CN111368865B - Remote sensing image oil storage tank detection method and device, readable storage medium and equipment - Google Patents
Remote sensing image oil storage tank detection method and device, readable storage medium and equipment Download PDFInfo
- Publication number
- CN111368865B CN111368865B CN201811600681.XA CN201811600681A CN111368865B CN 111368865 B CN111368865 B CN 111368865B CN 201811600681 A CN201811600681 A CN 201811600681A CN 111368865 B CN111368865 B CN 111368865B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- image
- node
- classifier
- storage tank
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a remote sensing image oil storage tank detection method, a remote sensing image oil storage tank detection device, a computer readable storage medium and a computer readable storage medium device, and belongs to the field of image processing and pattern recognition. The method comprises the following steps: unifying the remote sensing images into a single-channel gray scale image; sliding the detection template on the single-channel gray level graph in a certain step length to obtain a plurality of frames to be classified; extracting IPR characteristics of the frames to be classified, wherein the definition of the IPR characteristics is shown as the following formula 1/(1+y/x), and x and y are gray values of any two pixels in one frame to be classified; classifying the IPR features of the frames to be classified by using a classifier to obtain target candidate frames; and performing nms operation on the target candidate frame to obtain a final detection result. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
Description
Technical Field
The invention relates to the field of image processing and pattern recognition, in particular to a remote sensing image oil storage tank detection method, a remote sensing image oil storage tank detection device, a computer readable storage medium and a computer readable storage device.
Background
The remote sensing image is an earth surface real object image shot by a satellite from space, is a space information source which is important at present, has wide development prospect, and particularly has wide application in various fields such as national defense safety, military deployment, economic production, environmental disaster monitoring and the like. With the development of high and new technologies, the space remote sensing technology has made great progress, the speed of remote sensing information transmission is increased, the space resolution is improved, and the image information is amplified, so that the requirements on the extraction, analysis and processing of the remote sensing image information are promoted. The remote sensing image target detection and identification is an important branch in the remote sensing image processing technology, and has functions in the fields of national economy, national defense, military and the like. Due to the cross penetration of various technologies and knowledge, the remote sensing image target detection and identification technology is developed towards more reliability, high efficiency and generalization.
The oil storage tank is a typical military target in the remote sensing image, and the information of the quantity, the size, the position distribution and the like of the oil storage tank has great significance in military. The remote sensing image information is redundant, and how to efficiently detect the oil storage tank target has larger influence on the subsequent oil storage tank information extraction.
Aiming at the characteristics of the oil storage tank on the remote sensing image, the detection method can be divided into the following three main types:
(1) Morphological method
The method mainly starts from the angle of the circular or nearly circular geometric characteristics presented by the oil storage tank, and achieves the final oil storage tank detection purpose through circle detection. There are commonly known oil storage tank detection methods based on Hough transformation or based on improved Hough transformation, and the like. The Hough transformation is an effective circle detection method, and can still obtain ideal effects under the conditions of noise, deformation and incomplete target area, and has the disadvantages of large calculated amount and long time consumption.
(2) Gray threshold segmentation method
When the gray level distribution in the oil storage tank target is uniform, the gray level difference between the oil storage tank target and the background gray level difference is obvious, and the image noise is small, the oil storage tank and the surrounding background can be separated by using a threshold segmentation technology, and a remote sensing image oil tank target segmentation method based on the Ostu or improving the Ostu threshold segmentation is common.
(3) Template matching method
The method is suitable for the conditions of small deformation, uniform internal gray level change and the like of the oil storage tank, and the matching identification of the oil storage tank target is carried out in the image through the designed template. However, when the target form to be matched has large change and poor image quality, the matching process is easy to influence, and how to design a template with strong general performance and stable effect is one of the difficulties of using a template matching method.
In terms of detection precision analysis, the morphological method and the template matching method are easily influenced by deformation of the oil storage tank from the aspect of the target morphological characteristics, and particularly, for the template matching method, the designed template is difficult to adapt to the target, so that the recognition rate is low; when the image or target gray level change is large, the internal gray level is uneven and the texture is complex, the method based on gray threshold segmentation is directly affected, and in addition, when the gray level distribution difference between images is large, the self-adaption capability of the traditional method for the images is poor, and the recognition capability is reduced. In terms of detection time performance, although the detection method based on gray threshold segmentation is not stable enough, the time utility is better, and the method based on morphology and template matching has large calculated amount, large memory occupation amount and poor instantaneity. That is, the conventional methods have the advantages of not easily achieving a detection effect with better time and precision balance due to the large limitation of the self-applicable conditions.
