CN114359731A - Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing - Google Patents
Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing Download PDFInfo
- Publication number
- CN114359731A CN114359731A CN202210010640.5A CN202210010640A CN114359731A CN 114359731 A CN114359731 A CN 114359731A CN 202210010640 A CN202210010640 A CN 202210010640A CN 114359731 A CN114359731 A CN 114359731A
- Authority
- CN
- China
- Prior art keywords
- thermal infrared
- image
- infrared image
- aerial vehicle
- unmanned aerial
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000010865 sewage Substances 0.000 title claims abstract description 20
- 238000001514 detection method Methods 0.000 title claims abstract description 14
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 65
- 230000002159 abnormal effect Effects 0.000 claims abstract description 20
- 238000001228 spectrum Methods 0.000 claims description 22
- 238000003384 imaging method Methods 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 8
- 230000000007 visual effect Effects 0.000 claims description 8
- 210000000746 body region Anatomy 0.000 claims description 7
- 230000007613 environmental effect Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013136 deep learning model Methods 0.000 claims description 4
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 230000029142 excretion Effects 0.000 claims description 2
- 238000003702 image correction Methods 0.000 claims description 2
- 230000005855 radiation Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 230000009466 transformation Effects 0.000 description 6
- 238000013507 mapping Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000009499 grossing Methods 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000007599 discharging Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- RZVHIXYEVGDQDX-UHFFFAOYSA-N 9,10-anthraquinone Chemical compound C1=CC=C2C(=O)C3=CC=CC=C3C(=O)C2=C1 RZVHIXYEVGDQDX-UHFFFAOYSA-N 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000010842 industrial wastewater Substances 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000009418 renovation Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000003643 water by type Substances 0.000 description 1
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a hidden sewage discharge detection and tracing method based on unmanned aerial vehicle thermal infrared remote sensing, which comprises the following steps: acquiring a decimeter-level thermal infrared image along a preset airway based on an unmanned aerial vehicle and an infrared imager; extracting homonymous feature points in the thermal infrared image overlapping area, and correcting and splicing the thermal infrared images; obtaining a thermal infrared image having a geographic parameter; extracting a water body area in the thermal infrared image, segmenting the water body area based on the temperature value, and extracting a water area with abnormal temperature; and analyzing the thermal infrared image with the geographic parameters based on the water area with the abnormal temperature, and searching a temperature abnormal linear target with connectivity on the land as a suspected candid image pipeline, wherein the end of the suspected candid image pipeline is a suspected candid image main body. By utilizing the portable characteristic of the unmanned aerial vehicle, the operation can be flexibly carried out, the operation cost is reduced for the industrial sewage steal-discharge detection work, and the time efficiency is improved; by utilizing the radiation inversion characteristic of thermal infrared remote sensing, the detection capability of pipelines and immersed steal drainage is improved.
Description
Technical Field
The invention belongs to the technical field of environmental monitoring, and particularly relates to a hidden sewage discharge detection and tracing method based on unmanned aerial vehicle thermal infrared remote sensing.
Background
The rapid monitoring of the water quality of natural waters is of great importance to the conservation of water resources and related land resources. Today, the discharge of industrial waste water is an important cause of water quality pollution. Some sewage outlets and pipelines which do not meet the requirements of laws and regulations are usually concealed, and the sewage discharge period is unstable, such as avoiding the discharge in daytime, and the phenomena increase the difficulty of the inspection and the cleaning. In order to meet the requirement of protecting the water environment, the environmental protection department needs to dynamically monitor the stealing and hidden discharging of the industry, and then support data is provided for the renovation work.
