CN114897918B - Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system - Google Patents
Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system Download PDFInfo
- Publication number
- CN114897918B CN114897918B CN202210818928.5A CN202210818928A CN114897918B CN 114897918 B CN114897918 B CN 114897918B CN 202210818928 A CN202210818928 A CN 202210818928A CN 114897918 B CN114897918 B CN 114897918B
- Authority
- CN
- China
- Prior art keywords
- dirty
- area
- robot
- edge
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004140 cleaning Methods 0.000 title claims abstract description 94
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims description 47
- 238000005192 partition Methods 0.000 claims description 32
- 150000001875 compounds Chemical class 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 14
- 230000011218 segmentation Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 4
- 238000011109 contamination Methods 0.000 claims description 3
- 239000002689 soil Substances 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 238000003064 k means clustering Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 description 8
- 230000002776 aggregation Effects 0.000 description 6
- 238000004220 aggregation Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000001276 controlling effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S40/00—Components or accessories in combination with PV modules, not provided for in groups H02S10/00 - H02S30/00
- H02S40/10—Cleaning arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
-
- 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
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Electric Vacuum Cleaner (AREA)
Abstract
The invention relates to a photovoltaic cleaning robot brush power adjustment method and system based on artificial intelligence.
Description
Technical Field
The application relates to the field of photovoltaic cleaning and artificial intelligence, in particular to a method and a system for adjusting brush power of a photovoltaic cleaning robot based on artificial intelligence.
Background
The battery pack is under the sunshine, and the part programming rate that is covered by the dirt of dust is far greater than not covered the part, and the dark spot that burns out will appear in the high temperature, and the photovoltaic cell who is covered can become the resistance that does not generate electricity, and the electric power that the consumption battery produced generates heat, leads to the ageing aggravation of panel, and the generated energy reduces, can cause the subassembly to burn out when serious. Therefore, the photovoltaic cleaning robot plays an important role in cleaning the dirty battery plate.
Photovoltaic cleaning machines people mainly clears away panel surface dirt etc. through the brush, but because factors such as the dirty type in panel surface and dirty position, can cause the dirty degree of difficulty that clears away in panel surface to be different. Therefore, aiming at the problems, the invention dynamically adjusts the brush power of the cleaning robot through the dirty characteristic data of each area of the battery panel, realizes the regulation and control of the brush power of the photovoltaic cleaning robot according to the requirements, improves the working efficiency of the cleaning robot, reduces the energy consumption of the cleaning robot, and further ensures the cleaning effect.
Disclosure of Invention
The invention provides a photovoltaic cleaning robot brush power adjusting method and system based on artificial intelligence, and aims to solve the problems of low efficiency, high power consumption, incapability of cleaning according to requirements and the like in the prior art.
The invention discloses a photovoltaic cleaning robot brush power adjusting method based on artificial intelligence, which adopts the following technical scheme:
the method comprises the following steps: acquiring images of all photovoltaic cell panels;
step two: obtaining a dirty area in the photovoltaic cell panel image by semantic segmentation, analyzing the dirty area to obtain dirty area information and dirty area position distribution information, and extracting a dirty thickness index of the dirty area;
step three: extracting edge lines in the battery plate image, and calculating the sum of dirty pixels on the battery plate edge lines in each dirty area;
step four: taking the dirty pixel sum as a dirty position characteristic value, and calculating an edge dirty index;
step five: establishing a dirty area characteristic vector according to the dirty area information, the dirty position characteristic value, the dirty thickness index and the edge dirty index;
step six: adjusting the power of a robot brush for cleaning each dirty area by using the dirty characteristic vector of the dirty area;
the brush power adjusting method comprises the following steps:
dividing the area of the cell panel;
the dirty characteristic vector of each region is formed into the dimension of the region asDirty feature matrix of:
In the formula (I), the compound is shown in the specification,is the dirty feature matrix for the kth region,carrying out mean value calculation on each column of the matrix for the number of dirty areas in the kth area of the cell panel:
in the formula (I), the compound is shown in the specification,is a dirty featureThen obtain the dirty feature vector index of the area:
Constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next partition.
