CN111611910B - Yellow river ice dam image feature recognition method - Google Patents
Yellow river ice dam image feature recognition method Download PDFInfo
- Publication number
- CN111611910B CN111611910B CN202010422701.XA CN202010422701A CN111611910B CN 111611910 B CN111611910 B CN 111611910B CN 202010422701 A CN202010422701 A CN 202010422701A CN 111611910 B CN111611910 B CN 111611910B
- Authority
- CN
- China
- Prior art keywords
- mode
- yellow river
- ice dam
- image
- river ice
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
The invention discloses a yellow river ice dam image feature recognition method, which comprises the following steps: s1, acquiring a yellow river ice dam sample image, and preprocessing the acquired yellow river ice dam sample image; s2, constructing a mode classifier based on the minimum distance of a decision theory, wherein the mode classifier is used for identifying the image characteristics of the yellow river ice dam, calculating a mode vector sample based on the preprocessed yellow river ice dam sample image in the step S1, and training the constructed mode classifier through the mode vector sample; and step S3, performing mode primary identification on the yellow river ice dam image to be detected based on the trained mode classifier, and performing false identification mode elimination on the mode primary identification result to obtain a yellow river ice dam image characteristic identification result to be detected. The method effectively ensures the robustness, stability and instantaneity of the yellow river ice dam image mode identification, and has strong adaptability.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a yellow river ice dam image feature recognition method.
Background
Ice condition monitoring is a key link of anti-icing decision, water resource optimal scheduling and water engineering operation management. In view of the importance of ice condition monitoring, various countries have carried out ice condition analysis and monitoring work successively, and various monitoring methods are adopted. Currently, the main ice condition observation methods in the world mainly comprise visual observation, instrument observation, remote sensing observation and nuclear technology observation.
The visual inspection is a traditional ice condition monitoring method, and the visual inspection is carried out according to the ice condition observation specification and completely depends on eyes and experience of observers. The visual observation method has wide content, but the observation data are often different due to different observation individuals, so that quantification of the observation result is difficult. Modern technology has higher and higher requirements on data quality, so that visual inspection cannot meet the requirements of ice condition monitoring.
The instrument observation is a basic observation method for reading ice condition data by using simple measuring tools such as theodolites, level gauges, cameras and the like and manually operated by observers. Although the automatic measurement cannot be completely realized, and the visual field range is limited, the obtained data is accurate and objective, and continuous data with higher precision can be obtained, so that the method is widely applied.
The remote sensing observation is to observe ice condition by means of monitoring equipment carried on an airplane and a satellite. The ice condition is observed in a large range by means of optical photography, infrared scanning and the like, and the ice condition data is inverted by processing sound, light and electric signals such as images, reflection and the like to obtain data such as water temperature, ice shape, ice range and the like.
Nuclear technology observations apply nuclear radiation properties to measure ice density and ice velocity by varying reactions through the medium and examining the ability of low concentrations of isotopes.
At present, the ice condition monitoring field in China mainly uses a remote sensing observation method as a data acquisition means, but an ice dam image has a complex and changeable target background, threshold information is difficult to determine by using priori knowledge, the ice dams are various in variety and complex in shape, yellow river water has large sand content, a complete ice dam target is difficult to extract by simply relying on edge information or gray information, and over-segmentation or false segmentation is easy to cause. Therefore, a method for identifying the image features of the yellow river ice dam with strong robustness and effectiveness in real time is needed at present so as to realize accurate identification of the ice dam.
Disclosure of Invention
The invention aims to provide a yellow river ice dam image feature recognition method, which solves the problems existing in the prior art and can effectively improve the robustness, stability and instantaneity of yellow river ice dam image pattern recognition.