The existing law enforcement department steal detection method mainly depends on a pull net type field manual inspection drain outlet, and workers can search by walking or riding a monitoring ship along the shore. The method has the following disadvantages: (1) the manual search has low working efficiency and large workload for large-scale detection and investigation; (2) for the condition that the sewage draining exit is hidden under water or in weeds, the condition is difficult to find in time and easy to miss; (3) the working environment on the coastal beach has certain personal danger for the working personnel; (4) partially adopting a satellite remote sensing means, the requirements of on-site law enforcement cannot be met due to the limitation of time and spatial resolution; (5) a few units adopt an unmanned aerial vehicle thermal infrared method to pay more attention to monitoring of a water area sewage discharge outlet, the tracing support of an enterprise stealing and discharging to the sewage discharge root is lacked, particularly, the source is difficult to accurately lock under the condition that a plurality of enterprises exist in the peripheral distribution of the sewage discharge outlet, the automation degree is not high enough, and a certain gap is left in the requirement of the law enforcement on the aspect of sudden impact on the technology.
Disclosure of Invention
The invention aims to provide a sewage dark drainage detection tracing method based on unmanned aerial vehicle thermal infrared remote sensing, and aims to solve the problems in the prior art.
In order to achieve the purpose, the invention provides a sewage dark drainage detection tracing method based on unmanned aerial vehicle thermal infrared remote sensing, which comprises the following steps:
acquiring decimeter-level thermal infrared image data along a preset airway based on an unmanned aerial vehicle and an infrared imager;
extracting homonymous feature points of an overlapping area in the thermal infrared image data, and correcting and splicing the thermal infrared images;
converting the data in the spliced thermal infrared images into a geographical coordinate system frame of a high-definition map to obtain thermal infrared images with geographical parameters;
extracting a water body area in the thermal infrared image with the geographic parameters, segmenting the water body area based on a temperature value, and extracting a temperature abnormal water area;
and analyzing the land part based on the water area with abnormal temperature and the thermal infrared image with the geographic parameters, and searching a temperature abnormal linear target with connectivity on the land as a suspected candid camera pipeline, wherein the end of the suspected candid camera pipeline is a suspected candid excretion main body.
Optionally, the process of collecting thermal infrared image data along a preset airway based on the unmanned aerial vehicle and the infrared imager comprises:
carrying the infrared imager by the unmanned aerial vehicle to construct an unmanned aerial vehicle imaging system, wherein the unmanned aerial vehicle is internally provided with a positioning system;
acquiring environmental information of a region to be measured in advance, and planning a preset airway based on the environmental information and visible light image data;
the unmanned aerial vehicle imaging system flies along the air route and collects decimeter-level thermal infrared image data.
Optionally, the unmanned aerial vehicle imaging system flies along the air route, and the process of collecting thermal infrared image data includes:
guarantee that there is overlap region adjacent twice shooting, and infrared imager will trigger the timestamp of shooing and bind with unmanned aerial vehicle's positioning data when shooing hot infrared image at every turn, obtain the positioning data of every image at the formation of image constantly.
Optionally, the extracting the feature points with the same name in the overlapping area in the thermal infrared image data, and the process of splicing the thermal infrared images includes:
performing feature extraction in the overlapping area of the two adjacent thermal infrared images based on a SHIF operator to obtain homonymous pixels;
based on a random consensus (RANSAC) method, removing wrong matching corresponding points in the homonymous pixels to obtain homonymous feature points;
and calculating the characteristic points with the same name based on an aerial photography method, and correcting and splicing the images.
Optionally, the process of converting the data in the spliced thermal infrared image into a geographic coordinate system frame of a high-definition map to obtain a thermal infrared image with geographic parameters includes:
and based on an interactive point selection mode, selecting matching points corresponding to the same name on the spliced thermal infrared image and the high-definition map to obtain a conversion relation from the spliced thermal infrared image to the high-definition map, and based on the conversion relation, converting data in the spliced thermal infrared image into a geographic coordinate system frame of the high-definition map to obtain the thermal infrared image with geographic parameters.
Optionally, the process of extracting the water body region in the thermal infrared image with the geographic parameter includes:
acquiring a deep learning model based on an SU-Net framework to extract a water body region in a visible light image;
distinguishing the water body area from the land area and marking;
and extracting the water body area from the thermal infrared image with the geographic parameters based on the marked content.
Optionally, the water body region is segmented based on the temperature value, and the process of extracting the water body with abnormal temperature includes:
performing threshold segmentation on the abnormal region of the temperature value based on a visual saliency model algorithm of the spectrum residual error to obtain a binary image;
and processing the binary image to obtain a temperature abnormal water area.