The dirty region feature vector is:
in the formula (I), the compound is shown in the specification,is as followsA dirty region feature vector for each dirty region,is as followsThe area of the individual dirty regions,is as followsA dirty position characteristic value of each dirty region,is as followsAn indication of the thickness of the soil in each of the dirty regions,is a firstEdge smudge indicators for individual smudge zones.
battery plateRGBConverting the image into a gray map, and segmenting by adopting a maximum entropy threshold segmentation algorithm to obtain a panel edge binary map;
carrying out opening operation denoising on the binary image;
randomly selecting edge line pixel points in the image to perform curve fitting to obtain an edge line equation;
substituting the coordinates of the pixel points in the dirty area into an edge line equation, and judging whether the dirty pixel points are positioned on the edge line of the battery plate;
carrying out pixel summation on pixel points of each dirty area on the edge line of the cell panel to obtain dirty characteristic values。
multiplying the semantically segmented image with the original image to obtain the dirtRGBAn image;
for dirty and dirtyRGBImage processingHISColor conversion to extract brightness information of dirty regionAnd saturation information;
in the formula (I), the compound is shown in the specification,respectively luminance and saturation information of the dirty region,to be at leastAnd (5) adjusting parameters.
in the formula (I), the compound is shown in the specification,is as followsThe edge dirty index of each dirty region,is as followsLength of dirt in individual dirt areas at the edge line of the panel, i.e. secondThe sum of the number of pixels on the panel edge line within each dirty region.
The brush power adjusting method comprises the following steps:
carrying out area division on the cell panel;
the dirty characteristic vector of each region is formed into the dimension of the region asDirty feature matrix of:
In the formula (I), the compound is shown in the specification,is as followskA dirty signature matrix for each of the regions,for the battery platekThe number of dirty areas in each area is calculated by the mean value of each row of the matrix:
in the formula (I), the compound is shown in the specification,is a dirty featureThen obtain the dirty feature vector index of the region:
Constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next partition.
The region division method comprises the following steps:
acquiring a distribution set of the dirty area according to the dirty area position distribution information;
by passingK-meansClustering analysis is carried out on the distribution situation of the dirty positions of the battery plates through a clustering algorithm to obtain a plurality of clusters;
randomly selecting a cluster as a core to find out the dirt in the clusterTwo pairs of dirty pixel points which are closest to each other in the two adjacent clusters and the pixel are obtained, two straight lines perpendicular to a line segment formed by the two pairs of dirty pixel points are obtained, the intersection point of the two straight lines is recorded as o, and a pair of dirty pixel points which are closest to each other in the two adjacent clusters is found, and the crossing point is obtainedoAnd (5) making a vertical line of a line segment formed by the pair of dirty pixel points, and finishing the region division of the cell panel.
The network prediction model adopts a full-connection prediction network model, takes the characteristic vector index of a dirty area as input, takes the power of a brush in the area as output, adopts a cross entropy loss function in the network training process, continuously updates parameters, and carries out iterative training.
The method for acquiring the adjusting time comprises the following steps:
when the cleaning robot finishes cleaning in the current subarea, the robot is positioned at the terminal point of the current subarea, and the coordinates of the current terminal point are obtained;
acquiring the coordinate of a point of the next partition closest to the current end point coordinate;
the brush power regulation time formula obtained according to the current end point coordinate and the coordinate closest to the end point coordinate in the next subarea is as follows:
in the formula (I), the compound is shown in the specification,is the coordinate information of the closest point of the next partition from the robot,coordinates of a central point of the robot when the robot reaches the current subarea cleaning end point, v is the running speed of the robot,is a delay factor.
This technical scheme still provides photovoltaic cleaning machines people brush power adjustment system based on artificial intelligence, including image acquisition module, image processing module, dirty characteristic extraction module, brush power calculation module, clean strategy module:
the image acquisition module acquires images of all the battery panels through the monitoring camera;
the image processing module identifies the dirty information of the battery panel through a semantic segmentation network;
the battery panel is characterized in that the dirty characteristic extraction module extracts dirty characteristic data including a dirty area, a dirty characteristic value, a dirty thickness index and an edge dirty index by analyzing dirty information of the battery panel, and a dirty area characteristic vector model is constructed according to the dirty characteristic data;
the cleaning strategy module:
the method comprises the steps that the battery panel is divided into areas, and a dirty feature matrix of each area is formed by using dirty feature vectors of the areas;
carrying out mean value calculation on each row of the dirty characteristic matrix to obtain a dirty characteristic vector index of the region;
constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next partition.
The invention has the beneficial effects that:
according to the invention, the position information of the dirt on the surface of the battery panel is considered, when the dirt is positioned at the edge position of the joint of the adjacent battery panels, the edge position is uneven according to the priori knowledge, so that the dirt is difficult to clean, the battery panels are not completely cleaned, and meanwhile, for the area with higher dirt thickness, the power of the brush is properly increased in the cleaning process of the cleaning robot, so that the dirt on the surface of the battery panel is ensured to be removed. In order to improve the cleaning efficiency of the robot, the invention carries out cluster analysis on the battery panel according to the dirty distribution characteristics of the battery panel, and carries out partition processing on the battery panel by adopting a support vector machine algorithm, and carries out targeted dynamic adjustment on the brush power based on dirty characteristic vector indexes of each partition to realize cleaning according to needs.