In order to achieve the above object, the present invention provides the following solutions: the invention provides a yellow river ice dam image feature recognition method, which comprises the following steps:
s1, acquiring a yellow river ice dam sample image, and preprocessing the acquired yellow river ice dam sample image;
s2, constructing a mode classifier based on the minimum distance of a decision theory, wherein the mode classifier is used for identifying the image characteristics of the yellow river ice dam, calculating a mode vector sample based on the preprocessed yellow river ice dam sample image in the step S1, and training the constructed mode classifier through the mode vector sample;
and step S3, performing mode primary identification on the yellow river ice dam image to be detected based on the trained mode classifier, and performing false identification mode elimination on the mode primary identification result to obtain a yellow river ice dam image characteristic identification result to be detected.
Preferably, the preprocessing the obtained yellow river ice dam sample image in step S1 includes: noise elimination, enhancement and gray scale transformation.
Preferably, in the step S2, the minimum distance construction mode classifier based on the decision theory specifically includes the following steps:
for W mode classes of the yellow river ice dam image, completing the construction of the classifier by calculating W decision functions d (Z), wherein Z represents an n-dimensional mode vector in the yellow river ice dam image; if d i (z)＞d j (z)，j∈[1,W]，i∈[1,W]And j is not equal to i, Z belongs to the ith pattern class w i The method comprises the steps of carrying out a first treatment on the surface of the Pattern class w j From the mean vector m j Sum covariance matrix C j Characterization, as shown in formulas 1 and 2:
wherein N is j Representing from pattern class w j The number of training pattern vector samples;
mode recognition of yellow river ice dam images is completed based on mode vector samples and mahalanobis distance D of mode types j (z) is represented by formula 3:
decision function d j (z) is expressed as: d, d j (z)＝1/D j (z)。
Preferably, in step S2, the mode vector sample is a four-dimensional mode vector, and the mode vector calculating method includes the following steps:
let (x, y) represent the coordinates of a point in the yellow river ice dam image, S xy Representing a k x k neighborhood centered at (x, y), then corresponds to the pattern vector z at the (x, y) point xy As shown in formula 4:
z xy ＝(R，G，B，V)S xy 4
wherein R, G, B each represents a neighborhood S xy Mean value of middle red, green and blue color components, V represents neighborhood S xy The mean square error of the gray value in (c).
Preferably, the mode class of the yellow river ice dam image comprises two modes of an ice dam mode and other modes.
Preferably, step S3 performs mode primary identification on the yellow river ice dam image to be detected based on the trained mode classifier, and performs false identification mode elimination on the mode primary identification result, and specifically includes the following steps:
step S31, preprocessing the yellow river ice dam image to be detected according to the step S1, carrying out neighborhood division on the preprocessed yellow river ice dam image to be detected, and calculating a mode vector based on the neighborhood to obtain a mode vector sample to be detected;
s32, inputting a mode vector sample to be detected into a trained mode classifier, and performing mode primary identification on the yellow river ice dam image;
and step S33, performing false recognition mode elimination on the primary recognition result of the yellow river ice dam image based on morphological open operation to obtain the feature recognition result of the yellow river ice dam image to be detected.
The invention discloses the following technical effects:
according to the method, a reasonable mode identification model is built based on the minimum Markov distance of the decision theory, the mode vector sample is obtained through the mode vector sample statistical algorithm, the representativeness, the anti-interference capability, the calculation efficiency and the compatibility of the classifier of the mode vector sample are comprehensively considered, the mode classification problem is simplified, the false identification mode elimination is carried out based on morphological open operation, and therefore robustness, stability and instantaneity of the yellow river ice dam image mode identification are effectively guaranteed, adaptability is high, and the real-time performance of the ice condition remote sensing monitoring and the requirement of local river segment fine monitoring are met.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a yellow river ice dam image feature recognition method of the invention;
FIG. 2 is a flow chart of the training of the mode classifier of the present invention;
FIG. 3 is a training image and a selected region of interest according to an embodiment of the present invention;
FIG. 4 is a flow chart of feature recognition of a yellow river ice dam image based on a trained yellow river ice dam image feature recognition model according to the present invention;
