vector of the jth patch region and the weight coefficient corresponding to the initial feature vector of the jth patch region.
2. The method for detecting diseases of acer truncatum according to claim 1, wherein plaque detection is performed on the surface image to obtain plaque areas in acer truncatum leaves to be detected, and the method comprises the following steps:
clustering the surface images to obtain two cluster types, calculating average pixel values corresponding to the two cluster types according to pixel values of all pixel points in the two cluster types, determining the pixel points in the cluster type corresponding to the larger average pixel value as abnormal pixel points, and determining the pixel points in the cluster type corresponding to the smaller average pixel value as normal pixel points;
and carrying out connected domain processing on each abnormal pixel point in the surface image to obtain each connected domain in the surface image, and taking a closed region formed by the connected domains as a plaque region to obtain each plaque region in the acer truncatum leaves to be detected.
3. The method for detecting diseases of acer truncatum according to claim 2, wherein determining the filling degree corresponding to each plaque area according to the pixel points in each plaque area comprises the following steps:
counting the number of normal pixels and the number of abnormal pixels in each patch area, calculating the number difference between the number of abnormal pixels and the number of normal pixels in each patch area, taking the number difference as a ratio molecule, taking the number of abnormal pixels in each patch area as a denominator of the ratio, and taking the ratio as the filling degree corresponding to each patch area.
4. The method for detecting diseases of acer truncatum according to claim 1, wherein the step of obtaining edge gradient sequences corresponding to each plaque region according to each plaque region comprises the steps of:
performing edge detection on each plaque area to obtain each edge pixel point corresponding to each plaque area, calculating the gradient value of each edge pixel point in the direction outside the plaque area, taking the gradient value in the direction outside the plaque area as a target gradient value, obtaining the target gradient value of each edge pixel point corresponding to each plaque area, and taking a sequence formed by the target gradient values of each edge pixel point as an edge gradient sequence corresponding to the corresponding plaque area.
5. The method for detecting a powdery mildew of acer truncatum leaves according to claim 1, wherein the step of judging whether powdery mildew exists in acer truncatum leaves to be detected according to classification results comprises the following steps:
if the first characteristic vector of the powdery mildew exists in the classification result, judging that the powdery mildew exists in the acer truncatum leaves to be detected, otherwise, judging that the powdery mildew does not exist in the acer truncatum leaves to be detected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310145539.5A CN115861308B (en) | 2023-02-22 | 2023-02-22 | Acer truncatum disease detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310145539.5A CN115861308B (en) | 2023-02-22 | 2023-02-22 | Acer truncatum disease detection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115861308A CN115861308A (en) | 2023-03-28 |
CN115861308B true CN115861308B (en) | 2023-05-12 |
Family
ID=85658591
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310145539.5A Active CN115861308B (en) | 2023-02-22 | 2023-02-22 | Acer truncatum disease detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115861308B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117437248B (en) * | 2023-12-19 | 2024-03-08 | 中国海洋大学 | Disease area segmentation method for blueberry quality detection |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104398271A (en) * | 2014-11-14 | 2015-03-11 | 西安交通大学 | Method using three-dimensional mechanics and tissue specific imaging of blood vessels and plaques for detection |
CN104598908A (en) * | 2014-09-26 | 2015-05-06 | 浙江理工大学 | Method for recognizing diseases of crop leaves |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106778786A (en) * | 2016-12-29 | 2017-05-31 | 西京学院 | Apple disease recognition methods based on log-spectral domain laminated gradient direction histogram |
CN108764275B (en) * | 2018-04-10 | 2020-08-04 | 甘肃农业大学 | Leaf disease identification method and system |
WO2020206666A1 (en) * | 2019-04-12 | 2020-10-15 | 深圳市汇顶科技股份有限公司 | Depth estimation method and apparatus employing speckle image and face recognition system |
CN111402152B (en) * | 2020-03-10 | 2023-10-24 | 北京迈格威科技有限公司 | Processing method and device of disparity map, computer equipment and storage medium |
-
2023
- 2023-02-22 CN CN202310145539.5A patent/CN115861308B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104598908A (en) * | 2014-09-26 | 2015-05-06 | 浙江理工大学 | Method for recognizing diseases of crop leaves |
CN104398271A (en) * | 2014-11-14 | 2015-03-11 | 西安交通大学 | Method using three-dimensional mechanics and tissue specific imaging of blood vessels and plaques for detection |
Also Published As
Publication number | Publication date |
---|---|
CN115861308A (en) | 2023-03-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106295124B (en) | The method of a variety of image detecting technique comprehensive analysis gene subgraph likelihood probability amounts | |
CN101630363B (en) | Rapid detection method of face in color image under complex background | |
CN110120042B (en) | Crop image pest and disease damage area extraction method based on SLIC super-pixel and automatic threshold segmentation | |
CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
CN108537751B (en) | Thyroid ultrasound image automatic segmentation method based on radial basis function neural network | |
CN112464942B (en) | Computer vision-based overlapped tobacco leaf intelligent grading method | |
CN110598560A (en) | Night monitoring and identifying method and system based on neural network enhancement | |
CN106340016A (en) | DNA quantitative analysis method based on cell microscope image | |
CN113723573A (en) | Tumor tissue pathological classification system and method based on adaptive proportion learning | |
CN110675421B (en) | Depth image collaborative segmentation method based on few labeling frames | |
CN115861308B (en) | Acer truncatum disease detection method | |
Naufal et al. | Preprocessed mask RCNN for parking space detection in smart parking systems | |
CN111199245A (en) | Rape pest identification method | |
CN115908371A (en) | Plant leaf disease and insect pest degree detection method based on optimized segmentation | |
CN115410096B (en) | Satellite remote sensing image multi-scale fusion change detection method, medium and electronic device | |
CN110458064B (en) | Low-altitude target detection and identification method combining data driving type and knowledge driving type | |
CN112183237A (en) | Automatic white blood cell classification method based on color space adaptive threshold segmentation | |
CN110827327B (en) | Fusion-based long-term target tracking method | |
CN107679467B (en) | Pedestrian re-identification algorithm implementation method based on HSV and SDALF | |
CN113344047A (en) | Platen state identification method based on improved K-means algorithm | |
CN106446832B (en) | Video-based pedestrian real-time detection method | |
CN113313678A (en) | Automatic sperm morphology analysis method based on multi-scale feature fusion | |
CN110866435B (en) | Far infrared pedestrian training method for self-similarity gradient orientation histogram | |
CN113095332A (en) | Saliency region detection method based on feature learning | |
CN112307894A (en) | Pedestrian age identification method based on wrinkle features and posture features in community monitoring scene |
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 |