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, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of a method for adjusting brush power of a photovoltaic cleaning robot based on artificial intelligence;
FIG. 2 is a schematic view of a partition in the artificial intelligence-based photovoltaic cleaning robot brush power adjustment method of the invention;
FIG. 3 is a schematic diagram of a cleaning path in the artificial intelligence-based photovoltaic cleaning robot brush power adjustment method of the invention;
FIG. 4 is a block diagram of a brush power adjustment system of a photovoltaic cleaning robot based on artificial intelligence according to the present invention;
fig. 5 is an application scene diagram of the photovoltaic cleaning robot brush power adjustment method based on artificial intelligence.
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 embodiment of the invention relates to a method for adjusting the brush power of a photovoltaic cleaning robot based on artificial intelligence, as shown in fig. 1, comprising the following steps:
the method comprises the following steps: acquiring images of all photovoltaic cell panels;
the purpose of this step is through each surveillance camera collection panel image that photovoltaic power plant installed, the installation of camera is arranged the implementer and is set up by oneself according to the regional scope of photovoltaic power plant panel, guarantee that photovoltaic power plant's surveillance camera shooting scope can contain all panels, this embodiment is at first shot the panel surface, adopt each surveillance camera to realize the image acquisition to the panel surface, and number each panel and handle, make cleaning robot discern judgement to each panel, automatic shutdown when so that cleaning robot cleaned last panel.
It should be noted that wireless data can be transmitted between each monitoring camera in the photovoltaic cell area and the cleaning robot, and the next adjacent cell panel number and the specific situation of the cell panel, which have the same running trend as the robot, can be transmitted to the cleaning robot according to the current position information of the robot, so that the self-adaptive adjustment of the power of the robot brush can be performed based on the dirt characteristics of the cell panel in the following process.
Step two: obtaining a dirty area in the photovoltaic cell panel image by semantic segmentation, analyzing the dirty area to obtain dirty area information and dirty area position distribution information, and extracting a dirty thickness index of the dirty area;
the method aims to sense the surface contamination of the battery panel by adopting a semantic segmentation network, realize the identification of the distribution information of the contaminated area and the position of the battery panel and construct a contamination thickness index according to an image.
The specific content of the semantic segmentation network is as follows:
(1) the collected battery plate images are used as training data sets, the data sets are labeled, the battery plates with clean surfaces are labeled as 0, the pixels of the areas with dirt on the surfaces of the battery plates are labeled as 1, 80% of the data sets are randomly selected as the training sets, and the remaining 20% of the data sets are selected as verification sets.
(2) Inputting image data and tag data into a network, smudging perceptionEncoderExtracting the features of the image and outputting the image asFeature map(ii) a And then detected by soilDecoderTo pairFeature mapAnd performing upsampling, and outputting the semantic segmentation image of the surface dirt of the battery plate by a network.
(3) Of pollution-aware networkslossThe function is trained using a cross entropy loss function.
It should be noted that, the smudges on the surface of the battery panel can be identified and judged according to semantic segmentation, each smudged connected domain is obtained to obtain a smudged region, and the area of each smudged region of the battery panel is calculated through connected domain analysisAnd distribution of each dirty position,Is as followsThe number of the dirty areas is reduced,is a firstCoordinates of the center point of each dirty region.
Furthermore, the method analyzes the dirt thickness index of each dirt area, is used for detecting the dirt degree, and multiplies the dirt semantic perception effect image by the original image to obtain the dirtRGBThe image is used for visually analyzing the dirty thickness index, and in order to accurately obtain the dirty thickness index of the battery panel, the image is firstly processedHISAnd (4) color conversion.
HISThe color conversion comprises the following specific steps:
(1) to pairRGBThe images were normalized to ensure that their values were all normalized to [0,1 ]]Internal;
(2) then the following treatment is carried out:
taking into account the existence of the result of the calculation of the tone functionThe following will therefore be done for the tone function:
(3) switching over a soiled area toHSIAfter space, based onHSIThe color model analyzes the color characteristics of the battery panel dirt, and the saturation and brightness pairs of the dirt areas with different degrees are greatly different, so that the dirt thickness index is analyzed and calculated based on the color characteristics of the battery panel surface as the dirt characteristics of the subsequent battery panel analysis, and the following dirt thickness index analysis model is established in the invention:
in the formula (I), the compound is shown in the specification,respectively luminance and saturation information of the dirty region,the adjustable parameters of the model are selected by an implementer, and the invention is set as. The higher the value of the model function is, the larger the dirt thickness index of the dirt area is, and the higher the dirt degree is.