fig. 5 is an ice dam pattern recognition result of two aerial yellow river ice dam images according to the embodiment of the present invention, where fig. 5 (a) is a first aerial image, 5 (b) is an ice dam pattern recognition result of the first aerial image, 5 (c) is an ice dam pattern initial recognition result of the first aerial image, and 5 (d) is an ice dam pattern erroneous recognition elimination result of the first aerial image; fig. 5 (e) is a second aerial image, 5 (f) is an ice dam pattern recognition result of the second aerial image, 5 (g) is an ice dam pattern initial recognition result of the second aerial image, and 5 (h) is an ice dam pattern erroneous recognition elimination result of the second aerial image;
FIG. 6 shows the results of identifying the ice dam pattern of three indoor scene images according to the embodiment of the present invention, wherein FIG. 6 (a) shows the first indoor scene image and FIG. 6 (b) shows the results of identifying the ice dam pattern of the first indoor scene image; FIG. 6 (c) is a second image of an indoor scene, and FIG. 6 (d) is a result of identifying the ice dam pattern of the second indoor scene; fig. 6 (e) is a third indoor scene image, and fig. 6 (f) is an ice dam pattern recognition result of the third indoor scene;
FIG. 7 shows the result of identifying the ice dam pattern of two outdoor scene images according to the embodiment of the present invention, wherein FIG. 7 (a) is the first outdoor scene image, and FIG. 7 (b) is the result of identifying the ice dam pattern of the first outdoor scene image; fig. 7 (c) shows a second outdoor scene image, and fig. 7 (d) shows a result of identifying an ice dam pattern of the second outdoor scene.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Referring to fig. 1, the embodiment provides a yellow river ice dam image feature recognition method, which includes the following steps:
s1, acquiring a yellow river ice dam sample image, and preprocessing the acquired yellow river ice dam sample image;
the yellow river ice dam sample image comprises remote sensing images obtained through a plurality of modes of MODIS, TM, IRSP, CBERS and China remote sensing first-number radar satellites; the preprocessing comprises noise elimination, enhancement and gray scale transformation.
Compared with other images, the yellow river ice dam image has the following characteristics:
(1) The ice dam image has a complex and changeable target background, the sediment content of each river segment of the yellow river is different, the geographic environment is different, and the states of the ice dam and the river are different. In addition, the water surface in the image can also present different gray level distribution along with the transformation of illumination at different time periods, and the threshold information is difficult to determine by using priori knowledge.
(2) The ice dams have various types and complex shapes, for example, the gray value of the annular ice edges is smaller, the gray value of the inner part is larger and is similar to the water surface, and the textures of the surfaces of the ice dams are complex and have a plurality of filament structures, so that the ice dams are very easy to interfere with a segmentation algorithm, and oversplitting or mis-segmentation is caused.
(3) The yellow river water has larger sand content, part of river water in the river reach is turbid, the dam is hidden and becomes weak, and the complete dam target is difficult to extract by simply relying on edge information or gray information.
Therefore, the model and algorithm suitable for the yellow river ice dam image feature recognition are considered in the aspects of robustness, stability, real-time performance and the like, and the effectiveness of the algorithm is improved in the aspects of self-adaption to a threshold value, the integrity of a segmentation target, the sensitivity to the texture of the ice dam target and the like.
S2, constructing and training a yellow river ice dam image feature recognition model; the method specifically comprises the following steps:
s21, constructing a mode classifier based on a minimum distance of a decision theory, wherein the mode classifier is used for identifying the image features of the yellow river ice dam;
in various yellow river ice dam images, the ice dams are usually represented as gray texture areas with certain roughness, and the pattern features are different from other environmental pattern features such as water surface, river bank, sky and the like, and can be depicted by the ice dam colors and textures. In addition, in order to realize the ice dam mode identification, an effective mode classifier is also required to be established, so that the ice dam mode can be distinguished from other environment modes with higher accuracy.
The mode classifier consists of a mode vector Z and a decision function d (Z);
let z= (z) 1 ，z 2 ，...，z n ) T Representing an n-dimensional mode vector in the yellow river ice dam image, and W mode classes W for the yellow river ice dam image 1 ，w 2 ，...，w W The basic problem of decision theory pattern recognition is to find W decision functions, d 1 (z)，d 2 (z)，...，d W (z) these decision functions have the following properties: if d i (z)＞d j (z)，j∈[1,W]，i∈[1,W]And j is not equal to i, Z belongs to w i Class.