Optionally, the processing the binary image to obtain the abnormal-temperature water area includes:
and eliminating noise in the binary image based on corrosion operation, and filling the small holes in the noise-reduced binary image by adopting expansion operation to obtain a final binary image serving as the temperature abnormal water area.
The invention has the technical effects that:
by utilizing the portable characteristic of the unmanned aerial vehicle, the operation can be flexibly carried out, the operation cost is reduced for the industrial sewage steal-discharge detection work, and the time efficiency is improved;
the radiation inversion characteristic of thermal infrared remote sensing is utilized, and the detection capability of pipelines and immersed steal drainage is improved;
based on an aerial photogrammetry theory and a computer vision method, a plurality of images with overlapped parts are registered in a function mode, and relevant information such as the position, the area and the length of a polluted area and suspected source information can be provided;
by utilizing a deep learning calculation method, the water body range in the image can be extracted, the polluted area is divided from the infrared target in the water body range by a visual significance model method based on the spectrum residual error, and the stealing, draining and tracing are realized by combining the thermal infrared image of the land part and the connectivity of the thermal infrared image.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application. In the drawings:
FIG. 1 is a schematic flow chart in example 1 of the present invention;
fig. 2 is a schematic diagram of an SU-Net deep learning model in embodiment 1 of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer-executable instructions and that, although a logical order is illustrated in the flowcharts, in some cases, the steps illustrated or described may be performed in an order different than presented herein.
Example 1
As shown in fig. 1, the present embodiment provides a method for detecting and tracing to a source of dark sewage based on unmanned aerial vehicle thermal infrared remote sensing, which includes:
step S1: building an unmanned aerial vehicle-mounted imaging equipment system;
a fixed-wing unmanned aerial vehicle is used as a flying carrier, and a thermal infrared imager is equipped. And then, calibrating camera parameters of the imager on the ground. Meanwhile, the unmanned aerial vehicle carries a global positioning system positioning device, such as a GPS or Beidou positioning module. Binding a timestamp triggering the imager to shoot with positioning data to provide positioning parameters of the imaging time of the image, and storing the thermal infrared image shot in the flight process by using a built-in memory card of the imager. The thermal infrared image shot by the airborne unmanned aerial vehicle can observe that a single thermal infrared image which is not subjected to registration splicing can only reflect the scene of a small-range area; in addition, a certain degree of overlapping exists between adjacent images, and therefore necessary homonymy point searching basis is provided for correcting splicing.
The fixed wing unmanned aerial vehicle in this embodiment can additionally carry on visible light camera for gather the visible light image in step.
Step S2: environmental information acquired by data is investigated, route planning is prepared, the unmanned aerial vehicle performs flight operation, a flight task is executed, and thermal infrared image data are acquired.
The method comprises the steps of collecting weather conditions and terrain information in advance, avoiding carrying out data acquisition in time under severe weather conditions, avoiding potential dangerous objects capable of causing flight safety to the unmanned aerial vehicle, designing appropriate flight altitude by combining sensor parameters to ensure that image ground resolution can reach a decimetric level (better than 30cm), determining air route planning information, and providing necessary preparation information for unmanned aerial vehicle flight field data acquisition.
And starting field collection, carrying equipment by using an unmanned aerial vehicle, flying according to the path planned by the air route, obtaining a thermal infrared image data sequence of the observation area, and storing the thermal infrared image data sequence in an onboard memory card.
To ensure that there is more than 50% overlap between two adjacent thermal infrared images, the field angle of the thermal infrared images is known
The following conditions are satisfied by the setting of the flying speed v, the flying height h and the thermal infrared image acquisition interval t of the unmanned aerial vehicle:
the overlapping degree between two adjacent thermal infrared images can be set to be more than 50%, and is recommended to be not less than 30%, otherwise, image splicing of the global observation area is affected.
Step S3: and extracting homonymous characteristic points between the thermal infrared images with the superposition to splice the thermal infrared images.
