CN113343930A - Unmanned aerial vehicle image processing method based on Gaussian denoising - Google Patents
Unmanned aerial vehicle image processing method based on Gaussian denoising Download PDFInfo
- Publication number
- CN113343930A CN113343930A CN202110755666.8A CN202110755666A CN113343930A CN 113343930 A CN113343930 A CN 113343930A CN 202110755666 A CN202110755666 A CN 202110755666A CN 113343930 A CN113343930 A CN 113343930A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- image
- dimensional
- phase
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
- Processing Or Creating Images (AREA)
Abstract
The unmanned aerial vehicle image processing method based on Gaussian denoising, which is claimed by the invention, obtains an unmanned aerial vehicle three-dimensional geographic image sample classifier by training an unmanned aerial vehicle three-dimensional geographic image sample, then tests the unmanned aerial vehicle three-dimensional geographic image sample classifier, and obtains an unmanned aerial vehicle three-dimensional geographic image sample classification test result and an on-site three-dimensional geographic evaluation result by using Gaussian and median filtering denoising; performing wavelet detection analysis to obtain a dynamic detection result of the on-site three-dimensional geography; and (4) carrying out environment detection and analysis to obtain an image processing result of the on-site three-dimensional geography. The invention achieves good multi-feature fusion clustering analysis effect by applying the advantages of a wavelet transform processing mode to unmanned aerial vehicle image processing, not only focuses on image features, but also integrates an environment sensor to assist in unmanned aerial vehicle image combination, obtains a comprehensive processing result, identifies unmanned aerial vehicle ground images in an all-around manner, and achieves the effect of multi-dimension accurate determination of an examination strategy.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicles, in particular to an unmanned aerial vehicle image processing method based on Gaussian denoising.
Background
The unmanned plane is called unmanned plane for short, and is an unmanned plane operated by radio remote control equipment and a self-contained program control device. The machine has no cockpit, but is provided with an automatic pilot, a program control device, a signal acquisition device and other equipment. The personnel on the ground, the naval vessel or the mother aircraft remote control station can track, position, remotely control, telemeter and digitally transmit the personnel through equipment such as a radar. The aircraft can take off like a common airplane under the radio remote control or launch and lift off by a boosting rocket, and can also be thrown into the air by a mother aircraft for flying. In the prior art, an unmanned aerial vehicle has an image tracking technology, and tracking images to acquire tracking contents.
The method for acquiring the geographic position image comprises the steps of controlling the periodic change of the brightness of each light source of at least two light sources which are separately arranged on a space, and respectively acquiring images for three-dimensional reconstruction by adopting three cameras at least three positions. The image-based three-dimensional reconstruction is a process of automatically calculating and matching by a computer according to two or more two-dimensional images shot by an object or a scene, calculating two-dimensional geometric information and depth information of the object or the scene, and establishing a three-dimensional stereo model. Along with the development of unmanned aerial vehicle technique, consumption level unmanned aerial vehicle price constantly descends to and laser radar's miniaturized portableization, and medium-size and small-size unmanned aerial vehicle carries on laser radar and carries out geographical survey and drawing and possibly. At present unmanned aerial vehicle Lidar survey and drawing generally uses ground mark target location and ground basic station complex mode, but uses not strong to the adaptability nature of different landforms, needs artificial exploration in advance, goes to establish the mark target around the survey and drawing place, and the loaded down with trivial details inefficiency of process. Meanwhile, the acquired point cloud data needs to be processed offline at a base station, and the real-time performance is not good enough.
Meanwhile, the processing of the unmanned aerial vehicle image lacks precision denoising and other processing, so that subsequent image correlation operation is facilitated; the detection of the current geographic image of the unmanned aerial vehicle is mainly limited to the singularity of feature recognition, namely, the geographic geology and other recognition are carried out only by tracking the image, and the comprehensive three-dimensional geographic environment information processing cannot be carried out according to multiple sensors.
Disclosure of Invention
The invention discloses an unmanned aerial vehicle image processing method based on Gaussian denoising, and aims to solve the problems of inaccurate detection and three-dimensional geographic exploration strategy deviation in the tracking of a geographic image by a current unmanned aerial vehicle.