Assume that each pattern class w j From the mean vector m j Sum covariance matrix C j Characterization, as shown in formula (1) and formula (2):
wherein N is j Representing from pattern class w j The number of training pattern vector samples.
An effective way to determine which class the unknown pattern vector Z belongs to is to divide it into the nearest pattern classes, this embodiment employing the mahalanobis distance D j (z) performing measurement as shown in formula (3):
then, decision function d j (z) is defined as d j (z)＝1/D j (z)。
Since patterns in yellow river ice dam images are generally not likely to be characterized by some accurate measurement, the use of mahalanobis distance can better increase the compatibility of the pattern classifier.
In various yellow river ice dam images, the main characteristics of the two modes are in the color and texture properties of the images, so that the mode vector is set as a four-dimensional mode vector, and (x, y) represents the coordinate of a certain point in the yellow river ice dam image, S xy Representing a k x k neighborhood centered at (x, y), then corresponds to the pattern vector z at the (x, y) point xy As shown in formula (4):
z xy ＝(R，G，B，V)S xy (4)
wherein R, G, B each represents a neighborhood S xy Average values of the middle red, green and blue color components are shown in formulas (5) to (7)) The following is shown:
r s，t 、g s，t 、b s，t respectively represent the neighborhoods S xy Red, green, blue color component values at the middle coordinates (s, t); v represents the neighborhood S xy In the gray value mean square error, neighborhood S xy Gray value I at coordinates (s, t) s，t As shown in formula (8):
I s，t ＝0.2989×r s，t +0.5870×g s，t +0.1140×b s，t (8)
then neighborhood S xy The gray value average value I of the color filter is shown as a formula (9):
neighborhood S xy The gray value mean square error V of (b) is shown in formula (10):
because each component of the mode vector comes from statistics of local areas of the yellow river ice dam image, the influence of interference and noise can be well resisted.
In various yellow river ice dam images, a large number of different modes exist, and in order to simplify the problem of yellow river ice dam image mode identification, the modes in the images are divided into two types, namely an ice dam mode and other modes.
Step S22, training a mode classifier;
the specific training process is shown in fig. 2, and includes:
first, training samples for training a pattern classifier, i.e., pattern vector samples representing an ice dam pattern, are acquired. The learning sample is obtained by interactively selecting the region of interest, i.e. the representative ice dam region, on the yellow river ice dam sample image preprocessed in step S1 according to experience, and the black box is the selected training region of interest as shown in fig. 3.
Next, using image data statistics in the region of interest, mode vector samples of the ice bank mode are calculated as shown in equations (4) to (10). In this embodiment, k=9, i.e. 9×9 neighborhood S is set xy Completion pattern vector sample z xy Is a statistic of (1). The neighborhood is set by considering the representativeness of the mode vector sample, the anti-interference capability of the recognition algorithm and the timeliness of the recognition algorithm.
And training the mode classifier through the calculated mode vector samples, and calculating a mean vector m and a covariance matrix C through training to obtain the trained mode classifier.
And S3, carrying out feature recognition on the yellow river ice dam image based on the trained yellow river ice dam image feature recognition model.
The specific identification method is shown in fig. 4, and specifically comprises the following steps:
step S31, preprocessing the yellow river ice dam image to be detected according to the step S1, carrying out neighborhood division on the preprocessed yellow river ice dam image to be detected, and calculating a mode vector based on the neighborhood to obtain a mode vector sample to be detected;
s32, inputting a mode vector sample to be detected into a trained mode classifier, and performing mode primary identification on the yellow river ice dam image;
in the mode classification process, as the ice dam mode identification problem is simplified into the 0-1 judgment problem between two types of modes, an adaptive threshold T is introduced for improving the identification speed of the mode classifier,
the main diagonal of covariance matrix C contains the variances of R, G, B and V components in the pattern vector samples, so the adaptive threshold T is the smallest integer value that is greater than the mean square of each component.