The homonymous feature points between the thermal infrared images with the overlap are extracted. The acquisition of the image data is carried out continuously, and the thermal infrared images in the observation range need to be registered and spliced in order to obtain a complete large image of the observation region. In the overlapping area between adjacent images, according to the characteristic matching technology in the image processing method, a SHIF operator is used for carrying out characteristic extraction, the pixels with the same name between the overlapping images are found, and the RANSAC method is combined to remove wrong matching corresponding points.
Step S4: geographic parameter registration
And the data on the unmanned aerial vehicle is converted into the geographic coordinate system frame where the existing high-definition map is located. After the large images are spliced, the existing high-definition map, such as a topographic map or a satellite image of an observation area, is used, and an interactive point selection method is adopted, matching points corresponding to the same name are selected on the thermal infrared large images which are spliced in a registering mode and the existing geographic reference high-definition map, and a conversion relation of mapping the spliced thermal infrared large images to the existing geographic reference map is obtained, so that data carried by the unmanned aerial vehicle is converted into a geographic coordinate system frame where the existing map is located.
Step S5: extracting thermal infrared image water body area
And extracting a water body region in the visible light image through a deep learning model, distinguishing a water body part from a land part, and outputting a classification mark as 0 (non-water body) or 1 (water body). And then the thermal infrared image water body area is divided.
The water body extraction method based on the SU-Net architecture comprises the following steps:
the input image is mapped to a pixel classification map, where w is the width, h is the height, and c is the number of input image features. For the prediction of the water body, the output classification label is 0 (non-water body) or 1 (water body). The convolutional network can be described as a mapping function, given by:
Y＝f(X；θ)
where θ is a parameter of the mapping function. The function f can be decomposed into a plurality of sub-functions f1 and f2, and by increasing the number of sub-functions, deeper networks can be obtained, and the fitting effect is enhanced.
Y＝f2(f1(X；θ)；θ1)
Therefore, the mapping function is changed into a combination of a plurality of simple subfunctions or layers, and finally a multi-layer combined deep layer model is obtained. The most important layer in these models is the convolutional layer, which generates a feature map by convolving the input with a specified number of kernels. In addition to convolutional layers, pooling layers are also used to down-sample feature maps. By spatially dividing the feature map into 2 × 2 blocks and taking the maximum value to pass down, the data amount of the feature map is reduced by 75% by this operation.
The right half of the U-Net network structure is restored to the original size of the image by up-sampling the image by deconvolution. And in the final convolution layer, using Sigmoid as an activation function and outputting the predicted water body probability. SU-Net differs from U-Net in that convolution and deconvolution operations are performed interleaved so that, at the same number of layers, SU-Net can appear more connections to retain more information. Calculating calculated loss values of the artificially interpreted water body mask and the predicted water body mask using a binary cross entropy loss function in the training model:
wherein
Is a determined water body area which is interpreted manually, Y is the model predicted water body result, w is the width of the image, and h is the height of the image.
The SU-Net network model shown in FIG. 2 is composed of layers (convolution path and deconvolution path) with three scales, including 19 convolution layers. The network input layer is an image of 256 × c, where c is the number of channels (features) and takes different values for different experiments c. The deconvolution layer and the convolution layer are corresponding and mainly consist of two convolutions of 3x 3; each convolution is followed by a linear function (ReLU) and a 2x2 max pool operation. In each down-sampling step, we double the number of eigen-channels, shrinking the feature size. For the deconvolution operation, each step contains one upsampled feature map, followed by a 2x2 convolution ("convolution"), which halves the number of feature channels, concatenated with the corresponding clipped feature map from the systolic path, and two 3x3 convolutions, each reduced by the ReLU. At the final level, each 32-component feature vector is mapped to the required class number using a 1x1 convolution. The output level is a category confidence map of 256 × 1. Thus, each pixel has a confidence value as to whether it is a body of water.