The invention requests to protect an unmanned aerial vehicle image processing method based on Gaussian denoising, which is characterized by comprising the following steps:
training a three-dimensional geographic image sample of the unmanned aerial vehicle to obtain a three-dimensional geographic image sample classifier of the unmanned aerial vehicle;
acquiring an on-site three-dimensional geographic sampling image of an unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by adopting the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, and acquiring a classification test result and an on-site three-dimensional geographic evaluation result of the three-dimensional geographic image sample of the unmanned aerial vehicle;
performing wavelet detection analysis on the evaluation result of the on-site three-dimensional geography to obtain a dynamic detection result of the on-site three-dimensional geography;
and carrying out environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography.
Preferably, the training of the three-dimensional geographic image sample of the unmanned aerial vehicle to obtain the three-dimensional geographic image sample classifier of the unmanned aerial vehicle further includes:
acquiring an original phase image of the three-dimensional geographic image sample of the training unmanned aerial vehicle, and extracting dynamic characteristics and color characteristics of the original phase image;
respectively carrying out normalization processing on the dynamic features and the color features of the original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample to obtain the dynamic features and the color features of the normalized original phase image;
fusing the dynamic characteristics and the color characteristics of the normalized original phase image to form dual-phase unmanned aerial vehicle image characteristics;
manually acquiring C double-phase positive training samples and D double-phase negative training samples from an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, wherein 0< C <500, and 0< D < 1000;
expressing the C double-phase forward training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain the characteristics of the double-phase forward training samples;
representing all (C + D) double-phase training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain all double-phase training sample characteristics;
and constructing a support classifier SVC by adopting the characteristics of the two-phase forward training sample and all the characteristics of the two-phase training sample.
Preferably, the acquiring an on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by using the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, and acquiring a classification test result and an on-site three-dimensional geographic evaluation result of the unmanned aerial vehicle further includes:
controlling a lens to capture a digital image, carrying out graying, drying removal and image size resetting on the acquired digital image, and removing noise by using Gaussian filtering and median filtering;
judging the defocusing degree of the image by using an image definition evaluation algorithm, finding a focus position by using a focus searching method, and driving the unmanned aerial vehicle lens to an optimal imaging position;
inputting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle into a wavelet neural network of a first layer by adopting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle, and respectively calculating the output of the hidden layer of the first layer and the output of the output layer by utilizing the initial weights of the hidden layer and the input layer nodes of the first layer network, the initial weights of the nodes of the output layer and the hidden layer, and the scaling factor and the shifting factor of the wavelet activation function;
inputting the output of the first network hidden layer and the output of the output layer into the support classifier SVC constructed by the characteristics of the double-phase forward training sample and the characteristics of all the double-phase training samples to obtain a final unmanned aerial vehicle image test result;
and analyzing the test result in the field of three-dimensional geography, clustering the field of three-dimensional geography, and acquiring texture characteristics and hardness characteristics of the field of three-dimensional geography.
Preferably, the performing wavelet detection analysis on the evaluation result of the solid three-dimensional geography to obtain a dynamic detection result of the solid three-dimensional geography further includes:
performing wavelet detection analysis on the content of the solid three-dimensional geographic evaluation result, initializing the content of the solid three-dimensional geographic evaluation result, and packaging into a calling data file;
sending the packaged calling data file to a wavelet detection analysis program, reading the calling data file, and performing wavelet packet decomposition and reconstruction to complete feature extraction;
sending out a reconstruction signal by adopting a principal component analysis method, establishing a covariance matrix after average preprocessing to obtain a characteristic value eigenvector, and taking data obtained after projection as a new data sample to reduce the dimension of original data;
and acquiring a dynamic detection result of the solid three-dimensional geography according to the data obtained after the projection, wherein the dynamic detection result of the solid three-dimensional geography shows the potential dynamic change of the solid three-dimensional geography.
The performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography further comprises:
acquiring a dynamic detection result of the on-site three-dimensional geography, and carrying out cluster analysis on the dynamic detection result of the on-site three-dimensional geography;
when the dynamic detection result of the solid three-dimensional geography belongs to a first cluster, performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography, detecting the odor concentration in the current environment, planning a path to track and position an odor source, and finally obtaining an image processing result of the solid three-dimensional geography;
and when the dynamic detection result of the on-site three-dimensional geography belongs to a second cluster, detecting the odor concentration in the current environment without detection.