Step three: extracting edge lines in the battery plate image, and calculating the sum of dirty pixels on the battery plate edge lines in each dirty area;
the purpose of this step is for the convenience of follow-up directly perceived statistics panel edge dirty pixel, draws the curve at edge pixel place through the method of fitting simultaneously, can accurately acquire panel edge line place equation, and this method can prevent effectively that the dirt from sheltering from the influence to the edge line information integrality who draws, improves the whole precision of system, and concrete step is as follows:
(1) and extracting the edge line of the battery plate so as to judge the dirty position of the battery plate. The method comprises the steps of converting an RGB image of the battery panel into a gray level image, segmenting the battery panel image by adopting a maximum entropy threshold segmentation algorithm, extracting battery panel edge pixels in the image, wherein the pixel value of the battery panel edge position is 1, the pixel values of other areas of the battery panel are 0, and finally obtaining a binary image of the battery panel edge.
(2) And after the image is segmented, performing morphological processing on the panel edge binary image to eliminate fine white noise points caused by grid lines in the image. The method carries out morphological operation, specifically opening operation, on the binary image, firstly carries out corrosion operation on the image, and then carries out expansion processing, thereby eliminating fine white noise in the image. Therefore, the edge lines of the battery panel can be extracted through the image processing, and a binary image only containing the edge lines of the battery panel is obtained.
(3) In the embodiment, the situation that dirty shielding exists in the process of extracting the edge line of the solar panel is considered, so that the extracted edge information of the solar panel is incomplete. Therefore, in order to accurately judge whether the dirt is positioned at the edge position of the battery panel, the invention fits the curve of the battery panel edge according to the edge pixel point in the battery panel edge binary image so as to accurately obtain the battery panel edge equation for judging the subsequent dirt position characteristics. And obtaining coordinates of edge pixel points of the battery panel according to the binary image, and fitting the obtained scattered points of the edge pixels to obtain an edge line equation.
(4) The pixel coordinates in the dirty area are brought into a battery panel edge line equation, whether dirty pixel points are located on an edge line or not is verified, and the sum of dirty pixels located at the edge line position in each dirty area is counted。
The battery plate edge line fitting process specifically comprises the following steps: firstly, randomly selectingFitting a curve with points, and calculating the distances from all the points to the curve, wherein the distances are less than a threshold valueJudging that the point belongs to the curve, recording the number of the points belonging to the curve, then replacing the selected points, refitting a new curve, calculating the number of the points belonging to the curve, and after the search is finished, selecting the curve equation containing the most points as the edge line equation.
Step four: taking the dirty pixel sum as a dirty position characteristic value, and calculating an edge dirty index;
the purpose of this step is to reduce the number of stepsAs the dirty position characteristic value of the dirty region, in the present embodiment, considering that the higher the degree of aggregation of the dirty at the edge, the more difficult it is to clean the cleaning robot, the degree of aggregation of the dirty at the edge of the battery panel in each connected domain is analyzed, and this is used as the edge dirty indexConstructing an edge dirt aggregation degree analysis model for analyzing the edge dirt aggregation degreeThe degree of accumulation of dirt at the rim location. Analyzing each edge smudge-containing area on the panel, and calculating the concentration of edge smudge in each area:
in the formula (I), the compound is shown in the specification,is as followsThe edge dirty index of each dirty region,is a firstThe dirty length of the edge line of the battery plate in each dirty area isThe sum of the number of dirty pixels on the edge line of the battery panel in each dirty area is higher, the higher the model function value is, the higher the dirty aggregation degree of the inner edge of the corresponding dirty area is, and the higher the difficulty of the cleaning robot in cleaning work is.
Step five: establishing a dirty area characteristic vector according to the dirty area information, the dirty position characteristic value, the dirty thickness index and the edge dirty index;
the purpose of this step is, in order to clean the dirty of panel accurately, improve the precision of cleaning method, this embodiment constructs the model according to the relevant characteristic index in dirty region, makes things convenient for follow-up regulation power.
Wherein, dirty regional area information, dirty position characteristic value, marginal dirty degree of aggregation and dirty thickness index have been obtained in step one to step four, do further analysis to each dirty region on the panel, obtainFeature vector of dirty regionFor an entire individual panel, a set of panel dirty feature vectors may be acquired,And the number of the dirty connected domains is used for adjusting and controlling the power of the cleaning robot subsequently based on the number of the dirty connected domains.