As shown in table 1, the parameters involved in the convolution are listed in the table. The left and right sides of table 1 describe the configurations of the encoder and decoder, respectively. Each conv or deconv layer is followed by a RELU layer that is merged into its corresponding convolutional or deconvolution layer. Starting with the input image, the features flow down through the left layer, then up through the right layer, and finally to the output layer. As indicated by the arrows in the table below, skipping the connection allows the function to pass directly from the same level of coding layer to the decoding layer without reducing the spatial resolution. By skipping the coding layer, feature details are preserved to ensure accurate segmentation.
And improper application of the features can reduce the recognition rate of the target object, and on the basis of the model, the model performance of several different feature combinations is researched by inputting different feature quantity training models. Therefore, SU-Net will be studied based on its performance in 3 different frequency band combinations, mainly (i) red/green/blue (RGB) bands, (ii) RGB + GLCM, (iii) RGB + Gabor.
TABLE 1
The left and right columns in the table correspond to the encoder and decoder, respectively. The encoded convolutional layers evolve downward, while the decoded convolutional layers evolve upward. Each conv (or deconv) layer is followed by a RELU layer. The arrows indicate the connections. Each convolution/deconvolution layer lists 3 numbers. The first 2 table filter sizes, the last one represents the number of filters for that layer.
Step S6: segmenting the water body area based on temperature values, and extracting abnormal pixel areas of the temperature field
The contaminated water body area may be understood as a target area and the uncontaminated water body area is defined as a background area. The temperature values of the pixels in each area have the characteristic of being relatively consistent, the relative temperature values of the target area and the background area are differentiated to a certain degree, and the area range where the target is located can be divided by utilizing a visual saliency model algorithm based on the spectrum residual error. The binary black-and-white image is obtained after threshold segmentation, wherein white pixel point parts comprise approximate regions of targets and some unnecessary isolated noise points, and places where the white regions are not communicated exist, weak particle pixel noise in the binary image is eliminated by adopting corrosion operation in morphological processing, small holes in the binary image are filled by adopting expansion operation, isolated points and discontinuous regions of the target regions are eliminated, subsequent processing of the image saliency map is completed, and a final binary image is obtained.
Example 2
Acquiring thermal infrared data, making a flight plan according to existing visible light image data (satellite remote sensing images or unmanned aerial vehicle remote sensing images) and pre-acquired flight environment information, carrying a calibrated thermal infrared sensor by using an unmanned aerial vehicle to execute a flight task, acquiring thermal infrared image data, and recommending that night flight continues to use flight control related parameters in daytime flight;
splicing the thermal infrared images, extracting homonymous feature points between the thermal infrared images with the overlapped areas by combining a SHIF (short distance image correlation) and random consistency test (RANSAC) method, calculating matching parameters by using an aerial photography method, and correcting and splicing the images;
the process of calculating matching parameters based on the aerial photography method and performing image correction and splicing comprises the following steps:
the image matching adopted by the invention is a technology for synthesizing a whole wide-view-angle large image by processing a series of images with overlapped areas in the same scene, and has the advantages that: the infrared image has strong anti-interference capability and can work at night, but the infrared image splicing technology is an effective way due to the small detection field, and the method is adopted.
The specific process of image matching comprises the following steps: and registering two or more images with overlapped parts in a functional form to realize the one-to-one corresponding mapping relation of the same parts of the two images. Since matching between images is actually matching between pixel coordinates of images, if there is a point P (x, y) on an image I, and transformation is performed by rotation, scaling, translation, affine, etc., there is a similar point P (x ', y ') on an image I ', and transformation in the form of rotation, scaling, translation, affine, etc. can establish a functional relation between the coordinates of the point P on I and I ' by mathematical relations, denoted as f (x), and if there are sets of similar points P on I and I ', and the sets are represented by I (x, y) and I ' (x ', y '), there are I (x, y) ═ I ' (x ', y ') f (x), and the relation between two matching images is decomposed as follows:
in the formula, T: translating the model; r: rotating the model; z: scaling the model; v: stretching, etc. The transformation of the image is generally unknown, and there are local transformation and stretching and overall transformation and stretching, but for oblique aerial images, certain change rules are generally followed, so that the optimal solution of the function can be conveniently found. The core of the image matching adopted by the invention is to carry out different algorithms around the expansion so as to improve the calculation efficiency, robustness, stability and the like.