The unmanned aerial vehicle image processing method based on Gaussian denoising, which is claimed by the invention, obtains an unmanned aerial vehicle three-dimensional geographic image sample classifier by training an unmanned aerial vehicle three-dimensional geographic image sample, then tests the unmanned aerial vehicle three-dimensional geographic image sample classifier, and obtains an unmanned aerial vehicle three-dimensional geographic image sample classification test result and an on-site three-dimensional geographic evaluation result by using Gaussian and median filtering denoising; performing wavelet detection analysis to obtain a dynamic detection result of the on-site three-dimensional geography; and (4) carrying out environment detection and analysis to obtain an image processing result of the on-site three-dimensional geography. The invention achieves good multi-feature fusion clustering analysis effect by applying the advantages of a wavelet transform processing mode to unmanned aerial vehicle image processing, not only focuses on image features, but also integrates an environment sensor to assist in unmanned aerial vehicle image combination, obtains a comprehensive processing result, identifies unmanned aerial vehicle ground images in an all-around manner, and achieves the effect of multi-dimension accurate determination of an examination strategy.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flowchart of a method for processing an image of an unmanned aerial vehicle based on Gaussian denoising according to the present invention;
fig. 2 is a first embodiment of an unmanned aerial vehicle image processing method based on gaussian denoising according to the present invention;
fig. 3 is a second embodiment of the unmanned aerial vehicle image processing method based on gaussian denoising according to the present invention;
fig. 4 is a third embodiment of the unmanned aerial vehicle image processing method based on gaussian denoising according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms "first" and "second," and the like in the description and claims of the present invention and the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not set forth for a listed step or element but may include steps or elements not listed.
Referring to the attached figure 1, the invention requests to protect an unmanned aerial vehicle image processing method based on Gaussian denoising, and is characterized in that:
training a three-dimensional geographic image sample of the unmanned aerial vehicle to obtain a three-dimensional geographic image sample classifier of the unmanned aerial vehicle;
acquiring an on-site three-dimensional geographic sampling image of an unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by adopting the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, and acquiring a classification test result and an on-site three-dimensional geographic evaluation result of the three-dimensional geographic image sample of the unmanned aerial vehicle;
performing wavelet detection analysis on the evaluation result of the on-site three-dimensional geography to obtain a dynamic detection result of the on-site three-dimensional geography;
and carrying out environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography.
Preferably, referring to fig. 2, in a first embodiment of the unmanned aerial vehicle image processing method based on gaussian denoising, the training of the unmanned aerial vehicle three-dimensional geographic image sample to obtain the unmanned aerial vehicle three-dimensional geographic image sample classifier further includes:
acquiring an original phase image of the three-dimensional geographic image sample of the training unmanned aerial vehicle, and extracting dynamic characteristics and color characteristics of the original phase image;
respectively carrying out normalization processing on the dynamic features and the color features of the original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample to obtain the dynamic features and the color features of the normalized original phase image;
fusing the dynamic characteristics and the color characteristics of the normalized original phase image to form dual-phase unmanned aerial vehicle image characteristics;
manually acquiring C double-phase positive training samples and D double-phase negative training samples from an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, wherein 0< C <500, and 0< D < 1000;
expressing the C double-phase forward training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain the characteristics of the double-phase forward training samples;
representing all (C + D) double-phase training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain all double-phase training sample characteristics;
and constructing a support classifier SVC by adopting the characteristics of the two-phase forward training sample and all the characteristics of the two-phase training sample.
The method comprises the following steps of obtaining an original phase image of a three-dimensional geographic image sample of the training unmanned aerial vehicle, extracting dynamic characteristics and color characteristics of the original phase image, and specifically comprising the following steps:
extracting a gray value vector of an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, and using the gray value vector as a dynamic feature;
carrying out Gabor transformation in M scales and N directions on an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, selecting a central pixel, and extracting mean information and variance information of a high-pass sub-band RGB coefficient on a window;
taking pixel points of original phase images of all the three-dimensional geographic image samples of the training unmanned aerial vehicle in the images as central pixel points, and extracting the mean information and the variance information to obtain a mean vector and a variance vector;
and jointly forming the color characteristics of the original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample by using the mean vector and the variance vector.
The normalizing processing is respectively carried out on the dynamic features and the color features of the original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, and the obtaining of the dynamic features and the color features of the normalized original phase image comprises the following steps:
wherein the content of the first and second substances,
representing dynamic features of an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample;
representing color features of an original phase image of the trained drone three-dimensional geographic image sample;
representing dynamic characteristics of an original phase image of the normalized training unmanned aerial vehicle three-dimensional geographic image sample;
representing color features of the normalized original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample.