Disclosure of Invention
In order to solve the technical problems, the invention provides a remote sensing image oil storage tank detection method, a remote sensing image oil storage tank detection device, a readable storage medium and a remote sensing image oil storage tank detection device.
The technical scheme provided by the invention is as follows:
in a first aspect, the present invention provides a method for detecting an oil storage tank in a remote sensing image, the method comprising:
unifying the remote sensing images into a single-channel gray scale image;
sliding the detection template on the single-channel gray level graph in a certain step length to obtain a plurality of frames to be classified;
extracting IPR characteristics of the frames to be classified, wherein the IPR characteristics are defined as follows:
wherein x and y are gray values of any two pixels in a frame to be classified;
classifying the IPR features of the frames to be classified by using a classifier to obtain target candidate frames;
and performing non-maximum value inhibition operation on the target candidate frame to obtain a final detection result.
Further, the classifier is a RST classifier, the RST classifier is a tree classifier, and node characteristics are determined according to generated random numbers when each node is constructed by the RST classifier; and if the generated random number is larger than a preset bias threshold value, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
Further, when the RST classifier learns, each attribute of the node learns a corresponding optimal threshold interval, the optimal threshold interval minimizes error loss of the left leaf node and the right leaf node, when the sample characteristic value is in the optimal threshold interval, the sample is divided into the right leaf node, otherwise, the sample is divided into the left leaf node, and a fitting value is given by the leaf node; the calculation formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
d is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
Further, in the node splitting process of the RST classifier, if the sample sets in the nodes are of the same class, or the total number of samples does not meet the minimum sample number required by the specified splitting, or the corresponding tree depth reaches the specified maximum tree depth, the node splitting is stopped.
Further, unifying the remote sensing images into a single-channel gray scale image includes:
if the remote sensing image is a full-color image and is not processed, if the remote sensing image is a multispectral color image, the remote sensing image is converted into a single-channel gray scale image by the following method:
converting the multispectral color image into a gray level image by adopting a component weighted average mode;
Calculating a global threshold T of the gray image by using an Ostu method * ；
And performing power law transformation on the gray level image according to the global threshold value to obtain the single-channel gray level image, wherein the power law transformation function is as follows:
wherein I is the pixel gray value of the gray image, I * The pixel gray values of the single-channel gray map are represented by a, b which are power law transformation coefficients, a < 1, b > 1, and a, beta which are set function segmentation thresholds.
In a second aspect, the present invention provides a remote sensing image oil storage tank detection device, the device comprising:
the single-channel gray level image acquisition module is used for unifying the remote sensing images into a single-channel gray level image;
the frame to be classified acquisition module is used for sliding the detection template on the single-channel gray level diagram in a certain step length to obtain a plurality of frames to be classified;
the IPR feature extraction module is used for extracting the IPR features of the frames to be classified, and the definition of the IPR features is as follows:
wherein x and y are gray values of any two pixels in a frame to be classified;
the classification module is used for classifying the IPR characteristics of the frames to be classified by using a classifier to obtain target candidate frames;
and the non-maximum value suppression module is used for performing non-maximum value suppression operation on the target candidate frame to obtain a final detection result.
Further, the classifier is a RST classifier, the RST classifier is a tree classifier, and node characteristics are determined according to generated random numbers when each node is constructed by the RST classifier; and if the generated random number is larger than a preset bias threshold value, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
Further, when the RST classifier learns, each attribute of the node learns a corresponding optimal threshold interval, the optimal threshold interval minimizes error loss of the left leaf node and the right leaf node, when the sample characteristic value is in the optimal threshold interval, the sample is divided into the right leaf node, otherwise, the sample is divided into the left leaf node, and a fitting value is given by the leaf node; the calculation formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
d is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
Further, in the node splitting process of the RST classifier, if the sample sets in the nodes are of the same class, or the total number of samples does not meet the minimum sample number required by the specified splitting, or the corresponding tree depth reaches the specified maximum tree depth, the node splitting is stopped.
Further, the single-channel gray scale image acquisition module includes:
if the remote sensing image is a full-color image and is not processed, if the remote sensing image is a multispectral color image, the remote sensing image is converted into a single-channel gray scale image through the following units:
the weighted average unit is used for converting the multispectral color image into a gray image by adopting a component weighted average mode;
an Ostu unit for calculating a global threshold T of the gray image by an Ostu method * ；
And the power law transformation unit is used for performing power law transformation on the gray level image according to the global threshold value to obtain the single-channel gray level image, and the power law transformation function is as follows:
wherein I is the pixel gray value of the gray image, I * The pixel gray values of the single-channel gray map are represented by a, b which are power law transformation coefficients, a < 1, b > 1, and a, beta which are set function segmentation thresholds.