Geographic parameter registration, so that data carried by the unmanned aerial vehicle is converted to a geographic coordinate system frame where the existing high-definition map is located;
performing thermal infrared image water body segmentation, and indirectly extracting the water body of the spliced large image by using a deep learning water body extraction method based on an SU-Net framework;
and (3) recognizing the stealing sewage pollution, namely performing infrared target segmentation calculation of a visual saliency model based on spectrum residual error in each water area, and extracting a water pixel area with a temperature difference value larger than a certain threshold value in each water area, wherein the water pixel area is an extracted candidate stealing sewage area.
Calculating the infrared target segmentation based on the visual saliency model of the spectrum residual:
since thermal infrared imaging uses the radiant energy of the target, making the target region significant with respect to the background, a significance model can be used to segment the region where the target is located. A visual saliency model based on spectrum Residual error, namely a spectrum Residual error method (SR), which is simple and rapid in calculation and high in accuracy rate is selected in the experiment. The basic idea of the SR method is as follows: the log spectrum of each image has a similar trend, while the varying parts of the image, i.e. the salient regions, are caused by spectral residuals. By using the inverse fourier transform, an output image called a saliency map can be constructed in the spatial domain.
The spectrum obtained from the thermal infrared image i (x) after fourier transform can be decomposed into two parts, namely an amplitude spectrum a (f) and a phase spectrum p (f).
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
And then carrying out logarithmic transformation on the magnitude spectrum, and subtracting the logarithmic magnitude spectrum L (f) from the logarithmic magnitude spectrum which is subjected to convolution and smoothing by a low-pass filter to obtain a spectrum residual error.
L(f)＝log A(f)R(f)＝L(f)-hn(f)*L(f)
And finally, performing inverse Fourier transform on the frequency spectrum residual error and the initial phase spectrum to obtain a saliency map of the original thermal infrared image, and smoothing the saliency map through a Gaussian filter to obtain a better visual effect.
S(x)＝g(x)*|FFT-1[exp(R(f))+i×P(f)]|2
Wherein, A (f), P (f), L (f), R (f), S (x) A are respectively the amplitude spectrum, phase spectrum, logarithmic phase spectrum, spectrum residual error and saliency map of the original thermal infrared image; FTT, FFT-1Representing the fourier transform and its inverse; h isn(f) It is shown that the low-pass filter in the frequency domain is here chosen as the mean filter; g (x) is a spatial low pass filter, which may be a gaussian smoothing filter.
According to the obtained saliency map S (x), a simple threshold segmentation algorithm is adopted by utilizing a threshold pair S (x) obtained by the following formula, so that a binary image obtained by segmenting a conventional water body area and stealing and excluding candidate areas by using an original thermal infrared image is obtained.
threshold＝3×E(S(x))
Tracing the steal enterprise, comparing and analyzing the steal candidate area and the land part thermal infrared image, and searching a temperature abnormal linear target which is communicated with the land part on the land as a suspected steal pipeline, wherein the enterprise pointed by the other end of the target is a suspected steal main body.
Because the buried depth of the hidden pipe, the concealing mode, the using material and the like can all image the characteristics of the hidden pipe on the thermal infrared image, the tracing mainly adopts the mode of superposing, analyzing and visually interpreting visible light and thermal infrared images, and the specific method comprises the following steps: searching towards the land and the bank direction by taking the steal candidate area as the center on the thermal infrared image; extracting a linear target with abnormal temperature, wherein the temperature is higher when the stolen sewage passes through the concealed pipe; further identifying a linear target which has connectivity with the stealing candidate area as a suspected hidden pipe, wherein the connected position is a hidden pipe terminal point; continuously searching along the linear target in the direction opposite to the stealing direction, and mainly checking the breakpoint of the linear target; if no other linear target with similar characteristics exists around the breakpoint, analyzing whether the breakpoint is a large machine, a sewage pool, a factory workshop or not by combining with a visible light image, comprehensively judging whether the breakpoint is suspected to be a concealed conduit starting point or not, and recording the position information of the point; and combining auxiliary information such as a map, an electronic map and the like, wherein the enterprise at the point is a suspected steal subject.