Preferably, referring to fig. 3, in a second embodiment of the unmanned aerial vehicle image processing method based on gaussian denoising according to the present invention, the acquiring an on-site three-dimensional geographic sampling image of an unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by using the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, and acquiring a classification test result and an on-site three-dimensional geographic evaluation result of the three-dimensional geographic image sample of the unmanned aerial vehicle, further includes:
controlling a lens to capture a digital image, carrying out graying, drying removal and image size resetting on the acquired digital image, and removing noise by using Gaussian filtering and median filtering;
judging the defocusing degree of the image by using an image definition evaluation algorithm, finding a focus position by using a focus searching method, and driving the unmanned aerial vehicle lens to an optimal imaging position;
inputting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle into a wavelet neural network of a first layer by adopting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle, and respectively calculating the output of the hidden layer of the first layer and the output of the output layer by utilizing the initial weights of the hidden layer and the input layer nodes of the first layer network, the initial weights of the nodes of the output layer and the hidden layer, and the scaling factor and the shifting factor of the wavelet activation function;
inputting the output of the first network hidden layer and the output of the output layer into the support classifier SVC constructed by the characteristics of the double-phase forward training sample and the characteristics of all the double-phase training samples to obtain a final unmanned aerial vehicle image test result;
and analyzing the test result in the field of three-dimensional geography, clustering the field of three-dimensional geography, and acquiring texture characteristics and hardness characteristics of the field of three-dimensional geography.
Wherein the output of the output layer is calculated in a manner
h (i) represents the output of output node i, where h is a general representation of the output layer, and h is the output of the output layer of the first layer network1Indicating that h is output from the output layer of the second layer network2Denotes that p is the number of hidden nodes, W ″)ijRepresenting the weights between the output node i and the hidden node j, where W' is a general representation of the weights of the output node and the hidden node, and W is used as the weights of the output node and the hidden node of the first layer network1"indicates, the output layer node and hidden layer node weight W of the second layer network2"means, # j denotes the output of hidden node j, where # is a general representation of the hidden node output, and # is used for the first-layer network hidden node output1Indicating that the hidden node output of the second layer network is psi2And (4) showing.
Preferably, with reference to fig. 4, in a third embodiment of the unmanned aerial vehicle image processing method based on gaussian denoising according to the present invention, the performing wavelet detection analysis on the evaluation result of the real three-dimensional geography to obtain the dynamic detection result of the real three-dimensional geography further includes:
performing wavelet detection analysis on the content of the solid three-dimensional geographic evaluation result, initializing the content of the solid three-dimensional geographic evaluation result, and packaging into a calling data file;
sending the packaged calling data file to a wavelet detection analysis program, reading the calling data file, and performing wavelet packet decomposition and reconstruction to complete feature extraction;
sending out a reconstruction signal by adopting a principal component analysis method, establishing a covariance matrix after average preprocessing to obtain a characteristic value eigenvector, and taking data obtained after projection as a new data sample to reduce the dimension of original data;
and acquiring a dynamic detection result of the solid three-dimensional geography according to the data obtained after the projection, wherein the dynamic detection result of the solid three-dimensional geography shows the potential dynamic change of the solid three-dimensional geography.
Preferably, the performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography further includes:
acquiring a dynamic detection result of the on-site three-dimensional geography, and carrying out cluster analysis on the dynamic detection result of the on-site three-dimensional geography;
when the dynamic detection result of the solid three-dimensional geography belongs to a first cluster, performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography, detecting the odor concentration in the current environment, planning a path to track and position an odor source, and finally obtaining an image processing result of the solid three-dimensional geography;
and when the dynamic detection result of the on-site three-dimensional geography belongs to a second cluster, detecting the odor concentration in the current environment without detection.
Performing wavelet detection analysis on the content of the solid three-dimensional geographic evaluation result, and initializing the content of the solid three-dimensional geographic evaluation result comprises the following steps: collecting by an acoustic emission sensor, and sending an acoustic emission signal to an A/D converter to obtain a discrete time sequence;
performing frequency domain analysis on the redundancy signal by Fast Fourier Transform (FFT);
the frequency domain characteristics of the redundancy signals are obtained through calculation, the frequency domain characteristic values of the non-solid three-dimensional geographic activity redundancy material are large in whole and large in fluctuation range, and the characteristic values of the solid three-dimensional geographic activity redundancy material in the frequency domain are small in whole and small in fluctuation range. The characteristic value of the active redundancy signal is extracted in the frequency domain, so that the characteristic parameter for distinguishing the redundancy material is increased.
Sending the packaged calling data file to a wavelet detection analysis program for wavelet packet decomposition and reconstruction, wherein the step of completing feature extraction comprises the following steps:
and windowing the active redundancy signal to obtain an n-frame redundancy signal. And performing fast Fourier transform on each frame of the redundant signals and calculating an energy spectrum of each frame of the redundant signals. And then, filtering the energy spectrum by using a triangular band-pass filter, and simultaneously carrying out logarithmic operation and discrete cosine transform on the filtered redundant signals to finally obtain the MFCC coefficient. The above steps are repeated until the MFCC characteristic values are extracted from the redundancy signals of all the frames. When the MFCC feature extraction is carried out on the active redundancy signals, the number of the band-pass filters and the dimensionality of the active redundancy signals after discrete cosine transform influence the final result. When the number of the filters is large, the time for obtaining the MFCC characteristic value is prolonged, and the numerical value is also increased; the transformed dimension determines the refinement degree of the MFCC, and the larger the dimension is, the better the refinement degree is.
Specifically, the performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography further includes:
acquiring a dynamic detection result of the on-site three-dimensional geography, and carrying out cluster analysis on the dynamic detection result of the on-site three-dimensional geography;
when the dynamic detection result of the solid three-dimensional geography belongs to a first cluster, performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography, detecting the odor concentration in the current environment, planning a path to track and position an odor source, and finally obtaining an image processing result of the solid three-dimensional geography;
and when the dynamic detection result of the on-site three-dimensional geography belongs to a second cluster, detecting the odor concentration in the current environment without detection.
When the dynamic detection result of the solid three-dimensional geography belongs to a first cluster, performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography, detecting the odor concentration in the current environment and planning a path to track and position an odor source, and finally obtaining an image processing result of the solid three-dimensional geography, wherein the image processing method comprises the following steps: the first cluster represents that the dynamic detection result of the current on-site three-dimensional geography represents that the on-site three-dimensional geography has geological instability; in the flying process of the quad-rotor unmanned aerial vehicle, the sensor can continuously obtain a series of odor plume Concentration Line integrals CL (Concentration Line integrals) in different measurement angle directions by means of the rotation of the conductive slip ring, and an annular sequence of CLI measurement values is constructed; in each sampling period, the gas sensor obtains a CLI measured value, and then elements on the corresponding angle in the CLI annular sequence are updated; four rotor unmanned aerial vehicle are in the flight in-process, and the gaseous pollutants sensor is continuously through rotatory sampling and update CLI annular sequence. When the pollutant gas sensor detects the smell plume to be detected for the first time (the CLI measured value is greater than or equal to the set CLI threshold value), the rotor unmanned aerial vehicle is switched to the plume tracking stage from the plume finding stage.
At each control cycle, the CLI loop sequence is scanned for CLI segments not less than the CLI threshold. And if the number of CLI fragments is more than zero, switching from a smoke plume finding stage to a smoke plume tracking stage. The next direction of motion of the drone is determined by the following method. First, the angle corresponding to the center of each CLI segment, referred to herein as the CLI segment center angle, is calculated. And then finding the central angle closest to the upwind direction from the plurality of calculated CLI segment central angles to serve as the next motion direction of the unmanned aerial vehicle, thereby tracking the smell plume and gradually approaching the smell source.
Adopting RBF neural network algorithm to estimate environment wind vector, adopting RBF function as Gaussian kernel function, and expressing function in form of
xcExpressing the center of the Gaussian kernel function, and expressing the expansion constant or width parameter of the radial basis function by sigma, acting on the radial action range of the control function, with the smaller the sigma, the smaller the function width
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (5)
1. An unmanned aerial vehicle image processing method based on Gaussian denoising is characterized in that:
training a three-dimensional geographic image sample of the unmanned aerial vehicle to obtain a three-dimensional geographic image sample classifier of the unmanned aerial vehicle;
acquiring an on-site three-dimensional geographic sampling image of an unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by adopting the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, and acquiring a classification test result and an on-site three-dimensional geographic evaluation result of the three-dimensional geographic image sample of the unmanned aerial vehicle;
performing wavelet detection analysis on the evaluation result of the on-site three-dimensional geography to obtain a dynamic detection result of the on-site three-dimensional geography;
and carrying out environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography.