Step six: and adjusting the power of the robot brush for cleaning each dirty area by using the dirty characteristic vector of the dirty area.
The purpose of this step is based on each dirty distribution information realizes the regional division to the panel, according to the dirty distribution condition on the panel, and the pertinence is handled in subregion, compares in fixed area's the segmentation have stronger generalization nature, improves the adaptability of system to, through effectively carrying out the subregion of panel according to dirty characteristic, can pertinence carry out dynamic adjustment to photovoltaic cleaning robot brush power, be convenient for realize that cleaning robot power realizes adjusting as required.
The area division comprises the following specific steps:
firstly, acquiring each dirty connected domain of the cell panel, analyzing the connected domains to obtain the central point of each dirty region, and acquiring the distribution set of the dirty regions ;
Then, clustering analysis is carried out through a clustering algorithm based on the dirty position distribution condition of the battery panel, three clustering center points are initially selected, and the method adoptsK-meansCarrying out clustering analysis on the clusters by an algorithm to obtain a plurality of clusters;
after each category is obtained, the dynamic area division of the battery panel is realized based on each dirt distribution information in the embodiment, the main purpose of the step is to perform partition processing in a targeted manner according to the dirt distribution situation on the battery panel, and compared with the division of a fixed area, the method has stronger generalization and improves the adaptability of the system.
Considering that only the smudges of the battery panel are classified after the clustering analysis, only a few clusters on the battery panel can be obtained, but the photovoltaic cleaning robot needs to clean and sweep the battery panel comprehensively in the process of cleaning the battery panel, so that the whole battery panel is partitioned based on the classified categories in order to clean the whole battery panel conveniently by the follow-up cleaning robot. The partitioning treatment specific process of the invention is as follows:
for the result after clustering, as shown in fig. 2, in this embodiment, first, a cluster is arbitrarily selected as a core, two pairs of dirty pixels closest to each other in the dirty pixel and two adjacent clusters in the cluster are found, two straight lines perpendicular to a line segment formed by the two pairs of dirty pixels are obtained, and an intersection point of the two straight lines is recorded as a pointOThen finding out a pair of dirty pixel points with the shortest distance between two adjacent clusters, and passing the dirty pixel pointsOAnd making a vertical line of a line segment formed by the pair of dirty pixel points, and realizing the partition processing of the battery panel based on the dirty clustering result.
Further, obtaining the feature vector indexes of each partitioned dirty area:
for the partitioned battery panel, each partition is divided into the dirty characteristic vectors to form a partition dimension ofThe dirty feature matrix of (a):
in the formula (I), the compound is shown in the specification,is as followskA dirty signature matrix for each of the regions,is a battery platekThe number of dirty areas in each area is calculated by the mean value of each row of the matrix:
in the formula (I), the compound is shown in the specification,is a dirty featureAnd finally acquiring the dirty characteristic vector indexes of the partition as follows:
further, in this embodiment, a network prediction model is constructed based on dirty characteristic vector indexes of each partition of the battery panel to predict and adjust the brush power of the cleaning robot:
the input of the prediction network model is a regional dirty characteristic vector indexThe network output is the working power of the brush in the subareaThe working power data of the brush in each partition can be obtained by installing a micro power detector on the robot, iterative training is carried out by adopting a cross entropy loss function in the network training process, network parameters are continuously updated, a prediction network can adopt a full-connection prediction network model, the relation between a dirt characteristic index and corresponding power can be obtained by training, in practical application, the dirt characteristic index is input into the prediction model, and the corresponding power of the brush of the cleaning robot in the area can be obtained, so that dynamic adjustment is carried out.
It should be noted that, for the brush power prediction adjustment model for each partition, an implementer may also construct a brush power prediction adjustment function model based on the feature vector indexes of each region:the brush power in the cleaning process can be predicted and regulated through the mathematical model, and the effect of cleaning as required can be achieved.
And adjusting the robot brush power of the next partition needing to be cleaned according to the prediction model and the dirty characteristic vector index of the next partition needing to be cleaned by the robot.
It should be noted that, for the brush power of the photovoltaic cleaning robot in each partition, the embodiment dynamically adjusts the brush power according to the dirty feature vector index in each partition, and since each battery panel is divided into three partitions in the embodiment, the number of times of adjusting the brush power corresponding to each battery panel is at most 3 in the cleaning process of the photovoltaic battery panel by the cleaning robot. The number of the partitions is the same as that of the clustering centers, and the setting implementer of the number of the clustering centers can set the number of the clustering centers by himself.