In a third aspect, the present invention provides a computer readable storage medium for remote sensing image oil storage tank detection, including a processor and a memory for storing instructions executable by the processor, the instructions when executed by the processor implementing the steps comprising the remote sensing image oil storage tank detection method of the first aspect.
In a fourth aspect, the present invention provides a device for detecting a remote sensing image oil storage tank, including at least one processor and a memory storing computer executable instructions, where the processor implements the steps of the remote sensing image oil storage tank detection method according to the first aspect when executing the instructions.
The invention has the following beneficial effects:
according to the invention, remote sensing images are unified into a single-channel gray level image, then a frame to be classified is obtained, IPR characteristics of the frame to be classified are extracted, the IPR characteristics are classified by using a classifier, and finally non-maximum suppression is carried out, so that a final detection frame is obtained. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
Drawings
FIG. 1 is a flow chart of a method for detecting a remote sensing image oil storage tank;
FIG. 2 is a schematic diagram of an oil storage tank image;
FIG. 3 is a schematic view of an IPR feature matrix of an oil storage tank;
FIG. 4 is a schematic illustration of pretreatment;
FIG. 5 is a schematic diagram of a capped and uncapped storage tank inspection;
FIG. 6 is a schematic diagram of a tank detection for size and gray scale variation;
FIG. 7 is a schematic diagram of cylinder and sphere tank detection;
fig. 8 is a schematic diagram of a remote sensing image oil storage tank detection device according to the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention.
Example 1:
the embodiment of the invention provides a remote sensing image oil storage tank detection method, as shown in fig. 1, which comprises the following steps:
step S100: and unifying the remote sensing images into a single-channel gray scale image.
The invention processes gray level images, so that the remote sensing images are required to be unified into a single-channel gray level image.
Step S200: and sliding the detection template on the single-channel gray level graph in a certain step length to obtain a plurality of frames to be classified.
The size of the detection template can be set according to the detection requirement, and when the detection template slides, the step size is determined according to the size of the detection template and the size of the remote sensing image.
Step S300: extracting IPR characteristics of a plurality of frames to be classified, wherein the definition of the IPR characteristics is as follows:
wherein x and y are gray values of any two pixels in a frame to be classified.
As shown in fig. 2, the shape and internal gray scale distribution characteristics of the oil storage tank are obvious, and the main gray scale is generally greatly different from the background area, so that the two can be distinguished by comparing the pixel gray scale between the object of the oil storage tank area and the background.
Based on this, the invention designs an improved pixel ratio (Improved Pixel Ratio, IPR) as a feature of an image, which is a variation of the gray ratio of any two pixels, and also describes the order relationship between the two pixels, and the mathematical formula is defined as follows:
And x, y is more than or equal to 0 and represents the gray value of any two pixels, f (x, y) is the improved pixel ratio of x and y, and the IPR characteristic of the frame to be classified can be obtained by traversing each pixel in the frame to be classified by x, y is more than or equal to 0.
From the defined formula, the IPR feature has a finite (f e 0, 1), gray scale invariance, and f (x, y) =1-f (y, x), i.e. one way of calculating the IPR feature value between two pixels can be linearly expressed in another way, so the IPR feature dimension can be d=p× (p-1)/2, where p represents the total number of pixels of the image. Taking fig. 2 as an example, an IPR feature matrix visual image of the oil storage tank is shown in fig. 3.
The invention adopts original IPR characteristic as image characteristic, which is a deformation of pixel gray ratio, examines the sequence relation between any two pixels in the image, and experiments prove that the characteristic is very effective for detecting round objects or nearly round objects, such as oil storage tanks. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, and has higher detection precision.
Step S400: and classifying the IPR characteristics of the frames to be classified by using a classifier to obtain target candidate frames.
Step S500: and performing non-maximum value inhibition operation on the target candidate frame to obtain a final detection result.
Non-maximum suppression (Non-Maximum Suppression, nms), as the name implies, suppressing elements that are not maxima, can be understood as local maximum searches. After the IPR features of a plurality of frames to be classified are identified by using a classifier, each mania to be classified can obtain a score, however, because the step length of the detection template in sliding is smaller than the length of the sliding template, the situation that the target candidate frame and other target candidate frames are crossed is caused to be included or mostly included, and the target candidate frames which are crossed are processed by using non-maximum suppression, so that a final detection frame is obtained.
According to the invention, remote sensing images are unified into a single-channel gray level image, then a frame to be classified is obtained, IPR characteristics of the frame to be classified are extracted, the IPR characteristics are classified by using a classifier, and finally non-maximum suppression is carried out, so that a final detection frame is obtained. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
The remote sensing image comprises a multispectral color image and a panchromatic image, wherein the panchromatic image is a black-and-white image in the whole visible light wave band range (a blue light wave band is often abandoned for avoiding the influence of atmospheric scattering), the panchromatic image is a single-channel gray scale image, the multispectral color image is not processed, and the multispectral color image has three channels of RGB, and is converted into the single-channel gray scale image by the following method:
Step S110: the original R, G, B three-channel multispectral color image is converted into a gray image by adopting a component weighted average mode, and the mathematical formula is as follows:
Grayimage＝0.2989R+0.5870G+0.1140B
compared with full-color images, the gray level distribution of the gray level image converted by the step still has larger difference. Aiming at the problem, the invention provides a processing method based on power law transformation by taking a full-color image as a template, so that the gray distribution of the processed image is approximate to the gray distribution of the full-color image, and the experimental image is standardized, and the specific operation steps are as follows:
step S120: calculating global threshold T of gray image by using Ostu method * 。
The Otsu algorithm is an efficient algorithm for binarizing an image, and a global threshold is used to divide a gray image into a foreground and a background, where the background should be the most different from the foreground when the optimal global threshold is taken.
Step S130: and performing power law transformation on the gray image according to the global threshold value to obtain a single-channel gray image, wherein the power law transformation function is as follows:
wherein I is the pixel gray value of the gray image, I * For the pixel gray values of the single-channel gray map, a, b are power law transformation coefficients, a < 1, b > 1, α, β are set function segmentation thresholds, as a preferred example, a=0.63, b=1.37, α=0.3, β=0.4.
Fig. 4 illustrates two exemplary preprocessing processes, in which (a) and (d) are full-color images, (b) and (e) are ordinary gray-scale images converted from multispectral color images, and (c) and (f) are single-channel gray-scale images after gray-scale power law conversion. Compared with a common gray level image, the gray level distribution of the single-channel gray level image obtained by power law conversion in the embodiment of the invention is more similar to that of a full-color image.
The classifier of the invention is preferably a RST classifier (Random Suboptimal Tree ), which is a tree classifier, wherein the tree RST classifier determines node characteristics according to the generated random numbers when constructing each node; if the generated random number is larger than a preset bias threshold value rho, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
The optimal feature information (information of the current iteration) and the suboptimal feature information (information of the previous iteration) of the embodiment of the present invention may be features used when the node error loss is minimum, threshold interval parameters, left and right leaf node fitting scores, and the like (see later RST classifier learning algorithm examples for details).
The random sub-optimal tree, as the name implies, is the classification tree of the random selection sub-optimal characteristic information, and is the unique classifier of the invention. The invention selects node characteristics according to random disturbance factors (namely random numbers). When each node of RST is constructed, a random disturbance factor is introduced to interfere with the selection of the optimal characteristics of each node (random numbers are randomly given by a system and are subject to uniform distribution during learning), when the random number is larger than a given threshold value rho, the optimal characteristics are selected at the node, otherwise, the suboptimal characteristics are selected (in experiments, rho=0.3 is preferable).
The optimal characteristic information is the information of the iteration, is also the characteristic with the minimum error loss, the suboptimal characteristic information is the information of the previous iteration, is also the characteristic with the suboptimal error loss, and the random disturbance factor is introduced to improve the generalization capability of the model and prevent the occurrence of the overfitting phenomenon. The RST classifier combines with the IPR feature, so that the stability and generalization of the oil storage tank detection method disclosed by the invention are superior to those of the traditional oil storage tank detection method.
When the RST classifier is used for learning, each attribute of the node learns a corresponding optimal threshold interval, and the optimal threshold interval enables error loss of the left leaf node and the right leaf node to be minimum.
A tree classifier is a classification method, given a set of samples, each sample of the set of samples has a set of attributes and a class, which are predetermined, these attributes being classification criteria for the sample, the samples being separable into different classes according to different attributes, these all attributes constituting a set of candidate attributes. The RST classifier is obtained through sample set learning, and during learning, a node learns a corresponding optimal threshold interval on each attribute, the attribute of the node is the attribute in the candidate attribute set, and the node serves as a splitting (classifying) standard according to the attribute, namely the node divides the sample set into left and right leaf nodes according to the attribute.