2. The unmanned aerial vehicle image processing method based on gaussian denoising of claim 1, wherein:
training the three-dimensional geographic image sample of the unmanned aerial vehicle to obtain the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, further comprising:
acquiring an original phase image of the three-dimensional geographic image sample of the training unmanned aerial vehicle, and extracting dynamic characteristics and color characteristics of the original phase image;
respectively carrying out normalization processing on the dynamic features and the color features of the original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample to obtain the dynamic features and the color features of the normalized original phase image;
fusing the dynamic characteristics and the color characteristics of the normalized original phase image to form dual-phase unmanned aerial vehicle image characteristics;
manually acquiring C double-phase positive training samples and D double-phase negative training samples from an original phase image of the training unmanned aerial vehicle three-dimensional geographic image sample, wherein 0< C <500, and 0< D < 1000;
expressing the C double-phase forward training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain the characteristics of the double-phase forward training samples;
representing all (C + D) double-phase training samples by using the obtained dynamic characteristics and color characteristics of the normalized original phase image to obtain all double-phase training sample characteristics;
and constructing a support classifier SVC by adopting the characteristics of the two-phase forward training sample and all the characteristics of the two-phase training sample.
3. The unmanned aerial vehicle image processing method based on gaussian denoising of claim 1, wherein: the method comprises the steps of obtaining an on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, preprocessing the on-site three-dimensional geographic sampling image of the unmanned aerial vehicle, testing the three-dimensional geographic image sample classifier of the unmanned aerial vehicle by adopting the three-dimensional geographic image sample classifier of the unmanned aerial vehicle, obtaining a classification test result of the three-dimensional geographic image sample of the unmanned aerial vehicle and an on-site three-dimensional geographic evaluation result, and further comprises the following steps:
controlling a lens to capture a digital image, carrying out graying, drying removal and image size resetting on the acquired digital image, and removing noise by using Gaussian filtering and median filtering;
judging the defocusing degree of the image by using an image definition evaluation algorithm, finding a focus position by using a focus searching method, and driving the unmanned aerial vehicle lens to an optimal imaging position;
inputting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle into a wavelet neural network of a first layer by adopting the image characteristics of the double-phase unmanned aerial vehicle and the forward training sample characteristics of the double-phase unmanned aerial vehicle, and respectively calculating the output of the hidden layer of the first layer and the output of the output layer by utilizing the initial weights of the hidden layer and the input layer nodes of the first layer network, the initial weights of the nodes of the output layer and the hidden layer, and the scaling factor and the shifting factor of the wavelet activation function;
inputting the output of the first network hidden layer and the output of the output layer into the support classifier SVC constructed by the characteristics of the double-phase forward training sample and the characteristics of all the double-phase training samples to obtain a final unmanned aerial vehicle image test result;
and analyzing the test result in the field of three-dimensional geography, clustering the field of three-dimensional geography, and acquiring texture characteristics and hardness characteristics of the field of three-dimensional geography.
4. The unmanned aerial vehicle image processing method based on gaussian denoising of claim 1, wherein:
the wavelet detection analysis is carried out on the evaluation result of the on-site three-dimensional geography to obtain the dynamic detection result of the on-site three-dimensional geography, and the method further comprises the following steps:
performing wavelet detection analysis on the content of the solid three-dimensional geographic evaluation result, initializing the content of the solid three-dimensional geographic evaluation result, and packaging into a calling data file;
sending the packaged calling data file to a wavelet detection analysis program, reading the calling data file, and performing wavelet packet decomposition and reconstruction to complete feature extraction;
sending out a reconstruction signal by adopting a principal component analysis method, establishing a covariance matrix after average preprocessing to obtain a characteristic value eigenvector, and taking data obtained after projection as a new data sample to reduce the dimension of original data;
and acquiring a dynamic detection result of the solid three-dimensional geography according to the data obtained after the projection, wherein the dynamic detection result of the solid three-dimensional geography shows the potential dynamic change of the solid three-dimensional geography.