S33, performing false recognition mode elimination on a primary mode recognition result of the yellow river ice dam image;
based on the complexity of the yellow river ice dam image, it is impossible to completely avoid false identifications, such as at the junction of the target area and the non-target area, and there may be some sporadic patterns in the non-target area with characteristics highly similar to the target pattern. In the initial recognition result of the classifier, partial misrecognition points exist around the target region, and some scattered pseudo target regions also exist. Based on continuity of ice dam features, the embodiment eliminates scattered samples in a primary mode identification result of a yellow river ice dam image based on morphological opening operation, and specifically comprises the following steps:
first, let a denote the coordinate set of the initial recognition result matrix (binary matrix) outputted by the pattern classifier, and define an isotropic morphological structure element B (size 3×3), then a is calculated by B morphological opening as shown in equation (11):
it (B) q Representation point q= (q 1 ，q 2 ) Translation of set B, as shown in equation (12):
(B) q ＝{c|c＝b+q，b∈B} (12)
the result of morphological open operation is to eliminate the area with the dimension smaller than that of morphological structural element B in the initial recognition result matrix and also to eliminate the tiny protruding part around the target area, so as to eliminate most of false recognition mode vector samples and complete the feature recognition of the yellow river ice dam image.
In order to further verify the effectiveness and robustness of the method, the embodiment identifies the ice dam modes of two aerial yellow river ice dam images, the identification results are shown in fig. 5, wherein fig. 5 (a) is a first aerial image, 5 (b) is an ice dam mode identification result of the first aerial image, 5 (c) is an ice dam mode initial identification result of the first aerial image, and 5 (d) is an ice dam mode misidentification elimination result of the first aerial image; fig. 5 (e) is a second aerial image, 5 (f) is an ice dam pattern recognition result of the second aerial image, 5 (g) is an ice dam pattern initial recognition result of the second aerial image, and 5 (h) is an ice dam pattern erroneous recognition elimination result of the second aerial image. From fig. 5, it can be seen that the inventive identification method has excellent robustness and stability.
The yellow river ice dam image feature recognition method is tested on a PC (CPU (Central processing Unit) configuration: interi-53450 3.1G,RAM:16G) machine, and for 3888×2592 high-resolution color test images, only 0.5 second is needed from inputting the test images to outputting recognition results, so that the real-time requirement can be completely met.
Three indoor scene images are selected, the method is verified, and the ice dam mode recognition result is shown in fig. 6; fig. 6 (a) is a first indoor scene image, and fig. 6 (b) is an ice dam pattern recognition result of the first indoor scene; FIG. 6 (c) is a second image of an indoor scene, and FIG. 6 (d) is a result of identifying the ice dam pattern of the second indoor scene; fig. 6 (e) shows a third indoor scene image, and fig. 6 (f) shows a result of identifying an ice bank pattern of the third indoor scene. According to fig. 6, it can be seen that a better result is obtained for identifying the ice dam pattern in the image of the indoor experimental scene. Meanwhile, because of the scenes with color and texture characteristics very close to those of the ice dam mode, such as off-white railing and bright window reflection on water surface, some false identifications are caused in the images.
Two outdoor scene images are selected, the method is verified, and the ice dam mode recognition result is shown in fig. 7; fig. 7 (a) is a first outdoor scene image, and fig. 7 (b) is an ice dam pattern recognition result of the first outdoor scene; fig. 7 (c) shows a second outdoor scene image, and fig. 7 (d) shows a result of identifying an ice dam pattern of the second outdoor scene. According to fig. 7, it can be seen that the method of the present invention has very high adaptability, and also has good results for identifying the ice dam pattern in the outdoor experimental scene image.
According to the verification result, the yellow river ice dam image feature recognition method disclosed by the invention is used for constructing a reasonable pattern recognition model based on the minimum mahalanobis distance of the decision theory, acquiring a pattern vector sample through a pattern vector sample statistical algorithm, comprehensively considering the representativeness, the anti-interference capability, the calculation efficiency and the compatibility of a classifier of the pattern vector sample, simplifying the pattern classification problem, and carrying out false recognition pattern elimination based on morphological opening operation, so that the robustness, the stability and the instantaneity of yellow river ice dam image pattern recognition are effectively ensured, and the adaptability is strong.