The above description is only for the preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. A hidden sewage discharge detection and source tracing method based on unmanned aerial vehicle thermal infrared remote sensing is characterized by comprising the following steps:
acquiring decimeter-level thermal infrared image data along a preset airway based on an unmanned aerial vehicle and an infrared imager;
extracting homonymous feature points of an overlapping area in the thermal infrared image data, and correcting and splicing the thermal infrared images;
converting the data in the spliced thermal infrared images into a geographical coordinate system frame of a high-definition map to obtain thermal infrared images with geographical parameters;
extracting a water body area in the thermal infrared image with the geographic parameters, segmenting the water body area based on a temperature value, and extracting a temperature abnormal water area;
and analyzing the land part based on the water area with abnormal temperature and the thermal infrared image with the geographic parameters, and searching a temperature abnormal linear target with connectivity on the land as a suspected candid camera pipeline, wherein the end of the suspected candid camera pipeline is a suspected candid excretion main body.
2. The method of claim 1, wherein the acquiring thermal infrared image data along a preset airway based on the unmanned aerial vehicle and the infrared imager comprises:
carrying the infrared imager by the unmanned aerial vehicle to construct an unmanned aerial vehicle imaging system, wherein the unmanned aerial vehicle is internally provided with a positioning system;
the method comprises the steps of collecting environmental information of a region to be measured in advance, and planning a preset airway based on the environmental information and visible light image data;
the unmanned aerial vehicle imaging system flies along the air route and collects thermal infrared image data.
3. The method of claim 2, wherein the drone imaging system flies along the airway, and wherein collecting thermal infrared image data comprises:
guarantee that there is overlap region adjacent twice shooting, and infrared imager will trigger the timestamp of shooing and bind with unmanned aerial vehicle's positioning data when shooing hot infrared image at every turn, obtain the positioning data of every image at the formation of image constantly.
4. The method according to claim 1, wherein the extracting the homonymous feature points of the overlapping region in the thermal infrared image data and performing thermal infrared image correction and stitching comprises:
performing feature extraction in the overlapping area of the two adjacent thermal infrared images based on a SHIF operator to obtain homonymous pixels;
based on a random consensus (RANSAC) method, removing wrong matching corresponding points in the homonymous pixels to obtain homonymous feature points;
and calculating the characteristic points with the same name based on an aerial photography method, and correcting and splicing the images.
5. The method of claim 1, wherein the step of converting the data in the stitched thermal infrared image into a geographic coordinate system frame of a high definition map to obtain the thermal infrared image with geographic parameters comprises:
and based on an interactive point selection mode, selecting matching points corresponding to the same name on the spliced thermal infrared image and the high-definition map to obtain a conversion relation from the spliced thermal infrared image to the high-definition map, and based on the conversion relation, converting data in the spliced thermal infrared image into a geographic coordinate system frame of the high-definition map to obtain the thermal infrared image with geographic parameters.
6. The method of claim 1, wherein extracting the water body region in the thermal infrared image with the geographic parameters comprises:
acquiring a deep learning model based on an SU-Net framework to extract a water body region in a visible light image;
distinguishing the water body area from the land area and marking;
and extracting the water body area from the thermal infrared image with the geographic parameters based on the marked content.
7. The method of claim 1, wherein the water body region is segmented based on the temperature value, and the extracting the water body with abnormal temperature comprises:
performing threshold segmentation on the abnormal region of the temperature value based on a visual saliency model algorithm of the spectrum residual error to obtain a binary image;
and processing the binary image to obtain a temperature abnormal water area.
8. The method according to claim 7, wherein the processing the binary image to obtain the abnormal-temperature water area comprises:
and eliminating noise in the binary image based on corrosion operation, and filling the small holes in the noise-reduced binary image by adopting expansion operation to obtain a final binary image serving as the temperature abnormal water area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210010640.5A CN114359731A (en) | 2022-01-05 | 2022-01-05 | Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210010640.5A CN114359731A (en) | 2022-01-05 | 2022-01-05 | Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114359731A true CN114359731A (en) | 2022-04-15 |
Family
ID=81108137
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210010640.5A Pending CN114359731A (en) | 2022-01-05 | 2022-01-05 | Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114359731A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106094872A (en) * | 2016-08-24 | 2016-11-09 | 北京大学深圳研究生院 | A kind of sewage based on unmanned plane secretly arranges mouthful detection method and system |
| CN110889327A (en) * | 2019-10-16 | 2020-03-17 | 南京航空航天大学 | Intelligent detection method for sewage draining exit around water area based on thermal infrared image |
-
2022
- 2022-01-05 CN CN202210010640.5A patent/CN114359731A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106094872A (en) * | 2016-08-24 | 2016-11-09 | 北京大学深圳研究生院 | A kind of sewage based on unmanned plane secretly arranges mouthful detection method and system |
| CN110889327A (en) * | 2019-10-16 | 2020-03-17 | 南京航空航天大学 | Intelligent detection method for sewage draining exit around water area based on thermal infrared image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583369B (en) | Target identification method and device based on target area segmentation network | |
| CN109255317B (en) | Aerial image difference detection method based on double networks | |
| CN110889327B (en) | Intelligent detection method for sewage outlet around water area based on thermal infrared image | |
| Singh et al. | WSB-DA: water surface boundary detection algorithm using Landsat 8 OLI data | |
| Boggess | Identification of roads in satellite imagery using artificial neural networks: A contextual approach | |
| Agapiou et al. | Colour to Greyscale Pixels: Re‐seeing Greyscale Archived Aerial Photographs and Declassified Satellite CORONA Images Based on Image Fusion Techniques | |
| Bektas Balcik et al. | Determination of magnitude and direction of land use/land cover changes in Terkos Water Basin, Istanbul | |
| Oșlobanu et al. | Built-up area analysis using Sentinel data in metropolitan areas of Transylvania, Romania | |
| CN114359731A (en) | Sewage hidden discharge detection traceability method based on unmanned aerial vehicle thermal infrared remote sensing | |
| Parape et al. | Optimization of structure elements for morphological hit-or-miss transform for building extraction from VHR airborne imagery in natural hazard areas | |
| Xu et al. | Oil spill discrimination of multi-time-domain shipborne radar images using active contour model | |
| Hoque et al. | Seagrass propeller scar detection using deep convolutional neural network | |
| Yue et al. | The extraction of water information based on SPOT5 image using object-oriented method | |
| Barazzetti et al. | Identification of buildings damaged by natural hazards using very high-resolution satellite images: The case of earthquake in L’Aquila, Italy | |
| Ali et al. | Automatic extractions of road intersections from satellite imagery in urban areas | |
| Cardoso | Computer aided recognition of man-made structures in aerial photographs | |
| Mudau et al. | Extraction of low cost houses from a high spatial resolution satellite imagery using Canny edge detection filter | |
| Liu et al. | Building segmentation in agricultural land using high resolution satellite imagery based on deep learning approach | |
| Corbane et al. | Detection of changes in built-up areas with a fully convolutional network in the context of the European settlement map | |
| Triantafyllia-Maria et al. | Water Surface Level Monitoring of the Axios River Wetlands, Greece, Using Airborne and Space-Borne Earth Observation Data | |
| Zhong et al. | Rapid screening of potential discharge outlets into rivers on UAV images using a remote sensing change detection method | |
| Dippold et al. | Potential Exploration of Segmentation Driven Stereo Matching of Very High-Resolution Satellite Imagery | |
| Najibi et al. | A New approach to update urban digital maps using high resolution satellite images and GIS tools (Case study: Beijing City) | |
| Tuama | Bigdata Based Disaster Monitoring of Satellite Image Processing Using Progressive Image Classification Algorithm | |
| Jia | Automatic ground control points refinement for remote sensing imagery registration |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220415 |