5. The unmanned aerial vehicle image processing method based on gaussian denoising of claim 1, wherein: the performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography to obtain an image processing result of the solid three-dimensional geography further comprises:
acquiring a dynamic detection result of the on-site three-dimensional geography, and carrying out cluster analysis on the dynamic detection result of the on-site three-dimensional geography;
when the dynamic detection result of the solid three-dimensional geography belongs to a first cluster, performing environment detection analysis on the dynamic detection result of the solid three-dimensional geography, detecting the odor concentration in the current environment, planning a path to track and position an odor source, and finally obtaining an image processing result of the solid three-dimensional geography;
and when the dynamic detection result of the on-site three-dimensional geography belongs to a second cluster, detecting the odor concentration in the current environment without detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110755666.8A CN113343930A (en) | 2021-07-05 | 2021-07-05 | Unmanned aerial vehicle image processing method based on Gaussian denoising |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110755666.8A CN113343930A (en) | 2021-07-05 | 2021-07-05 | Unmanned aerial vehicle image processing method based on Gaussian denoising |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113343930A true CN113343930A (en) | 2021-09-03 |
Family
ID=77482487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110755666.8A Pending CN113343930A (en) | 2021-07-05 | 2021-07-05 | Unmanned aerial vehicle image processing method based on Gaussian denoising |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113343930A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630828A (en) * | 2023-05-30 | 2023-08-22 | 中国公路工程咨询集团有限公司 | Unmanned aerial vehicle remote sensing information acquisition system and method based on terrain environment adaptation |
-
2021
- 2021-07-05 CN CN202110755666.8A patent/CN113343930A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630828A (en) * | 2023-05-30 | 2023-08-22 | 中国公路工程咨询集团有限公司 | Unmanned aerial vehicle remote sensing information acquisition system and method based on terrain environment adaptation |
| CN116630828B (en) * | 2023-05-30 | 2023-11-24 | 中国公路工程咨询集团有限公司 | Unmanned aerial vehicle remote sensing information acquisition system and method based on terrain environment adaptation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107862705B (en) | Unmanned aerial vehicle small target detection method based on motion characteristics and deep learning characteristics | |
| CN106356757B (en) | A kind of power circuit unmanned plane method for inspecting based on human-eye visual characteristic | |
| CN108460341B (en) | Optical remote sensing image target detection method based on integrated depth convolution network | |
| CN109598241B (en) | Satellite image marine ship identification method based on Faster R-CNN | |
| Rudol et al. | Human body detection and geolocalization for UAV search and rescue missions using color and thermal imagery | |
| CN110207671B (en) | Space-based intelligent imaging system | |
| CN108537122A (en) | Include the image co-registration acquisition system and image storage method of meteorologic parameter | |
| CN114612835A (en) | Unmanned aerial vehicle target detection model based on YOLOv5 network | |
| CN112445241A (en) | Ground surface vegetation identification method and system based on unmanned aerial vehicle remote sensing technology and readable storage medium | |
| CN107527037A (en) | Blue-green algae identification and analysis system based on unmanned aerial vehicle remote sensing data | |
| CN115457428A (en) | Improved YOLOv5 fire detection method and device integrating adjustable coordinate residual attention | |
| CN112906795A (en) | Whistle vehicle judgment method based on convolutional neural network | |
| Pölönen et al. | Tree species identification using 3D spectral data and 3D convolutional neural network | |
| US9892340B2 (en) | Method for classifying objects in an imaging surveillance system | |
| Aissou et al. | Building roof superstructures classification from imbalanced and low density airborne LiDAR point cloud | |
| CN113343930A (en) | Unmanned aerial vehicle image processing method based on Gaussian denoising | |
| US11727665B2 (en) | Unmanned aircraft system (UAS) detection and assessment via temporal intensity aliasing | |
| CN113553912A (en) | Unmanned aerial vehicle image processing method based on RGB distribution | |
| CN113343929A (en) | Unmanned aerial vehicle image processing method based on wavelet transformation | |
| Comert et al. | Rapid mapping of forested landslide from ultra-high resolution unmanned aerial vehicle data | |
| CN111950386A (en) | Functional intelligence-based environment self-adaptive navigation scene recognition method for micro unmanned aerial vehicle | |
| Bazzarello et al. | Anomaly detection in Sonar images: application of saliency filters | |
| CN113486753A (en) | Unmanned aerial vehicle image processing method based on SAR | |
| CN113569649A (en) | Unmanned aerial vehicle image processing method based on fractal recognition | |
| Cirneanu et al. | CNN based on LBP for evaluating natural disasters |
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 |