In the description of the present invention, it should be understood that the terms "longitudinal," "transverse," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the present invention.
The above embodiments are only illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope defined by the claims of the present invention without departing from the design spirit of the present invention.
Claims (4)
1. The yellow river ice dam image feature recognition method is characterized by comprising the following steps of:
s1, acquiring a yellow river ice dam sample image, and preprocessing the acquired yellow river ice dam sample image;
s2, constructing a mode classifier based on the minimum distance of a decision theory, wherein the mode classifier is used for identifying the image characteristics of the yellow river ice dam, calculating a mode vector sample based on the preprocessed yellow river ice dam sample image in the step S1, and training the constructed mode classifier through the mode vector sample;
in the step S2, the minimum distance construction mode classifier based on the decision theory specifically includes the following steps:
for W mode classes of the yellow river ice dam image, completing the construction of the classifier by calculating W decision functions d (Z), wherein Z represents an n-dimensional mode vector in the yellow river ice dam image; if d i (z)＞d j (z)，j∈[1,W]，i∈[1,W]And j is not equal to i, Z belongs to the ith pattern class w i The method comprises the steps of carrying out a first treatment on the surface of the Pattern class w i From the mean vector m j Sum covariance matrix C j Characterization, as shown in formulas 1 and 2:
wherein N is j Representing from pattern class w j The number of training pattern vector samples;
mode recognition of yellow river ice dam images is completed based on mode vector samples and mahalanobis distance D of mode types j (z) is represented by formula 3:
D j (z)＝[(z-m j ) T C j -1 (2-m j )] 1/2 3
decision function dd j (z) is expressed as: d, d j (z)＝1/D j (z)；
In step S2, the mode vector sample is a four-dimensional mode vector, and the mode vector calculation method includes the following steps:
let (x, y) represent the coordinates of a point in the yellow river ice dam image, S xy Representing a k x k neighborhood centered at (x, y), then corresponds to the pattern vector z at the (x, y) point xy As shown in formula 4:
wherein R, G, B each represents a neighborhood S xy Mean value of middle red, green and blue color components, V represents neighborhood S xy The mean square error of gray values in (a);
and step S3, performing mode primary identification on the yellow river ice dam image to be detected based on the trained mode classifier, and performing false identification mode elimination on the mode primary identification result to obtain a yellow river ice dam image characteristic identification result to be detected.
2. The yellow river ice dam image feature recognition method according to claim 1, wherein the preprocessing of the obtained yellow river ice dam sample image in step S1 comprises: noise elimination, enhancement and gray scale transformation.
3. The yellow river ice dam image feature recognition method of claim 1, wherein the pattern classes of the yellow river ice dam image comprise both an ice dam pattern and other patterns.