Further, the adjusting time required by the completion of the power adjustment is calculated, and the adjusting time is adjusted in advance through the hairbrush power adjusting time, so that the robot is ensured to be cleaned as required, the energy consumption of the robot is reduced, and the cleaning efficiency of the robot is improved, wherein the method comprises the following steps:
when the cleaning robot finishes cleaning in the current partition, the terminal point of the robot in the current partition is obtained, the coordinates of the point of the next partition, which is closest to the robot, are obtained according to the running area of the robot, and based on the position information of the two, the adjusting time of the brush power of the robot in the next partition is analyzed in the embodiment:
in the formula (I), the compound is shown in the specification,is the coordinate information of the closest point of the next partition from the robot,coordinates of a center point of the robot when the robot reaches the current subarea cleaning end point,vfor the operation speed of the robot, the embodiment sets the robot to operate at a constant speed in the cleaning process. Considering the fact that delay phenomenon exists in the process of regulating and controlling the power of the robot brush, the invention is provided withDelay factors to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next subarea,the value implementer robot is set according to the actual situation, and the embodiment sets the value implementer robot into the value implementer robot。
Further, the embodiment further provides a method for setting the cleaning path of the robot after calculating the adjustment time, as shown in fig. 3:
after cleaning the current subarea, the cleaning robot reaches the terminal point of the current subarea, firstly, the closest point is obtained to ensure that the robot reaches the next adjacent subarea as soon as possible, in the process of going to the initial point of the next subarea, the robot retracts the brush, the brush is operated in a straight-line mode to reach the closest point with low power consumption, and the closest point is the cleaning initial point of the robot of the next adjacent subarea;
acquiring the adjusting time of the brush power of the next subarea according to the brush power adjusting time, and taking the point as the initial cleaning point of the robot of the next subarea after the closest point is reached, wherein the robot walks around the subarea boundary for a circle in the clockwise direction by adopting the pre-adjusted and controlled brush power;
then inward each time after reaching the initial pointMoved by a distanceLAnd then continuing to advance along the same path as the contour of the subarea, and so on until the center of the subarea is reached as the end point of the robot in the subarea, wherein the moving distance L of the robot at each time is determined according to the size of the photovoltaic cleaning robot.
So far, can realize carrying out regulation and control in advance to the brush power of robot when next subregion is clean based on each subregion surface dirty characteristic vector index and the position of robot place panel down to guarantee cleaning machines people's clean effect, realize cleaning as required, improve the life of robot simultaneously. The method can automatically and adaptively adjust the power of the robot brush in advance based on the dirt characteristics of each partition of the battery plate, and reduces the subjectivity and the false detection rate of artificial detection control.
Embodiment 2 as shown in fig. 4, the artificial intelligence-based power adjustment system for a brush of a photovoltaic cleaning robot includes:
image acquisition moduleS100Acquiring images of all the battery panels by a monitoring camera;
image processing moduleS101Identifying the dirty information of the battery plate through a semantic segmentation network;
dirty characteristic extraction moduleS102Extracting dirty characteristic data including dirty area, dirty characteristic value, dirty thickness index and edge dirty index by analyzing dirty information of the battery plate, and constructing a feature vector model according to the dirty area according to the dirty characteristic data;
the brush power calculation module S103 is used for obtaining a dirty area characteristic vector of a corresponding area according to the dirty area characteristic vector model by partitioning the battery panel, and obtaining the power of the brush in the corresponding area by utilizing a prediction network;
the cleaning strategy module S104:
the method comprises the steps that the battery panel is divided into areas, and a dirty feature matrix of each area is formed by using dirty feature vectors of the areas;
carrying out mean value calculation on each row of the dirty characteristic matrix to obtain a dirty characteristic vector index of the region;
constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time, ensuring that the robot just completes the adjustment of the brush power when reaching the starting point of the next partition, and setting a cleaning path.