Because the IPR characteristic function is a continuous function, the characteristic has the invariance of gray scale proportion, and the threshold interval shape classification method is softer and finer than a single threshold shape, the corresponding range of the characteristic is more accurate, and the method belongs to soft division. When the characteristic value of the sample is in the optimal threshold value interval, the sample is divided into a right leaf node, or else, the sample is divided into a left leaf node, and a fitting value is given out by the leaf node; the fitting value of the node is determined according to the distribution condition of positive and negative sample weights contained in the current node, the node loss function is calculated based on a weighted least square method, and specifically, the calculating formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
D is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
The RST classifier follows a tree building process from top to bottom, depth first and breadth second, and in the node splitting process, if the sample sets in the nodes are of the same class, or the total number of samples does not meet the minimum sample number required by specified splitting, or the corresponding tree depth reaches the specified maximum tree depth, the node splitting is stopped.
A specific RST classifier learning algorithm example is given here:
the oil storage tank detection classifier is mainly studied by using a Gentle AdaBoost algorithm framework, and meanwhile, a soft cascade structure can be introduced to further strengthen the performances of a plurality of weak RST classifiers obtained by the study, so that a strong classifier is constructed. When the RST classifier is used for learning, when the characteristic value of a sample is within a threshold value interval, the sample is divided into a right leaf node, otherwise, the sample is divided into a left leaf node, but no matter which leaf node is divided, the sample is not directly judged to be a positive type sample or a negative type sample, and a fitting value is given by the leaf node and is used as an experience reference value of the classification of the subsequent strong classifier.
According to the experience reference value output by the RST classifier, a threshold value is learned in each stage of the soft cascade, when the score of a sample in the stage is larger than the threshold value, the sample is judged to be a positive sample, and the judgment in the current stage can refer to the score in the previous stage because the soft cascade stage function is in a cumulative form, so that the model acceptance can be effectively improved, and the possibility of false rejection is reduced.
In order to adapt to targets with various sizes, the invention designs a multi-scale detection template by taking 15 x 15 pixels as a basic template, and the size difference between two adjacent scales is 1.2 times.
The invention is illustrated below in a specific experimental example:
the invention verifies the validity of the experiment on a remote sensing image library GED (Google earth database), the detection results are shown in fig. 5-7, fig. 5 is an example of detecting a covered and uncovered oil storage tank, fig. 6 is an example of detecting an oil storage tank with variable size and gray scale, and fig. 7 is an example of detecting a cylinder and sphere oil storage tank. Experiments prove that the method has good performance on the image detection of the oil storage tank in the unrestricted state, stronger robustness and generalization capability, higher detection precision and good time utility, and can be effectively applied to the target detection of the oil storage tank in the visible light remote sensing image.
Example 2:
the embodiment of the invention provides a remote sensing image oil storage tank detection device, as shown in fig. 8, which comprises:
the single-channel gray scale image acquisition module 10 is used for unifying the remote sensing images into a single-channel gray scale image;
the frame to be classified acquisition module 20 is used for sliding the detection template on the single-channel gray level chart in a certain step length to obtain a plurality of frames to be classified;
The IPR feature extraction module 30 is configured to extract IPR features of a plurality of frames to be classified, where the definition of the IPR features is as follows:
wherein x and y are gray values of any two pixels in a frame to be classified;
the classification module 40 is configured to classify IPR features of a plurality of frames to be classified by using a classifier, so as to obtain target candidate frames;
and the non-maximum value suppression operation module 50 is used for performing non-maximum value suppression operation on the target candidate frame to obtain a final detection result.
According to the invention, remote sensing images are unified into a single-channel gray level image, then a frame to be classified is obtained, IPR characteristics of the frame to be classified are extracted, the IPR characteristics are classified by using a classifier, and finally non-maximum suppression is carried out, so that a final detection frame is obtained. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
The classifier is a RST classifier, the RST classifier is a tree classifier, and node characteristics are determined according to generated random numbers when each node is constructed by the RST classifier; and if the generated random number is larger than a preset bias threshold value, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
When the RST classifier learns, each attribute of the node learns a corresponding optimal threshold interval, the optimal threshold interval minimizes error loss of the left leaf node and the right leaf node, when the characteristic value of the sample is in the optimal threshold interval, the sample is divided into the right leaf node, otherwise, the sample is divided into the left leaf node, and a fitting value is given by the leaf node; the calculation formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
d is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
In the node splitting process of the RST classifier, if the sample sets in the nodes are of the same class, or the total number of samples does not meet the minimum sample number required by the designated splitting, or the corresponding tree depth reaches the designated maximum tree depth, stopping the node splitting.
The single-channel gray scale image acquisition module of the invention comprises:
if the remote sensing image is a full-color image and is not processed, if the remote sensing image is a multispectral color image, the remote sensing image is converted into a single-channel gray scale image through the following units:
the weighted average unit is used for converting the multispectral color image into a gray image by adopting a component weighted average mode;
an Ostu unit for calculating a global threshold T of the gray image by using an Ostu method * ；
The power law transformation unit is used for performing power law transformation on the gray image according to the global threshold value to obtain a single-channel gray image, and the power law transformation function is as follows:
wherein I is the pixel gray value of the gray image, I * The pixel gray values of the single-channel gray map are represented by a, b which are power law transformation coefficients, a < 1, b > 1, alpha, beta which are set function segmentation thresholdsValues.
The device provided by the embodiment of the present invention has the same implementation principle and technical effects as those of the foregoing method embodiment, and for the sake of brevity, reference may be made to the corresponding content in the foregoing method embodiment where the device embodiment is not mentioned. It will be clear to those skilled in the art that, for convenience and brevity, the specific working procedures of the apparatus and units described above may refer to the corresponding procedures in the above method embodiments, and are not described herein again.
Example 3:
the method or apparatus according to the above embodiments provided in the present specification may implement service logic by a computer program and be recorded on a storage medium, where the storage medium may be read and executed by a computer, to implement the effects of the schemes described in the embodiments of the present specification. Therefore, the present invention also provides a computer readable storage medium for remote sensing image oil storage tank detection, comprising a processor and a memory for storing instructions executable by the processor, the instructions when executed by the processor implementing the steps comprising the remote sensing image oil storage tank detection method of embodiment 1.
According to the invention, remote sensing images are unified into a single-channel gray level image, then a frame to be classified is obtained, IPR characteristics of the frame to be classified are extracted, the IPR characteristics are classified by using a classifier, and finally non-maximum suppression is carried out, so that a final detection frame is obtained. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
The storage medium may include physical means for storing information, typically by digitizing the information before storing it in an electronic, magnetic, or optical medium. The storage medium may include: means for storing information using electrical energy such as various memories, e.g., RAM, ROM, etc.; devices for storing information using magnetic energy such as hard disk, floppy disk, magnetic tape, magnetic core memory, bubble memory, and USB flash disk; devices for optically storing information, such as CDs or DVDs. Of course, there are other ways of readable storage medium, such as quantum memory, graphene memory, etc.
The above description of the apparatus according to the method embodiments may also include other implementations. Specific implementation may refer to descriptions of related method embodiments, which are not described herein in detail.
Example 4:
the invention also provides equipment for detecting the remote sensing image oil storage tank, which can be a single computer, can also comprise an actual operating device and the like using one or more of the methods or one or more of the embodiment devices of the specification. The device for detecting the remote sensing image oil storage tank can comprise at least one processor and a memory for storing computer executable instructions, wherein the steps of the remote sensing image oil storage tank detection method in any one or more embodiments are realized when the processor executes the instructions.
According to the invention, remote sensing images are unified into a single-channel gray level image, then a frame to be classified is obtained, IPR characteristics of the frame to be classified are extracted, the IPR characteristics are classified by using a classifier, and finally non-maximum suppression is carried out, so that a final detection frame is obtained. The method can adaptively detect the oil storage tank targets with changeable morphology and gray scale distribution in the remote sensing image, has higher detection precision, and has simple data processing and good time utility.
The description of the above-mentioned apparatus according to the method or apparatus embodiment may further include other embodiments, and specific implementation manner may refer to the description of the related method embodiment, which is not described herein in detail.
It should be noted that, the description of the apparatus or the system according to the embodiments of the related method in this specification may further include other embodiments, and specific implementation manner may refer to the description of the embodiments of the method, which is not described herein in detail. In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the hardware + program class, the storage medium + program embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and reference is made to the partial description of the method embodiment for relevant points.
The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a car-mounted human-computer interaction device, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
For convenience of description, the above devices are described as being functionally divided into various modules, respectively. Of course, when one or more of the present description is implemented, the functions of each module may be implemented in the same piece or pieces of software and/or hardware, or a module that implements the same function may be implemented by a plurality of sub-modules or a combination of sub-units, or the like. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller can be regarded as a hardware component, and means for implementing various functions included therein can also be regarded as a structure within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method or apparatus comprising such elements.
One skilled in the relevant art will recognize that one or more embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
One or more embodiments of the present specification may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments. In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present specification. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
Finally, it should be noted that: the above examples are only specific embodiments of the present invention, and are not intended to limit the scope of the present invention, but it should be understood by those skilled in the art that the present invention is not limited thereto, and that the present invention is described in detail with reference to the foregoing examples: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the corresponding technical solutions. Are intended to be encompassed within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. The method for detecting the oil storage tank of the remote sensing image is characterized by comprising the following steps of:
unifying the remote sensing images into a single-channel gray scale image;
sliding the detection template on the single-channel gray level graph with a preset step length to obtain a plurality of frames to be classified;
Extracting IPR characteristics of the frames to be classified, wherein the IPR characteristics are defined as follows:
the IPR characteristic is improved pixel ratio, is a deformation of gray ratio of any two pixels, and represents sequence relation between the two pixels, and x and y are gray values of any two pixels in a frame to be classified;
classifying the IPR features of the frames to be classified by using a classifier to obtain target candidate frames;
performing non-maximum suppression operation on the target candidate frame to obtain a final detection result;
the classifier is a RST classifier, the RST classifier is a random suboptimal tree classifier, and node characteristics are determined according to generated random numbers when each node is constructed by the RST classifier; and if the generated random number is larger than a preset bias threshold value, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
2. The method for detecting the oil storage tank of the remote sensing image according to claim 1, wherein each attribute of the node learns a corresponding optimal threshold interval when the RST classifier learns, the optimal threshold interval minimizes error loss of the left leaf node and the right leaf node, when the characteristic value of the sample is within the optimal threshold interval, the sample is divided into the right leaf node, otherwise, the sample is divided into the left leaf node, and a fitting value is given by the leaf node; the calculation formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
D is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
3. The method for detecting the oil storage tank of the remote sensing image according to claim 2, wherein in the node splitting process of the RST classifier, if the sample sets in the nodes are of the same class, or the total number of samples does not meet the minimum sample number required by the specified splitting, or the corresponding tree depth reaches the specified maximum tree depth, the node splitting is stopped.
4. A method for detecting an oil storage tank in a remote sensing image according to any one of claims 1 to 3, wherein unifying the remote sensing image into a single-channel gray scale map comprises:
if the remote sensing image is a full-color image and is not processed, if the remote sensing image is a multispectral color image, the remote sensing image is converted into a single-channel gray scale image by the following method:
converting the multispectral color image into a gray level image by adopting a component weighted average mode;
calculating a global threshold T of the gray image by using an Ostu method * ；
And performing power law transformation on the gray level image according to the global threshold value to obtain the single-channel gray level image, wherein the power law transformation function is as follows:
wherein I is a gray scale imagePixel gray value, I * The pixel gray values of the single-channel gray map are represented by a, b which are power law transformation coefficients, a < 1, b > 1, and a, beta which are set function segmentation thresholds.
5. A remote sensing image oil storage tank detection device, the device comprising:
the single-channel gray level image acquisition module is used for unifying the remote sensing images into a single-channel gray level image;
the frame to be classified acquisition module is used for sliding the detection template on the single-channel gray level diagram in a preset step length to obtain a plurality of frames to be classified;
the IPR feature extraction module is used for extracting the IPR features of the frames to be classified, and the definition of the IPR features is as follows:
the IPR characteristic is improved pixel ratio, is a deformation of gray ratio of any two pixels, and represents sequence relation between the two pixels, and x and y are gray values of any two pixels in a frame to be classified;
the classification module is used for classifying the IPR characteristics of the frames to be classified by using a classifier to obtain target candidate frames;
the non-maximum value suppression operation module is used for performing non-maximum value suppression operation on the target candidate frame to obtain a final detection result;
the classifier is a RST classifier, the RST classifier is a random suboptimal tree classifier, and node characteristics are determined according to generated random numbers when each node is constructed by the RST classifier; and if the generated random number is larger than a preset bias threshold value, selecting optimal characteristic information as node characteristics, otherwise, selecting suboptimal characteristic information as node characteristics, wherein the optimal characteristic information is information of the current iteration, and the suboptimal characteristic information is information of the previous iteration.
6. The remote sensing image oil storage tank detection device according to claim 5, wherein the RST classifier learns a corresponding optimal threshold interval for each attribute of the node during learning, the optimal threshold interval minimizes error loss of the left leaf node and the right leaf node, and when the characteristic value of the sample is within the optimal threshold interval, the sample is divided into the right leaf node, otherwise, the sample is divided into the left leaf node, and a fitting value is given by the leaf node; the calculation formulas of the fitting value F (D) and the error loss L are as follows:
L＝W + ·(F(D)-1)·(F(D)-1)+W - ·(F(D)+1)·(F(D)+1)
d is a sample set in the node, W + Is the weight sum of the positive samples in D, W - Is the sum of the weights of the negative samples in D.
7. A computer readable storage medium for remote sensing image oil storage tank detection, comprising a processor and a memory for storing instructions executable by the processor, the instructions when executed by the processor implementing steps comprising the remote sensing image oil storage tank detection method of any one of claims 1-4.
8. An apparatus for remote sensing image storage tank inspection comprising at least one processor and a memory storing computer executable instructions that when executed by the processor perform the steps of the remote sensing image storage tank inspection method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600681.XA CN111368865B (en) | 2018-12-26 | 2018-12-26 | Remote sensing image oil storage tank detection method and device, readable storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600681.XA CN111368865B (en) | 2018-12-26 | 2018-12-26 | Remote sensing image oil storage tank detection method and device, readable storage medium and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111368865A CN111368865A (en) | 2020-07-03 |
| CN111368865B true CN111368865B (en) | 2023-08-22 |
Family
ID=71206014
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811600681.XA Active CN111368865B (en) | 2018-12-26 | 2018-12-26 | Remote sensing image oil storage tank detection method and device, readable storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111368865B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112749673A (en) * | 2021-01-20 | 2021-05-04 | 西安中科星图空间数据技术有限公司 | Method and device for intelligently extracting stock of oil storage tank based on remote sensing image |
| CN114550009B (en) * | 2022-04-01 | 2022-12-06 | 中国科学院空天信息创新研究院 | Lightweight target detection method, device, equipment and medium for satellite-borne remote sensing image |
| CN116681699B (en) * | 2023-07-31 | 2023-12-15 | 国网浙江省电力有限公司宁波供电公司 | Intelligent storage method and device for transformer substation oil samples |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107346434A (en) * | 2017-05-03 | 2017-11-14 | 上海大学 | A kind of plant pest detection method based on multiple features and SVMs |
| CN107358239A (en) * | 2017-06-07 | 2017-11-17 | 南京航空航天大学 | Image invariant feature extraction method based on circular track conversion |
| CN108304873A (en) * | 2018-01-30 | 2018-07-20 | 深圳市国脉畅行科技股份有限公司 | Object detection method based on high-resolution optical satellite remote-sensing image and its system |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
-
2018
- 2018-12-26 CN CN201811600681.XA patent/CN111368865B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107346434A (en) * | 2017-05-03 | 2017-11-14 | 上海大学 | A kind of plant pest detection method based on multiple features and SVMs |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN107358239A (en) * | 2017-06-07 | 2017-11-17 | 南京航空航天大学 | Image invariant feature extraction method based on circular track conversion |
| CN108304873A (en) * | 2018-01-30 | 2018-07-20 | 深圳市国脉畅行科技股份有限公司 | Object detection method based on high-resolution optical satellite remote-sensing image and its system |
Non-Patent Citations (1)
| Title |
|---|
| A new kind of Combination Classifier and its application in Classification of Remote Sensing Image;LI Chao-kui, et al;2018 Fifth International Workshop on Earth Observation and Remote Sensing Applications (EORSA);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111368865A (en) | 2020-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xie et al. | Multilevel cloud detection in remote sensing images based on deep learning | |
| Byeon et al. | Scene labeling with lstm recurrent neural networks | |
| CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
| CN110738207A (en) | character detection method for fusing character area edge information in character image | |
| CN105528575B (en) | Sky detection method based on Context Reasoning | |
| Jiang et al. | Focal-test-based spatial decision tree learning | |
| Cao et al. | A traffic-sign detection algorithm based on improved sparse r-cnn | |
| CN111368865B (en) | Remote sensing image oil storage tank detection method and device, readable storage medium and equipment | |
| Jiang et al. | Hyperspectral image classification with spatial consistence using fully convolutional spatial propagation network | |
| Ming et al. | Monocular depth-ordering reasoning with occlusion edge detection and couple layers inference | |
| JP6107531B2 (en) | Feature extraction program and information processing apparatus | |
| CN110852327A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| US20210272295A1 (en) | Analysing Objects in a Set of Frames | |
| CN112115805A (en) | Pedestrian re-identification method and system with bimodal hard-excavation ternary-center loss | |
| Zhao et al. | CRAS-YOLO: A novel multi-category vessel detection and classification model based on YOLOv5s algorithm | |
| Yang et al. | An improved algorithm for the detection of fastening targets based on machine vision | |
| Li et al. | Fast object detection from unmanned surface vehicles via objectness and saliency | |
| Fu et al. | A case study of utilizing YOLOT based quantitative detection algorithm for marine benthos | |
| CN110633727A (en) | Deep neural network ship target fine-grained identification method based on selective search | |
| CN111340139B (en) | Method and device for judging complexity of image content | |
| CN111353577B (en) | Multi-task-based cascade combination model optimization method and device and terminal equipment | |
| Lu et al. | Clustering based road detection method | |
| CN106340038B (en) | A kind of depth order inference method based on object in monocular image | |
| Sun et al. | Contextual models for automatic building extraction in high resolution remote sensing image using object-based boosting method | |
| Talebi et al. | Nonparametric scene parsing in the images of buildings |
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 |