4. The yellow river ice dam image feature recognition method according to claim 1, wherein the step S3 is based on a trained pattern classifier to perform pattern primary recognition on the yellow river ice dam image to be detected, and performs false recognition pattern elimination on a pattern primary recognition result, and specifically comprises the following steps:
step S31, preprocessing the yellow river ice dam image to be detected according to the step S1, carrying out neighborhood division on the preprocessed yellow river ice dam image to be detected, and calculating a mode vector based on the neighborhood to obtain a mode vector sample to be detected;
s32, inputting a mode vector sample to be detected into a trained mode classifier, and performing mode primary identification on the yellow river ice dam image;
and step S33, performing false recognition mode elimination on the primary recognition result of the yellow river ice dam image based on morphological open operation to obtain the feature recognition result of the yellow river ice dam image to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010422701.XA CN111611910B (en) | 2020-05-19 | 2020-05-19 | Yellow river ice dam image feature recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010422701.XA CN111611910B (en) | 2020-05-19 | 2020-05-19 | Yellow river ice dam image feature recognition method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111611910A CN111611910A (en) | 2020-09-01 |
| CN111611910B true CN111611910B (en) | 2023-04-28 |
Family
ID=72203229
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010422701.XA Active CN111611910B (en) | 2020-05-19 | 2020-05-19 | Yellow river ice dam image feature recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111611910B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB9307611D0 (en) * | 1993-04-13 | 1993-06-02 | Univ Strathclyde | Object recognition |
| JPH09134432A (en) * | 1995-07-17 | 1997-05-20 | Nec Corp | Pattern recognition method |
| WO2019232866A1 (en) * | 2018-06-08 | 2019-12-12 | 平安科技（深圳）有限公司 | Human eye model training method, human eye recognition method, apparatus, device and medium |
| CN110956101A (en) * | 2019-11-19 | 2020-04-03 | 广东省城乡规划设计研究院 | Remote sensing image yellow river ice detection method based on random forest algorithm |
-
2020
- 2020-05-19 CN CN202010422701.XA patent/CN111611910B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB9307611D0 (en) * | 1993-04-13 | 1993-06-02 | Univ Strathclyde | Object recognition |
| JPH09134432A (en) * | 1995-07-17 | 1997-05-20 | Nec Corp | Pattern recognition method |
| WO2019232866A1 (en) * | 2018-06-08 | 2019-12-12 | 平安科技（深圳）有限公司 | Human eye model training method, human eye recognition method, apparatus, device and medium |
| CN110956101A (en) * | 2019-11-19 | 2020-04-03 | 广东省城乡规划设计研究院 | Remote sensing image yellow river ice detection method based on random forest algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 开河期冰坝预测方法研究进展;茅泽育等;《水利水电科技进展》;20070620(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111611910A (en) | 2020-09-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105260709B (en) | Water meter calibration method, apparatus based on image procossing and system | |
| CN103442209B (en) | Video monitoring method of electric transmission line | |
| CN105260710B (en) | Water meter calibration method, apparatus based on image procossing and system | |
| CN102609701B (en) | Remote sensing detection method based on optimal scale for high-resolution SAR (synthetic aperture radar) | |
| CN110414411A (en) | The sea ship candidate region detection method of view-based access control model conspicuousness | |
| CN112434745B (en) | Occlusion target detection and identification method based on multi-source cognitive fusion | |
| CN112149543B (en) | Building dust recognition system and method based on computer vision | |
| CN107767400A (en) | Remote sensing images sequence moving target detection method based on stratification significance analysis | |
| CN111476159A (en) | Method and device for training and detecting detection model based on double-angle regression | |
| CN111242864A (en) | Finger vein image restoration method based on Gabor texture constraint | |
| CN110598613B (en) | Expressway agglomerate fog monitoring method | |
| CN109975196B (en) | Reticulocyte detection method and system | |
| CN108229551A (en) | A kind of Classification of hyperspectral remote sensing image method based on compact dictionary rarefaction representation | |
| CN102855485A (en) | Automatic wheat earing detection method | |
| CN114549446A (en) | Cylinder sleeve defect mark detection method based on deep learning | |
| Huang et al. | A new haze removal algorithm for single urban remote sensing image | |
| CN105512622A (en) | Visible remote-sensing image sea-land segmentation method based on image segmentation and supervised learning | |
| CN106526558B (en) | Gust front automatic identifying method based on Doppler weather radar data | |
| CN111611910B (en) | Yellow river ice dam image feature recognition method | |
| Li et al. | Surface defect detection of seals based on K-means clustering algorithm and particle swarm optimization | |
| CN114862883A (en) | Target edge extraction method, image segmentation method and system | |
| CN108765384A (en) | A kind of conspicuousness detection method of joint manifold ranking and improvement convex closure | |
| CN107239754B (en) | Automobile logo identification method based on sparse sampling intensity profile and gradient distribution | |
| CN112241691A (en) | Channel ice condition intelligent identification method based on unmanned aerial vehicle inspection and image characteristics | |
| CN113592032B (en) | Infrared imaging false alarm source classification method based on physical model constraint |
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 |