The embodiment can use a photovoltaic cleaning scene as shown in fig. 5, the technical scheme is based on artificial intelligence and image processing, the battery pollution area is accurately positioned, the power of the photovoltaic robot is adjusted in advance according to the pollution degree, cleaning according to needs is achieved, power consumption is reduced, and cleaning efficiency is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (9)
1. A photovoltaic cleaning robot brush power adjustment method based on artificial intelligence is characterized by comprising the following steps:
the method comprises the following steps: acquiring images of all photovoltaic cell panels;
step two: obtaining a dirty area in the photovoltaic cell panel image by semantic segmentation, analyzing the dirty area to obtain dirty area information and dirty area position distribution information, and extracting a dirty thickness index of the dirty area;
step three: extracting edge lines in the battery plate image, and calculating the sum of dirty pixels on the battery plate edge lines in each dirty area;
step four: taking the dirty pixel sum as a dirty position characteristic value, and calculating an edge dirty index;
step five: establishing a dirty area characteristic vector according to the dirty area information, the dirty position characteristic value, the dirty thickness index and the edge dirty index;
step six: adjusting the power of a robot brush for cleaning each dirty area by using the dirty characteristic vector of the dirty area;
the brush power adjusting method comprises the following steps:
dividing the area of the cell panel;
the dirty characteristic vector of each region is formed into the dimension of the region asDirty feature matrix of:
In the formula (I), the compound is shown in the specification,is the dirty feature matrix for the kth region,carrying out mean value calculation on each column of the matrix for the number of dirty areas in the kth area of the cell panel:
in the formula (I), the compound is shown in the specification,is a dirty featureThen obtain the dirty feature vector index of the region:
Constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next partition.
2. The artificial intelligence based photovoltaic cleaning robot brush power adjustment method according to claim 1, wherein the dirty region feature vector is:
in the formula (I), the compound is shown in the specification,is as followsA dirty region feature vector for each dirty region,is as followsThe area of each of the dirty regions is,is as followsA dirty position characteristic value of each dirty region,is as followsAn indication of the thickness of the soil in each of the soiled areas,is as followsEdge smudge indicators for individual smudge zones.
3. The artificial intelligence based photovoltaic cleaning robot brush power adjustment method according to claim 2, wherein the dirty position characteristic valueThe acquisition method comprises the following steps:
converting the RGB image of the battery panel into a gray map, and segmenting by adopting a maximum entropy threshold segmentation algorithm to obtain a battery panel edge binary map;
carrying out opening operation denoising on the binary image;
randomly selecting edge line pixel points in the image to perform curve fitting to obtain an edge line equation;
substituting the coordinates of the pixel points in the dirty area into an edge line equation, and judging whether the dirty pixel points are positioned on the edge line of the battery plate;
4. The artificial intelligence-based photovoltaic cleaning robot brush power adjustment method according to claim 2, wherein the dirty thickness indexThe acquisition steps are as follows:
multiplying the semantically segmented image with an original image to obtain a dirty RGB image;
performing HIS color conversion on the dirty RGB image, and extracting brightness information of the dirty regionAnd saturation information;
5. The artificial intelligence based photovoltaic cleaning robot brush power adjustment method according to claim 2, wherein the edge contamination indicatorThe calculation method comprises the following steps:
6. The artificial intelligence based photovoltaic cleaning robot brush power adjustment method according to claim 1, wherein the area division method is as follows:
acquiring a distribution set of the dirty areas according to the dirty area position distribution information;
carrying out clustering analysis on the distribution condition of the dirty positions of the battery plate by using a K-means clustering algorithm to obtain a plurality of clusters;
randomly selecting one cluster as a core, finding out two pairs of dirty pixel points which are closest to each other in the dirty pixels in the cluster and two adjacent clusters, obtaining two straight lines which are perpendicular to a line segment formed by the two pairs of dirty pixel points, recording the intersection point of the two straight lines as o, finding out one pair of dirty pixel points which are closest to each other in the two adjacent clusters, and making a vertical line of the line segment formed by the pair of dirty pixel points through the o point to complete the regional division of the battery panel.
7. The artificial intelligence-based photovoltaic cleaning robot brush power adjustment method according to claim 1, wherein the network prediction model adopts a full-connection prediction network model, a feature vector index of a dirty region is used as an input, the power of the brush in the region is used as an output, a cross entropy loss function is adopted in a network training process, parameters are continuously updated, and iterative training is performed.
8. The artificial intelligence based photovoltaic cleaning robot brush power adjustment method according to claim 1, wherein the adjustment time obtaining method is:
when the cleaning robot finishes cleaning in the current subarea, the robot is positioned at the terminal point of the current subarea, and the coordinates of the current terminal point are obtained;
acquiring the coordinate of a point of the next partition closest to the current end point coordinate;
the brush power regulation time formula obtained according to the current end point coordinate and the coordinate closest to the end point coordinate in the next subarea is as follows:
in the formula (I), the compound is shown in the specification,is the coordinate information of the closest point of the next partition from the robot,the coordinate of the center point of the robot when the robot reaches the current subarea cleaning end point, v is the running speed of the robot,is a delay factor.
