CN114879485A - Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network - Google Patents
Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network Download PDFInfo
- Publication number
- CN114879485A CN114879485A CN202210579995.6A CN202210579995A CN114879485A CN 114879485 A CN114879485 A CN 114879485A CN 202210579995 A CN202210579995 A CN 202210579995A CN 114879485 A CN114879485 A CN 114879485A
- Authority
- CN
- China
- Prior art keywords
- layer
- fuzzy
- aerial vehicle
- unmanned aerial
- neural network
- 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
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0205—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system
- G05B13/024—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Spray Control Apparatus (AREA)
Abstract
The invention discloses an unmanned aerial vehicle spraying quality monitoring system and a method thereof based on a RBF neural network, and the system comprises coating monitoring equipment which is additionally arranged at the front end of the unmanned aerial vehicle and comprises a thickness gauge and an industrial camera, a multi-degree-of-freedom parallel nozzle and a coating quality monitoring control unit; sampling the thickness of the coating by a thickness gauge, identifying images by an industrial camera, performing integrated processing on data of the thickness gauge and the image, establishing a membership function about the thickness, the proportion of the spraying area and the uniformity, analyzing the coating information in real time by adopting a RBF (radial basis function) fuzzy neural network, and timely adjusting the output state of spraying equipment. The invention can effectively improve the real-time monitoring and adjustment of the work of the spraying unmanned aerial vehicle, has high intelligent degree and strong stability, and is suitable for application and popularization.
Description
Technical Field
The invention belongs to the field of intelligent manufacturing, and relates to an unmanned aerial vehicle spraying quality monitoring system and method based on a Radial Basis Function (RBF) fuzzy neural network.
Background
The study of artificial neural networks began in the 40's of the 20 th century and belonged to a cross discipline. The neural network modes commonly used in general are mainly feedforward type, feedback type, self-organizing type and random type. In recent years, a relatively novel neural network has been formed by combining methods such as fuzzy control, adaptive control, and predictive control. The feed-forward network mainly comprises a BP network and an RBF network. However, the BP network has local minima and a slow convergence rate, and the RBF network can overcome the problems to some extent, and is therefore more emphasized.
Radial Basis Function (RBF) plays an important role in the field of neural networks, for example, the RBF neural network has the characteristic of unique optimal approximation, and the radial basis is used as a kernel function to map input samples to a high-dimensional feature space in an SVM (support vector machine), so that the problems of inseparability of original linearity are solved, and a plurality of new applications are derived around the technology.
In recent years, dynamic coating quality detection becomes a technical problem in the coating field, and particularly, a method for detecting the coating quality of an unmanned aerial vehicle has a great technical problem, and most of the existing detection methods and equipment are difficult to meet the actual requirements, and have the problems of poor applicability, unstable working state and the like.
Chinese patent application CN201911383678.1 discloses a metal surface painting quality detection apparatus. The device simple structure, the simple operation through mechanical transmission, accomplishes the operation to the crossing mar of the lacquer painting that has dried through automatically, and accomplishes the soft brush of mar and brush the operation, the inspection personnel's of being convenient for observation automatically after accomplishing the mar. But ultimately the quality of the coating is judged by the experience of the inspector, the efficiency is still low, and the method is not suitable for large objects.
Chinese patent CN202010283394.1 discloses a paint surface thickness detection device. The device determines a target part to be measured and a paint film instrument gesture corresponding to the target part; determining the displacement variation of the varnish film instrument between the current measurement time and the previous measurement time of the current measurement time; wherein, a first position of the paint film instrument at the current measuring moment and a second position of the paint film instrument at the previous measuring moment are both positioned on the target part; and if the current posture of the paint film instrument is matched with the posture of the paint film instrument corresponding to the target part and the displacement variation is larger than a preset displacement threshold, measuring the paint surface thickness at the first position on the target part to obtain a paint surface measuring result corresponding to the current measuring time. However, for the unmanned aerial vehicle, the load of the unmanned aerial vehicle has more unstable factors, and high-precision measurement through displacement variation is difficult.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides an unmanned aerial vehicle spraying quality monitoring system and method based on an RBF neural network, which can monitor the spraying quality of a coating under different illumination conditions and can control the pressure of a spray head in real time based on the quality of a layer.
In order to solve the technical problems, the invention adopts the following technical scheme.
The invention discloses an unmanned aerial vehicle spraying quality monitoring system based on a RBF neural network, which comprises a multi-rotor vector unmanned aerial vehicle, coating monitoring equipment, a multi-degree-of-freedom parallel nozzle and a coating quality monitoring control unit, wherein the coating monitoring equipment, the multi-degree-of-freedom parallel nozzle and the coating quality monitoring control unit are arranged on the unmanned aerial vehicle;
the coating monitoring equipment is arranged at the front end of the multi-rotor vector unmanned aerial vehicle and comprises a thickness gauge and an industrial camera; the multi-degree-of-freedom parallel nozzle can perform multi-degree-of-freedom rotary spraying, is arranged at the front end of the unmanned aerial vehicle through a connecting rod, and comprises a nozzle, an axial rotating part, a connecting rod and a steering engine;
the thickness gauge samples the thickness of the coating, utilizes an industrial camera to recognize images, integrates and processes data of the thickness gauge and the industrial camera, establishes a membership function related to thickness, spraying area ratio and uniformity, and a coating quality monitoring control unit analyzes coating information in real time by adopting a RBF fuzzy neural network and adjusts the output state of spraying equipment.
Preferably, the multi-degree-of-freedom parallel nozzle is arranged at the rear end of the multi-rotor vector unmanned aerial vehicle of 300-500mm, the axial rotating part of the multi-degree-of-freedom parallel nozzle is connected to the front end of a spray rod and can be controlled to rotate by a motor carried by the unmanned aerial vehicle, and the steering engine control connecting rod drives a plurality of groups of nozzle devices to perform large-angle direction adjustment so as to meet the spraying requirement.
Furthermore, a light intensity sensor is arranged in the industrial camera and used for monitoring the quality of the coating in real time; the monitoring system can calculate the difference value between the actual brightness amplitude and the predicted brightness reference value according to the brightness amplitude of the work area acquired by the light intensity sensor, and controls the voltage regulation mode of the diffuse light lamp through comparison of the brightness difference value to regulate the brightness of the diffuse light lamp.
The invention discloses an unmanned aerial vehicle spraying quality monitoring method based on a RBF neural network, which comprises the following steps:
step one, constructing a brightness adjusting device of a diffusion lamp: the method comprises the steps of collecting the brightness amplitude of a working area according to a light intensity sensor, calculating the difference value between the actual brightness amplitude and a predicted brightness reference value, and controlling the voltage regulation mode of the diffuse light lamp through comparison of the brightness difference value so as to regulate the brightness of the diffuse light lamp; and if the brightness amplitude data acquired by the light intensity sensor is not less than the preset brightness difference threshold, controlling the voltage regulation mode of the diffuse light lamp to regulate the brightness of the diffuse light lamp.
Step two, data sampling and image processing: acquiring coating picture data through an industrial camera, preprocessing the picture layer picture data, and further obtaining gray mean square error, gray mean value and gray mean value change rate image data of the image in the monitoring area;
step three, building an RBF fuzzy neural network: the fuzzy inference system comprises an input layer, a fuzzy inference layer and an output layer, wherein an action target function Gaussian basis function is constructed to form a local approximation neural network;
step four, carrying out sample training and optimizing step by step: collecting samples and dividing the samples, and approaching an object by using an RBF fuzzy network; the quality monitoring capability is trained by utilizing the samples of the training set, so that the precision and the speed of real-time monitoring and adjustment of the spraying work of the unmanned aerial vehicle are improved.
Specifically, in the second step, the coating image data collected by the industrial camera is sampled, quantized and encoded by using Python, and the processed data is grayed to obtain the grayscale mean square error x of the image in the spraying area 1 Mean value of gray x 2 Mean rate of change of gray scale x 3 。
Specifically, in the third step, the RBF fuzzy neural network is built to include an input layer, a deblurring layer, a fuzzy inference layer, and an output layer, which are respectively:
first layer, input layer:
the nodes of the input layer are directly linked with the input parameters, the input parameters are transmitted to the input layer, and the relation between the input and the output of the transmission nodes is as follows:
f(i)＝x i (1)
wherein x is i Is the ith input value;
second, fuzzy layer, i.e. membership function layer:
adopting a Gaussian function as a membership function, in a fuzzy layer, each node can carry out membership function calculation, and at the jth node:
wherein, c ij And c ij Respectively the mean value and standard deviation of the ith input variable and the jth fuzzy set Gaussian function;
the third layer and the fuzzy inference layer are the establishment rule layer:
each node of the fuzzy inference layer is connected with the fuzzy layer, and corresponding data are output by combining fixed fuzzy control rules, wherein the output result of the node j is the total product of all input signals of the node, namely:
wherein the content of the first and second substances,
N i the number of the ith input membership function in the input layer, namely the number of the fuzzification layer nodes;
fourth layer, output layer:
the weighted sum of the input signals received by each node is the output f of the node n The amount, namely:
wherein l is the number of output layer nodes, and W is the connection weight matrix of the output nodes and each node of the third layer.
Specifically, the fourth step specifically includes the following steps:
adopting an RBF fuzzy network object approximation method:
with fuzzy inputs y (k) and u (k) and fuzzy output y m (k)，
Taking a network structure as an input layer 2, a fuzzy layer 4 and an output layer 1, and taking y m (k)＝f 4 (ii) a The approximation error of the blur is:
e(k)＝y(k)-y m (k) (5)
wherein e (k) represents an approximation error, y m (k) Representing fuzzy output, y (k) representing actual output；
Weight learning algorithm of the output layer:
w(k)＝w(k-1)+Δw(k)+α(w(k-1)-w(k-2)) (6)
wherein k is 1,2,3 …, N is the number of samples, α is a momentum factor, α ∈ [0,1], w (k) represents the weight of the kth sample, and Δ w (k) represents the rate of change of the weight;
the adjustment mode of the output weight value is as follows:
wherein eta is a learning rate, and eta belongs to [0,1 ];
the adjustment mode of the membership function parameters is as follows:
wherein the content of the first and second substances,
Δc ij rate of change of center value, Δ b, of fuzzy subsets i Is the rate of change of width, net, of a Gaussian function j Is the error of the j-th layer unit and the upper layer unit, eta is the learning rate, eta belongs to [0,1]]；
Learning algorithm of membership function parameters:
c ij (k)＝c ij (k-1)+Δc ij (k)+α(c ij (k-1)-c ij (k-2)) (10)
b j (k)＝b j (k-1)+Δb j (k)+α(b j (k-1)-b j (k-2)) (11)
wherein: c. C ij Central value of fuzzy subset, b j Is the width of the Gaussian function, alpha is the momentum factor, alpha belongs to [0,1]]。
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the multi-degree-of-freedom parallel nozzle is arranged at the front end of the unmanned aerial vehicle, so that the effective spraying area can be remarkably increased during spraying operation, and the coating efficiency is improved; meanwhile, multi-degree-of-freedom synchronous rotation of the nozzle in a micro-angle is realized through the steering engine connecting rod mechanism, and the spraying point is adjusted in real time under the monitoring control of the coating, so that the coating quality is improved.
2. The invention can control the output of the coating by controlling the pressure of the multi-degree-of-freedom parallel nozzle through the control system based on the RBF fuzzy neural network, thereby achieving the effects of reducing the thickness difference of the coating and saving the coating.
3. The invention is constructed with a diffused light brightness adjusting device, collects the brightness amplitude of a working area by using a light intensity sensor of a coating monitoring device arranged at the front end of an unmanned aerial vehicle, calculates the difference value between the actual brightness amplitude and the predicted brightness reference value, controls the voltage adjusting mode of the diffused light lamp through the brightness difference value comparison, adjusts the brightness of the diffused light lamp, and is beneficial to monitoring the coating condition in real time.
4. The invention adopts the RBF fuzzy neural network to analyze the coating information in real time and adjust the output state of the spraying mechanism, thereby realizing the full-automatic monitoring work of spraying, having high intelligent degree and strong stability, and needing no frequent manual operation; simple use and is suitable for application and popularization.
Drawings
Fig. 1 is a schematic structural diagram of a painting drone according to an embodiment of the present invention.
Fig. 2 is a schematic structural diagram of a parallel rotary showerhead according to an embodiment of the present invention.
FIG. 3 is a method flow diagram of one embodiment of the present invention.
Fig. 4 is a schematic structural diagram of an RBF fuzzy neural network according to an embodiment of the present invention.
Fig. 5 is a schematic diagram of an RBF fuzzy network object approximation method according to an embodiment of the present invention.
The system comprises a 100-rotor-wing vector unmanned aerial vehicle, a 200-coating monitoring device and 300-degree-of-freedom parallel spray heads, wherein the spray heads are connected in parallel; 301 shower nozzle, axial rotating part 302, connecting rod 303, steering wheel 304, coupling rod 305.
Detailed Description
The invention relates to an unmanned aerial vehicle spraying quality monitoring system based on a RBF neural network and a method thereof, wherein the system comprises the following steps: many rotor vector unmanned aerial vehicle, set up coating monitoring facilities and the parallelly connected shower nozzle of multi freedom, coating quality monitoring control unit on unmanned aerial vehicle. The coating thickness is sampled by a thickness gauge additionally arranged at the front end of the unmanned aerial vehicle, an industrial camera is utilized for image recognition, data are subjected to integrated processing, a membership function about thickness, spraying area ratio and uniformity is established, a coating quality monitoring control unit adopts a RBF fuzzy neural network to analyze coating information in real time, and the output state of spraying equipment is adjusted.
As shown in fig. 1, a schematic structural diagram of a spraying drone according to an embodiment of the present invention includes a multi-rotor vector drone 100, a coating monitoring device 200, and a multiple-degree-of-freedom parallel nozzle 300. Wherein many rotor vector unmanned aerial vehicle can carry on spraying monitoring facilities and carry out spraying work, and coating monitoring facilities sets up the front end at many rotor vector unmanned aerial vehicle, including calibrator and industrial camera, sets up light intensity sensor in the industrial camera and be used for real-time supervision coating quality, and the parallelly connected shower nozzle of multi freedom sets up in many rotor vector unmanned aerial vehicle front end 300 and supplyes 500mm department, can carry out multi freedom and revolve to the spraying.
As shown in fig. 2, the multiple-degree-of-freedom parallel nozzle is arranged at the front end of the unmanned aerial vehicle through a coupling rod, and comprises a nozzle 301, an axial rotating component 302, a connecting rod 303, a steering engine 304 and a coupling rod 305. Wherein axial rotating part connects at the spray lance front end, can carry on motor control by unmanned aerial vehicle and rotate, and steering wheel control connecting rod drives three group shower nozzle devices and carries out the direction adjustment of wide angle for satisfy spraying control requirement.
As shown in fig. 3, the method for monitoring the spraying quality of the unmanned aerial vehicle based on the RBF neural network of the present invention includes the following steps:
step one, constructing a brightness adjusting device of a diffusion lamp: and calculating the difference value between the actual brightness amplitude value and the predicted brightness reference value according to the brightness amplitude value of the working area acquired by the light intensity sensor, and controlling the voltage regulation mode of the diffuse light lamp through comparing the brightness difference values so as to regulate the brightness of the diffuse light lamp.
Step two, data sampling and image processing: coating picture data are collected through an industrial camera, preprocessing is carried out on the picture layer picture data, and image data such as the gray mean square error, the gray mean value change rate and the like of the monitored area image are further obtained.
Step three, building an RBF fuzzy neural network: the method comprises an input layer, a fuzzy inference layer and an output layer, wherein a function target function Gaussian basis function is constructed to form a local approximation neural network, and the precision, robustness and adaptability of the system are further improved.
Step four, carrying out sample training and optimizing step by step: firstly, collecting samples and dividing the samples, approaching an object by using an RBF fuzzy network, and secondly, training the quality monitoring capability by using the samples of a training set, thereby improving the precision and speed of real-time monitoring and adjustment of the spraying work of the unmanned aerial vehicle.
The first step specifically comprises:
and acquiring brightness amplitude data by a light intensity sensor, calculating the difference value between the actual brightness amplitude and the predicted brightness reference value, and if the difference value is not less than a preset brightness difference value threshold value, controlling the voltage regulation mode of the diffuse light lamp to regulate the brightness of the diffuse light lamp, thereby enhancing the reliability of real-time monitoring of the coating.
The second step specifically comprises:
firstly, the image data of a working area is collected by an industrial camera, secondly, the image data is sampled, quantized and coded by Python, and finally, the processed data is subjected to graying processing to obtain the gray mean square error, the gray mean value and the change rate of the gray mean value of the image of the spraying area, so that the monitoring function is realized.
And step three, the established RBF fuzzy neural network comprises an input layer, a deblurring layer, a fuzzy inference layer and an output layer. As shown in fig. 4:
wherein x is 1 Is the mean square error of the gray scale of the image, x 2 Is the mean value of the gray levels of the image, x 3 The image gray level mean change rate;
a first layer: input layer
The nodes of the input layer directly establish the relation with the input parameters and transmit the input quantity to the input layer. The relationship between the input and output of the transmission node is:
f(i)＝x i (1)
wherein: x is the number of i Is the ith input value.
A second layer: layers of fuzzy, i.e. membership function layers
Adopting a Gaussian function as a membership function, in a fuzzy layer, each node can carry out membership function calculation, and at the jth node:
wherein: c. C ij And c ij The mean and standard deviation of the ith input variable and the jth fuzzy set gaussian function, respectively.
And a third layer: fuzzy inference layers, i.e. layers of rules
Each node of the fuzzy inference layer is connected with the fuzzy layer, and corresponding data are output by combining fixed fuzzy control rules, wherein the output result of the node j is the total product of all input signals of the node, namely:
wherein:
N i the number of the ith input membership function in the input layer, namely the number of the fuzzification layer nodes.
A fourth layer: output layer
The weighted sum of the input signals received by each node is the output f of the node n The amount, namely:
wherein: l is the number of output layer nodes, and W is the connection weight matrix of the output nodes and each node of the third layer.
As shown above, the establishment of the RBF fuzzy neural network structure is completed.
The fourth step specifically comprises the following steps:
the RBF fuzzy network object approximation method shown in FIG. 5 is adopted.
Wherein: the fuzzy inputs are y (k) and u (k), and the fuzzy output is y m (k)。
Taking a network structure as an input layer 2, an ambiguity layer 4 and an output layer 1, and taking y m (k)＝f 4 . The approximation error of the blur is:
e(k)＝y(k)-y m (k) (5)
wherein: e (k) represents the approximation error, y m (k) Representing the fuzzy output and y (k) the actual output.
Weight learning algorithm of the output layer:
w(k)＝w(k-1)+Δw(k)+α(w(k-1)-w(k-2)) (6)
where k is 1,2,3 …, N is the number of samples, α is a momentum factor, α ∈ [0,1], w (k) represents the weight of the kth sample, and Δ w (k) represents the rate of change in the weight.
The adjustment mode of the output weight value is as follows:
wherein eta is the learning rate, and eta belongs to [0,1 ].
The adjustment mode of the membership function parameters is as follows:
wherein:
Δc ij rate of change of center value, Δ b, of fuzzy subsets i Is the rate of change of width, net, of a Gaussian function j Is the error of the j-th layer unit and the upper layer unit, eta is the learning rate, eta belongs to [0,1]]。
Learning algorithm of membership function parameters:
c ij (k)＝c ij (k-1)+Δc ij (k)+α(c ij (k-1)-c ij (k-2)) (10)
b j (k)＝b j (k-1)+Δb j (k)+α(b j (k-1)-b j (k-2)) (11)
wherein: c. C ij Central value of fuzzy subset, b j Is the width of the Gaussian function, alpha is the momentum factor, alpha belongs to [0,1]]。
Therefore, the approximation training of the RBF fuzzy neural network is completed. The RBF fuzzy neural network approaches the training of the object, so that the monitoring accuracy and speed are continuously improved, and the real-time monitoring and control can be completed.
As mentioned above, the method of the invention adopts RBF fuzzy neural network control, firstly the industrial camera collects the coating picture and processes the data as the input value of the RBF fuzzy neural network; then, establishing a membership function based on a Gaussian function as a standard for monitoring the quality of the coating; secondly, in the fuzzy layer, each node can carry out membership function calculation; finally, combining through a fixed fuzzy control rule to output corresponding data, and taking the weighted sum of the input signals received by each node as the output quantity f of the node n The output quantity of the pressure valve is used for achieving the purpose of controlling the output of the spray head.
Claims (8)
1. An unmanned aerial vehicle spraying quality monitoring system based on a RBF neural network is characterized by comprising a multi-rotor vector unmanned aerial vehicle, coating monitoring equipment arranged on the unmanned aerial vehicle, a multi-degree-of-freedom parallel nozzle and a coating quality monitoring control unit;
the coating monitoring equipment is arranged at the front end of the multi-rotor vector unmanned aerial vehicle and comprises a thickness gauge and an industrial camera; the multi-degree-of-freedom parallel spray head can perform multi-degree-of-freedom rotary spraying, is arranged at the front end of the unmanned aerial vehicle through a connecting rod, and comprises a spray head (301), an axial rotating part (302), a connecting rod (303), a steering engine (304) and a connecting rod (305);
the coating monitoring equipment samples the thickness of the coating through a thickness gauge, utilizes an industrial camera to recognize images, integrates and processes data of the coating monitoring equipment, establishes membership functions related to thickness, spraying area ratio and uniformity, and adopts a RBF fuzzy neural network to analyze coating information in real time by a coating quality monitoring control unit so as to adjust the output state of the spraying equipment.
2. The unmanned aerial vehicle spraying quality monitoring system based on the RBF neural network as claimed in claim 1, wherein the multiple-degree-of-freedom parallel nozzle is arranged at the front end of the multi-rotor vector unmanned aerial vehicle by 300-mm distance, the axial rotating part of the multi-degree-of-freedom parallel nozzle is connected to the front end of a spray rod and can be controlled to rotate by a motor carried by the unmanned aerial vehicle, and the steering engine control connecting rod drives a plurality of groups of nozzle devices to perform large-angle direction adjustment so as to meet the spraying requirement.
3. The unmanned aerial vehicle spraying quality monitoring system based on the RBF neural network as claimed in claim 1, wherein a light intensity sensor is arranged in the industrial camera for monitoring the coating quality in real time; the monitoring system can calculate the difference value between the actual brightness amplitude and the predicted brightness reference value according to the brightness amplitude of the work area acquired by the light intensity sensor, and controls the voltage regulation mode of the diffuse light lamp through comparison of the brightness difference value to regulate the brightness of the diffuse light lamp.
4. An unmanned aerial vehicle spraying quality monitoring method based on a RBF neural network is characterized by comprising the following steps:
step one, constructing a brightness adjusting device of a diffusion lamp: the method comprises the steps of collecting the brightness amplitude of a working area according to a light intensity sensor, calculating the difference value between the actual brightness amplitude and a predicted brightness reference value, and controlling the voltage regulation mode of the diffuse light lamp through comparison of the brightness difference value so as to regulate the brightness of the diffuse light lamp;
step two, data sampling and image processing: acquiring coating picture data through an industrial camera, preprocessing the picture layer picture data, and further obtaining gray mean square error, gray mean value and gray mean value change rate image data of the image in the monitoring area;
step three, building an RBF fuzzy neural network: the fuzzy inference system comprises an input layer, a fuzzy inference layer and an output layer, wherein an action target function Gaussian basis function is constructed to form a local approximation neural network;
step four, carrying out sample training and optimizing step by step: collecting samples and dividing the samples, and approaching an object by using an RBF fuzzy network; the quality monitoring capability is trained by utilizing the samples of the training set, so that the precision and the speed of real-time monitoring and adjustment of the spraying work of the unmanned aerial vehicle are improved.
5. The unmanned aerial vehicle spraying quality monitoring method based on the RBF neural network as claimed in claim 4, wherein in the first step, if the brightness amplitude data obtained by the light intensity sensor is not less than the preset brightness difference threshold, the voltage regulation mode of the diffuse light lamp is controlled to regulate the brightness of the diffuse light lamp.
6. The unmanned aerial vehicle spraying quality monitoring method based on the RBF neural network as claimed in claim 4, wherein in the second step, the coating picture data collected by the industrial camera is sampled, quantized and encoded by Python, and finally the processed data is grayed to obtain the grayscale mean square error x of the spraying area image 1 Mean value of gray x 2 Mean rate of change of gray scale x 3 。
7. The unmanned aerial vehicle spraying quality monitoring method based on the RBF neural network as claimed in claim 4, wherein in the third step, the RBF fuzzy neural network is built to comprise an input layer, a deblurring layer, a fuzzy inference layer and an output layer, which are respectively:
first layer, input layer:
the node of the input layer directly establishes the relation with the input parameter quantity, and transmits the input quantity to the input layer, and the relation between the input and the output of the transmission node is as follows:
f(i)＝x i (1)
wherein x is i Is the ith input value, i is 1,2, 3;
second, fuzzy layer, i.e. membership function layer:
adopting a Gaussian function as a membership function, in a fuzzy layer, each node can carry out membership function calculation, and at the jth node:
wherein, c ij And c ij Respectively the mean value and standard deviation of the ith input variable and the jth fuzzy set Gaussian function;
the third layer and the fuzzy inference layer are the establishment rule layer:
each node of the fuzzy inference layer is connected with the fuzzy layer, and corresponding data are output by combining fixed fuzzy control rules, wherein the output result of the node j is the total product of all input signals of the node, namely:
wherein the content of the first and second substances,
N i the number of the ith input membership function in the input layer, namely the number of the fuzzification layer nodes;
fourth layer, output layer:
the weighted sum of the input signals received by each node is the output f of the node n The amount, namely:
wherein l is the number of nodes in the output layer, and W is the connection weight matrix of the output node and each node in the third layer.
8. The unmanned aerial vehicle spraying quality monitoring method based on the RBF neural network as claimed in claim 4, wherein the fourth step specifically comprises the following steps:
adopting an RBF fuzzy network object approximation method:
with fuzzy inputs y (k) and u (k) and fuzzy output y m (k)，
Taking a network structure as an input layer 2, a fuzzy layer 4 and an output layer 1, and taking y m (k)＝f 4 (ii) a The approximation error of the blur is:
e(k)＝y(k)-y m (k) (5)
wherein e (k) represents an approximation error, y m (k) Representing the fuzzy output, y (k) representing the actual output;
weight learning algorithm of the output layer:
w(k)＝w(k-1)+Δw(k)+α(w(k-1)-w(k-2)) (6)
wherein k is 1,2,3 …, N is the number of samples, α is a momentum factor, α ∈ [0,1], w (k) represents the weight of the kth sample, and Δ w (k) represents the rate of change of the weight;
the adjustment mode of the output weight value is as follows:
wherein eta is a learning rate, and eta belongs to [0,1 ];
the adjustment mode of the membership function parameters is as follows:
wherein the content of the first and second substances,
Δc ij rate of change of center value, Δ b, of fuzzy subsets i Is the rate of change of width, net, of a Gaussian function j Is the error of the j-th layer unit and the upper layer unit, eta is the learning rate, eta belongs to [0,1]]；
Learning algorithm of membership function parameters:
c ij (k)＝c ij (k-1)+Δc ij (k)+α(c ij (k-1)-c ij (k-2)) (10)
b j (k)＝b j (k-1)+Δb j (k)+α(b j (k-1)-b j (k-2)) (11)
wherein: c. C ij Central value of fuzzy subset, b j Is the width of the Gaussian function, alpha is the momentum factor, alpha belongs to [0,1]]。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210579995.6A CN114879485A (en) | 2022-05-25 | 2022-05-25 | Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210579995.6A CN114879485A (en) | 2022-05-25 | 2022-05-25 | Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114879485A true CN114879485A (en) | 2022-08-09 |
Family
ID=82677778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210579995.6A Pending CN114879485A (en) | 2022-05-25 | 2022-05-25 | Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114879485A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116029589A (en) * | 2022-12-14 | 2023-04-28 | 浙江问源环保科技股份有限公司 | Rural domestic sewage animal and vegetable oil online monitoring method based on two-section RBF |
| CN116045791A (en) * | 2023-04-03 | 2023-05-02 | 成都飞机工业（集团）有限责任公司 | Metal paint coating thickness assessment method |
-
2022
- 2022-05-25 CN CN202210579995.6A patent/CN114879485A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116029589A (en) * | 2022-12-14 | 2023-04-28 | 浙江问源环保科技股份有限公司 | Rural domestic sewage animal and vegetable oil online monitoring method based on two-section RBF |
| CN116029589B (en) * | 2022-12-14 | 2023-08-22 | 浙江问源环保科技股份有限公司 | Rural domestic sewage animal and vegetable oil online monitoring method based on two-section RBF |
| CN116045791A (en) * | 2023-04-03 | 2023-05-02 | 成都飞机工业（集团）有限责任公司 | Metal paint coating thickness assessment method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114879485A (en) | Unmanned aerial vehicle spraying quality monitoring system and method based on RBF neural network | |
| CN109375235B (en) | Inland ship freeboard detection method based on deep reinforcement neural network | |
| CN108898215B (en) | Intelligent sludge bulking identification method based on two-type fuzzy neural network | |
| CN111047012A (en) | Air quality prediction method based on deep bidirectional long-short term memory network | |
| CN107255923B (en) | RBF identification-based under-actuated unmanned ship track tracking control method of ICA-CMAC neural network | |
| CN111176115B (en) | Valve position control method based on fuzzy neural network and humanoid intelligent control | |
| CN111931411B (en) | Duhem dynamic hysteresis modeling method for piezoelectric driving micro-positioning platform | |
| CN111982302A (en) | Temperature measurement method with noise filtering and environment temperature compensation | |
| CN108227715B (en) | Wave-resistant energy-saving unmanned ship path tracking method | |
| CN116382071B (en) | Pneumatic parameter intelligent identification method for deep learning network correction compensation | |
| CN111608868B (en) | Maximum power tracking adaptive robust control system and method for wind power generation system | |
| CN108717262B (en) | Special-shaped curved surface tracking method and system based on moment feature learning neural network | |
| CN112182972A (en) | ADAM local weighted regression identification modeling method for ship maneuvering motion | |
| CN101122777A (en) | Large condenser underwater operation environment two-joint robot control method | |
| CN109600083B (en) | Two-degree-of-freedom bearingless permanent magnet synchronous motor suspension force subsystem decoupling controller | |
| CN114510864A (en) | Forecasting method for forecasting rolling force by neural network based on K-means clustering algorithm | |
| CN111538232B (en) | Unmanned driving anti-rolling positioning method and system based on self-adaptive neural fuzzy control | |
| CN116702320A (en) | Unmanned ship response model parameter identification method based on improved particle swarm algorithm | |
| CN116880201A (en) | Water network channel state control system based on fuzzy neural network | |
| CN108446506B (en) | Uncertain system modeling method based on interval feedback neural network | |
| CN116565877A (en) | Automatic voltage partition control method based on spectral cluster analysis | |
| CN111222525A (en) | Gluing quality database construction method based on improved extreme learning machine | |
| Zhang et al. | Multi-dimensional local weighted regression ship motion identification modeling based on particle swarm optimization | |
| CN116205914B (en) | Waterproof coating production intelligent monitoring system | |
| Blake et al. | Comparison between Regressive and Classifying Neural Networks for PID Controlled Path-Following |
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 |