9. The utility model provides a photovoltaic cleaning robot brush power adjustment system based on artificial intelligence which characterized in that, includes image acquisition module, image processing module, dirty characteristic extraction module, brush power calculation module, clean strategy module:
the image acquisition module acquires images of all the battery panels through the monitoring camera;
the image processing module identifies the dirty information of the battery panel through a semantic segmentation network;
the battery panel is characterized in that the dirty characteristic extraction module extracts dirty characteristic data including a dirty area, a dirty characteristic value, a dirty thickness index and an edge dirty index by analyzing dirty information of the battery panel, and a dirty area characteristic vector model is constructed according to the dirty characteristic data;
the cleaning strategy module:
the method comprises the steps that the battery panel is divided into areas, and a dirty feature matrix of each area is formed by using dirty feature vectors of the areas;
carrying out mean value calculation on each row of the dirty characteristic matrix to obtain a dirty characteristic vector index of the region;
constructing a network prediction model, and obtaining the power of the cleaning robot brush in a corresponding area through the dirty characteristic vector indexes of each area of the battery panel;
and calculating the adjusting time to ensure that the robot just completes the adjustment of the brush power when reaching the starting point of the next subarea.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210818928.5A CN114897918B (en) | 2022-07-13 | 2022-07-13 | Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210818928.5A CN114897918B (en) | 2022-07-13 | 2022-07-13 | Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114897918A CN114897918A (en) | 2022-08-12 |
CN114897918B true CN114897918B (en) | 2022-09-27 |
Family
ID=82729548
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210818928.5A Active CN114897918B (en) | 2022-07-13 | 2022-07-13 | Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114897918B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115082744B (en) * | 2022-08-22 | 2023-09-08 | 深圳烁圣数字能源有限公司 | Artificial intelligence-based solar heat collection efficiency analysis method and system |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9424660B2 (en) * | 2012-08-07 | 2016-08-23 | Intel Corporation | Media encoding using changed regions |
CN112232532A (en) * | 2020-10-14 | 2021-01-15 | 孙猛猛 | Photovoltaic cell panel night cleaning scheme automatic generation method based on visual perception |
CN112329618A (en) * | 2020-11-04 | 2021-02-05 | 夏永霞 | Photovoltaic cleaning robot control method and system based on artificial intelligence |
CN112734766A (en) * | 2020-12-14 | 2021-04-30 | 王富才 | Cleaning parameter adjusting method and system of photovoltaic cleaning robot based on artificial intelligence |
-
2022
- 2022-07-13 CN CN202210818928.5A patent/CN114897918B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN114897918A (en) | 2022-08-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104792792B (en) | A kind of road surface crack detection method of Stepwise Refinement | |
CN108040997B (en) | Insect pest monitoring method based on machine vision | |
CN110458839B (en) | Effective wire and cable monitoring system | |
CN107895376A (en) | Based on the solar panel recognition methods for improving Canny operators and contour area threshold value | |
CN112329618A (en) | Photovoltaic cleaning robot control method and system based on artificial intelligence | |
CN103699900B (en) | Building horizontal vector profile automatic batch extracting method in satellite image | |
CN108108772B (en) | Insulator pollution flashover state detection method based on aerial image of distribution line | |
CN114897918B (en) | Artificial intelligence-based photovoltaic cleaning robot brush power adjustment method and system | |
CN113313677B (en) | Quality detection method for X-ray image of wound lithium battery | |
CN116758071B (en) | Intelligent detection method for carbon electrode dirt under visual assistance | |
CN112588764A (en) | Multi-robot collaborative cleaning method and device on photovoltaic module based on artificial intelligence | |
CN112232532A (en) | Photovoltaic cell panel night cleaning scheme automatic generation method based on visual perception | |
CN111864896A (en) | Power load monitoring method and system | |
CN111242096B (en) | People number gradient-based people group distinguishing method | |
CN115082744B (en) | Artificial intelligence-based solar heat collection efficiency analysis method and system | |
CN105654099A (en) | Sugarcane segmentation and identification method based on improved vision | |
CN111738931A (en) | Shadow removal algorithm for aerial image of photovoltaic array unmanned aerial vehicle | |
CN116993745B (en) | Method for detecting surface leakage of water supply pipe based on image processing | |
CN116191680B (en) | Monitoring management system applied to photovoltaic power generation | |
CN110909751B (en) | Visual identification method, system and medium for transformer substation insulator cleaning robot | |
CN111192332B (en) | Smoke detection-based smoke machine control method and smoke machine | |
CN105844641A (en) | Adaptive threshold segmentation method in dynamic environment | |
CN108805136B (en) | Significance detection method for water surface pollutant monitoring | |
CN110570437B (en) | Electric power channel automatic inspection data processing method based on boundary identification | |
CN111985492B (en) | Cloud identification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |