Cleaning robot cleaning control method and system

Abstract

The invention provides a cleaning robot cleaning control method and system in the technical field of cleaning robots, and the method comprises the steps: S1, collecting an environment image by a cleaning robot, extracting stain features, and positioning a stain region; s2, clustering each stain feature to divide stain categories; s3, matching the adhesion strength index and the component influence factor based on the stain category, and identifying the thickness index of the stain based on the environment image; s4, calculating a stain strength index based on the thickness index, the adhesive strength index and the component influence factor; s5, dividing the environment image into a plurality of cleaning units, planning a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, matching a cleaning intensity index based on the stain category, and calculating the number of cleaning times based on the cleaning intensity index and the stain intensity index; and S6, cleaning operation is executed based on the cleaning path and the cleaning times. The cleaning robot has the advantage that the cleaning quality, flexibility and efficiency of the cleaning robot are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cleaning robots, and in particular to a cleaning control method and system for a cleaning robot. Background Art

[0002] Cleaning robots are intelligent devices that can automatically complete cleaning tasks and are widely used in homes, commercial spaces, public areas, and other scenarios. Cleaning robots typically complete cleaning tasks by using a robotic arm to extend and rotate, along with cleaning tools (such as brushes and suction heads). For example, in some industrial scenarios, the robotic arm will move to a designated location according to a preset program and use a brush to wipe or sweep. However, traditional cleaning robots have the following shortcomings during the cleaning process:

[0003] 1. Robotic arms often have fixed motions and paths, making them ineffective in complex environments or corners. 2. It's difficult to adjust the cleaning force based on the cleaning target and stain type; for example, using the same cleaning mode for both kitchen grease and living room dust can result in poor cleaning results. 3. Robotic arms have relatively fixed motions and paths, making them difficult to adapt to complex and changing environments. 4. Complex tasks require multiple round trips or repetitive operations, resulting in low efficiency.

[0004] Therefore, how to provide a cleaning robot cleaning control method and system to improve the cleaning quality, flexibility and efficiency of the cleaning robot has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a cleaning control method and system for a cleaning robot, so as to improve the cleaning quality, flexibility and efficiency of the cleaning robot.

[0006] In a first aspect, the present invention provides a cleaning robot cleaning control method, comprising the following steps:

[0007] Step S1: The cleaning robot collects environmental images in real time and pre-processes the environmental images;

[0008] Step S2: The cleaning robot extracts stain features including at least color, texture, and shape from the pre-processed environment image, and locates the stain area based on the stain features;

[0009] Step S3: The cleaning robot clusters the stain features using a K-means clustering algorithm to further classify the stains.

[0010] Step S4: The cleaning robot matches the corresponding adhesion strength index and component influencing factor based on the stain category, and identifies the thickness index of the stain based on the environmental image;

[0011] Step S5: The cleaning robot calculates a stain intensity index based on the thickness index, the adhesion strength index, and the component influencing factor;

[0012] Step S6: The cleaning robot divides the environment image into a plurality of cleaning units based on a preset segmentation size, plans a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, matches the corresponding cleaning intensity index based on the stain category, and calculates the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index;

[0013] Step S7: The cleaning robot performs a cleaning operation based on the cleaning path and the number of cleaning times, and dynamically adjusts the cleaning path based on the latest collected environment image.

[0014] Furthermore, the step S1 is specifically as follows:

[0015] The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement and brightness adjustment; the noise reduction is based on median filtering or Gaussian filtering; the contrast enhancement is based on histogram equalization or linear contrast stretching; and the brightness adjustment is based on linear transformation or gamma correction.

[0016] Furthermore, in step S2, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

[0017] Furthermore, in step S4, the thickness index is obtained by converting the shadow area in the environment image;

[0018] In step S5, the calculation formula of the stain intensity index is:

[0019] I=ɑ*T+β*A+γ*C;

[0020] Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

[0021] Furthermore, in step S6, the calculation formula for the number of cleaning times is:

[0022] times n =cei l(I n / effect n );

[0023] Among them, times nIndicates the number of cleaning times of the nth cleaning unit; ceil() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

[0024] In a second aspect, the present invention provides a cleaning robot cleaning control system, comprising the following modules:

[0025] An environmental image acquisition module is used by the cleaning robot to collect environmental images in real time and pre-process the environmental images;

[0026] A stain feature extraction module is used for the cleaning robot to extract stain features including at least color, texture and shape from the pre-processed environment image, and locate the stain area based on the stain features;

[0027] A stain classification module is used for the cleaning robot to cluster the stain features using a K-means clustering algorithm and then classify the stains into categories;

[0028] A data matching and identification module is used for the cleaning robot to match the corresponding adhesion strength index and component influencing factor based on the stain category and identify the thickness index of the stain based on the environmental image;

[0029] A stain intensity index calculation module, used for the cleaning robot to calculate the stain intensity index based on the thickness index, adhesion strength index and component influencing factor;

[0030] a cleaning path and number planning module, configured for the cleaning robot to divide the environment image into a plurality of cleaning units based on a preset segmentation size, plan a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, match the corresponding cleaning intensity index based on the stain category, and calculate the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index;

[0031] The cleaning operation module is used for the cleaning robot to perform a cleaning operation based on the cleaning path and the number of cleaning times, and dynamically adjust the cleaning path based on the latest collected environment image.

[0032] Furthermore, the environmental image acquisition module is specifically used to:

[0033] The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement and brightness adjustment; the noise reduction is based on median filtering or Gaussian filtering; the contrast enhancement is based on histogram equalization or linear contrast stretching; and the brightness adjustment is based on linear transformation or gamma correction.

[0034] Furthermore, in the stain feature extraction module, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

[0035] Furthermore, in the data matching and identification module, the thickness index is obtained by converting the shadow area in the environmental image;

[0036] In the stain intensity index calculation module, the calculation formula of the stain intensity index is:

[0037] I=ɑ*T+β*A+γ*C;

[0038] Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

[0039] Furthermore, in the cleaning path and times planning module, the calculation formula for the cleaning times is:

[0040] times n =cei l(I n / effect n );

[0041] Among them, times n Indicates the number of cleaning times of the nth cleaning unit; cei l() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

[0042] The advantages of the present invention are:

[0043] 1. The cleaning robot collects environmental images in real time and preprocesses them, extracts stain features including at least color, texture and shape from the preprocessed environmental images, locates the stain area based on the stain features, clusters the stain features using the K-means clustering algorithm and then divides the stain categories, matches the corresponding adhesion strength index and component influencing factors based on the stain categories, identifies the thickness index of the stain based on the environmental image, and calculates the stain intensity index based on the thickness index, adhesion strength index and component influencing factors; then the cleaning robot divides the environmental image into several cleaning units based on the preset segmentation size, plans the cleaning path based on the overlapping relationship between each cleaning unit and the stain area, and matches the stain category based on the stain category. It is equipped with a corresponding cleaning intensity index, and the cleaning times of each cleaning unit on the cleaning path are calculated based on the cleaning intensity index and the stain intensity index. Finally, the cleaning operation is performed based on the cleaning path and the cleaning times, and the cleaning path is dynamically adjusted based on the latest collected environmental image; that is, the cleaning robot automatically performs preprocessing, stain feature extraction, cleaning path planning, and cleaning times calculation through the real-time collected environmental image, and dynamically adjusts the cleaning path based on the latest environmental image. It can effectively cope with complex environments or corners, effectively adapt to complex and changing environments, and perform personalized cleaning for different types of stains based on the calculated cleaning times to avoid repeated operations, ultimately greatly improving the cleaning quality, flexibility and efficiency of the cleaning robot.

[0044] 2. By using median filtering / Gaussian filtering for noise reduction, histogram equalization for contrast enhancement, and gamma correction for brightness adjustment, we eliminate ambient light interference, ensure the robustness of subsequent feature extraction, improve image usability in low-light and high-noise scenarios, and reduce the false positive rate.

[0045] 3. By combining color (grayscale mean / variance), texture (grayscale co-occurrence matrix / LBP), and shape (contour algorithm) features, it covers the differences in physical and chemical properties of stains, avoids missed detection of a single feature (such as transparent oil stains through texture recognition and colored stains through color capture), and improves the comprehensiveness of stain detection.

[0046] 4. Classify stains into categories through unsupervised clustering, adapt to dynamic scenarios with unknown stain types, reduce reliance on manual labeling, and achieve data-driven self-learning of stain types.

[0047] 5. By introducing the thickness index (shadow conversion), adhesion strength index (physical adhesion), and component influencing factor (chemical corrosiveness), a comprehensive strength formula I = ɑ*T + β*A + γ*C is constructed to quantify the cleaning difficulty of different stains and avoid over-cleaning (waste of resources) or under-cleaning (residue) caused by fixed strength.

[0048] 6. By dividing the cleaning units by segmentation size, dynamically planning the path based on the overlapping relationship, and combining the intensity index to calculate the number of cleaning timesn=ceil(In / effectn), high-pollution areas can be treated first and ineffective movement can be reduced.

[0049] 7. By continuously acquiring new images while cleaning and dynamically adjusting the path and intensity, it can respond to sudden contamination (such as spilled liquid) and achieve closed-loop control.

[0050] 8. Support dynamic configuration through parameters such as weight coefficient (ɑ / β / γ) and segmentation size to adapt to different scenarios (home / factory).

[0051] 9. Accurately identify environmental stains through multimodal image preprocessing (noise reduction / contrast enhancement / brightness adjustment), combine color, texture, and shape multi-dimensional feature fusion with K-means clustering algorithm to adaptively divide stain types, and build a stain intensity quantification model based on thickness, adhesion strength, and composition factors (I = ɑ*T+β*A+γ*C). Dynamically plan the unit cleaning path and intelligently calculate the number of cleaning times. At the same time, through real-time image feedback closed-loop adjustment strategy, the coordinated optimization of stain recognition accuracy, cleaning efficiency and resource utilization is achieved. Its parameter scalability can adapt to the needs of multiple scenarios such as home and industry, solving the pain points of traditional robots such as single cleaning strength, rigid path, and poor environmental adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] Figure 1 The present invention is a flowchart of a cleaning control method for a cleaning robot.

[0054] Figure 2 It is a structural schematic diagram of a cleaning control system of a cleaning robot of the present invention. DETAILED DESCRIPTION

[0055] The technical solution in the embodiments of the present application has the following overall idea: the cleaning robot automatically performs preprocessing, stain feature extraction, cleaning path planning, and cleaning frequency calculation through real-time collected environmental images, and dynamically adjusts the cleaning path based on the latest environmental images. It can effectively cope with complex environments or corners, effectively adapt to complex and changing environments, and perform personalized cleaning for different types of stains based on the calculated cleaning frequency, avoiding repeated operations, so as to improve the cleaning quality, flexibility and efficiency of the cleaning robot.

[0056] Please refer to Figures 1 to 2 As shown, a preferred embodiment of a cleaning robot cleaning control method of the present invention includes the following steps:

[0057] Step S1: The cleaning robot collects an environmental image in real time and pre-processes the environmental image; the environmental image and the actual ground position are mapped based on a preset conversion relationship;

[0058] Step S2: The cleaning robot extracts stain features including at least color, texture, and shape from the pre-processed environment image, and locates the stain area based on the stain features;

[0059] Step S3: The cleaning robot clusters the stain features using a K-means clustering algorithm to further classify the stains.

[0060] Through unsupervised clustering, stain categories are divided to adapt to dynamic scenarios with unknown stain types, reduce dependence on manual labeling, and achieve data-driven self-learning of stain types.

[0061] Stains can also be classified based on thresholds, that is, classification is performed based on a threshold range of features. For example, stains can be classified as coffee stains, mud stains, etc. based on the range of color values. This method is simple and direct, but it requires a pre-determined threshold range. Clustering algorithms, on the other hand, calculate the distance information between different stain samples and classify them into K different categories. This method does not require a pre-determined threshold, but does require selecting an appropriate cluster center and number of categories. Clustering algorithms can automatically discover the distribution patterns of stain samples, thereby achieving stain classification.

[0062] Step S4: The cleaning robot matches the corresponding adhesion strength index and component influencing factor based on the stain category, and identifies the thickness index of the stain based on the environmental image;

[0063] Step S5: The cleaning robot calculates a stain intensity index based on the thickness index, the adhesion strength index, and the component influencing factor;

[0064] Quantitative assessment of stains is a key step in achieving efficient cleaning. Simply identifying and classifying stains is not enough; the severity of the stains also needs to be quantified to provide a scientific basis for the subsequent formulation of cleaning strategies. By quantifying stains into a series of numerical values, the cleaning difficulty and priority of the stains can be more intuitively assessed, thereby optimizing cleaning paths and methods. In addition, the quantitative data of stains can also be used to draw stain maps, providing a data basis for dynamic path planning.

[0065] Step S6: The cleaning robot divides the environment image into a plurality of cleaning units based on a preset segmentation size (for example, into a plurality of 5 mm × 5 mm small square cleaning units, which correspond one-to-one to the elements in a predefined two-dimensional array), plans a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, matches the corresponding cleaning intensity index based on the stain category, and calculates the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index;

[0066] By dividing the cleaning units by segmentation size, dynamically planning the path based on the overlapping relationship, and calculating the cleaning times timesn=ceil(In / effectn) in combination with the intensity index, high-pollution areas can be processed first and ineffective movement can be reduced.

[0067] In practice, a static path can be planned in a bow-shaped pattern based on the contours of the area to be cleaned, avoiding the problems of repeated cleaning or missed areas that may occur with traditional cleaning methods, thereby completing the cleaning task in the shortest possible time. When encountering an obstacle, the cleaning path (i.e., the dynamic path) is replanned only for the area covered by the static path that intersects the obstacle contour, while the original static path that does not intersect with the obstacle and has not been cleaned is retained. This reduces the number of back-and-forth travel paths of the cleaning robot between different areas and improves cleaning efficiency.

[0068] Step S7: The cleaning robot performs a cleaning operation based on the cleaning path and cleaning times, dynamically adjusts the cleaning path and cleaning times based on the most recently acquired environmental image, and verifies the cleaning effect through machine vision. Specifically, the cleaning robot can clean the stains in the cleaning unit step by step from top to bottom and from left to right.

[0069] In practice, the cleaning robot automatically records a map of the environment and determines its own position in real time. In complex indoor environments, such as those with obstacles like furniture and debris, the robot can adjust its cleaning path based on this real-time environmental information to ensure comprehensive coverage of the cleaning area and avoid missing areas due to obstructions. When an obstacle is detected while cleaning along a static path, the robot cleans along its edges to obtain its outline. It then replans a dynamic path for the area covered by the static path that intersects the obstacle outline and does not contain the obstacle. This precise path adjustment ensures the robot effectively cleans the area around the obstacle, improving cleaning quality. By optimizing the cleaning path, the robot reduces ineffective travel and repeated cleaning, thereby reducing energy consumption. During the cleaning process, the robot can allocate power appropriately based on actual needs, avoiding unnecessary energy waste. Proper cleaning path planning ensures even distribution and effective use of detergent, avoiding the problems of overuse or uneven distribution of detergent that can occur with traditional cleaning methods, thereby reducing detergent usage.

[0070] By continuously acquiring new images while cleaning is being performed and dynamically adjusting the path and intensity, it can respond to sudden contamination (such as spilled liquid) and achieve closed-loop control.

[0071] Through multimodal image preprocessing (noise reduction / contrast enhancement / brightness adjustment), environmental stains are accurately identified, and the multi-dimensional feature fusion of color, texture, and shape and the K-means clustering algorithm are combined to adaptively divide the stain types. A stain intensity quantification model based on thickness, adhesion strength, and composition factors is constructed (I=ɑ*T+β*A+γ*C). The unit cleaning path is dynamically planned and the number of cleaning times is intelligently calculated. At the same time, through the real-time image feedback closed-loop adjustment strategy, the coordinated optimization of stain recognition accuracy, cleaning efficiency and resource utilization is achieved. Its parameter scalability can adapt to the needs of multiple scenarios such as home and industry, solving the pain points of traditional robots such as single cleaning intensity, rigid path, and poor environmental adaptability.

[0072] The step S1 is specifically as follows:

[0073] The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement and brightness adjustment; the noise reduction is based on median filtering or Gaussian filtering; the contrast enhancement is based on histogram equalization or linear contrast stretching; and the brightness adjustment is based on linear transformation or gamma correction.

[0074] By using median filtering / Gaussian filtering for noise reduction, histogram equalization for contrast enhancement, and gamma correction for brightness adjustment, we can eliminate ambient light interference, ensure the robustness of subsequent feature extraction, improve image usability in low-light and high-noise scenarios, and reduce the misjudgment rate.

[0075] During environmental image acquisition, the image may contain some noise due to factors such as sensor noise and ambient light interference. This noise can interfere with the accuracy of feature extraction, so noise reduction is necessary to make the image clearer. Contrast measures the difference in brightness between pixels in an image. In low-contrast images, stains can be difficult to clearly identify. Contrast enhancement algorithms can increase the contrast between stains and the background in the image, making it easier to distinguish them. Different lighting conditions can cause brightness differences in the image. To ensure consistent feature extraction, the brightness of the image needs to be adjusted to an appropriate range.

[0076] In step S2, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

[0077] By combining color (grayscale mean / variance), texture (grayscale co-occurrence matrix / LBP), and shape (contour algorithm) features, the physical and chemical property differences of stains are covered, avoiding missed detection of a single feature (such as transparent oil stains through texture recognition and colored stains through color capture), thereby improving the comprehensiveness of stain detection.

[0078] Stains' color is an important visual feature. By analyzing the color distribution of pixels in an environmental image, we can extract color features related to stains. For example, coffee stains typically appear brown, while mud stains may appear darker. The grayscale mean reflects the overall brightness of the stain, while the grayscale variance reflects the uniformity of the stain's color. Histogram analysis can determine the color distribution of stains. Histogram analysis can also be used to extract color distribution features, such as peak location and distribution width.

[0079] The texture of a stain reflects the microscopic structure of its surface, and different types of stains have different textural characteristics. For example, a scratch typically has a linear texture, while a stain may have a mottled texture.

[0080] Gray-level co-occurrence matrix (GLCM) extracts texture features by analyzing the gray-level co-occurrence relationship of pixel pairs in the environmental image. GLCM can calculate a variety of texture features, such as contrast, correlation, and energy. Contrast reflects the clarity of the texture, correlation reflects the directionality of the texture, and energy reflects the uniformity of the texture.

[0081] Local Binary Pattern (LBP) extracts texture features by analyzing the relationship between each pixel and its neighboring pixels. Since LBP has rotation invariance and grayscale invariance, it is very helpful in describing the local features of texture. The histogram of LBP can be used as a descriptor of texture features for subsequent classification.

[0082] The shape of a stain can provide important information about its origin and nature. For example, a circular stain may be caused by a drop of liquid, while an irregularly shaped stain may be caused by dust accumulation. The spatial distribution of a stain in an image can reveal its location and extent on the surface. For example, a stain may be concentrated in a certain area of ​​the object or evenly distributed. Spatial distribution characteristics can be extracted by analyzing information such as the center position and distribution density of the stain.

[0083] In step S4, the thickness index is obtained by converting the shadow area in the environment image;

[0084] In step S5, the calculation formula of the stain intensity index is:

[0085] I=ɑ*T+β*A+γ*C;

[0086] Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

[0087] By introducing the thickness index (shadow conversion), adhesion strength index (physical adhesion), and component influencing factor (chemical corrosiveness), a comprehensive strength formula I = ɑ*T + β*A + γ*C is constructed to quantify the difficulty of cleaning different stains and avoid over-cleaning (waste of resources) or under-cleaning (residue) caused by fixed strength.

[0088] Dynamic configuration is supported through parameters such as weight coefficients (ɑ / β / γ) and segmentation size to adapt to different scenarios (home / factory).

[0089] The thickness of a stain is an important indicator of its severity. Thicker stains typically require longer cleaning times or stronger cleaning forces to remove. The thickness index can be calculated by measuring the thickness of the stain using optical or tactile sensors. For example, the thickness of the stain can be estimated by analyzing shadows in optical imaging or changes in the pressure distribution of a tactile sensor. The adhesion strength of a stain reflects its degree of adhesion to the surface. Stains with stronger adhesion are more difficult to remove and may require special cleaning agents or stronger mechanical force. The adhesion strength index can be calculated by measuring the force or energy required during the cleaning process. For example, the adhesion strength of a stain can be assessed by analyzing changes in the friction force detected by a tactile sensor during the cleaning process. The composition of the stain also has a significant impact on its cleaning difficulty. Stains of different compositions may require different cleaning methods and cleaning agents. The compositional impact factor can be determined using near-infrared spectroscopy or other chemical analysis methods. For example, oil stains and water stains require different cleaning methods, and the compositional impact factor can reflect this difference.

[0090] In step S6, the calculation formula for the number of cleaning times is:

[0091] times n =cei l(I n / effect n );

[0092] Among them, times n Indicates the number of cleaning times of the nth cleaning unit; ceil() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

[0093] A preferred embodiment of a cleaning robot cleaning control system of the present invention includes the following modules:

[0094] An environmental image acquisition module is used for the cleaning robot to collect environmental images in real time and pre-process the environmental images; the environmental images are mapped to the actual ground position based on a preset conversion relationship;

[0095] A stain feature extraction module is used for the cleaning robot to extract stain features including at least color, texture and shape from the pre-processed environment image, and locate the stain area based on the stain features;

[0096] A stain classification module is used for the cleaning robot to cluster the stain features using a K-means clustering algorithm and then classify the stains into categories;

[0097] Through unsupervised clustering, stain categories are divided to adapt to dynamic scenarios with unknown stain types, reduce dependence on manual labeling, and achieve data-driven self-learning of stain types.

[0098] Stains can also be classified based on thresholds, that is, classification is performed based on a threshold range of features. For example, stains can be classified as coffee stains, mud stains, etc. based on the range of color values. This method is simple and direct, but it requires a pre-determined threshold range. Clustering algorithms, on the other hand, calculate the distance information between different stain samples and classify them into K different categories. This method does not require a pre-determined threshold, but does require selecting an appropriate cluster center and number of categories. Clustering algorithms can automatically discover the distribution patterns of stain samples, thereby achieving stain classification.

[0099] A data matching and identification module is used for the cleaning robot to match the corresponding adhesion strength index and component influencing factor based on the stain category and identify the thickness index of the stain based on the environmental image;

[0100] A stain intensity index calculation module, used for the cleaning robot to calculate the stain intensity index based on the thickness index, adhesion strength index and component influencing factor;

[0101] Quantitative assessment of stains is a key step in achieving efficient cleaning. Simply identifying and classifying stains is not enough; the severity of the stains also needs to be quantified to provide a scientific basis for the subsequent formulation of cleaning strategies. By quantifying stains into a series of numerical values, the cleaning difficulty and priority of the stains can be more intuitively assessed, thereby optimizing cleaning paths and methods. In addition, the quantitative data of stains can also be used to draw stain maps, providing a data basis for dynamic path planning.

[0102] A cleaning path and frequency planning module is configured to divide the environmental image into a plurality of cleaning units based on a preset segmentation size (e.g., into a plurality of 5 mm × 5 mm small square cleaning units corresponding one-to-one to the elements in a predefined two-dimensional array), plan a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, match the corresponding cleaning intensity index based on the stain category, and calculate the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index;

[0103] By dividing the cleaning units by segmentation size, dynamically planning the path based on the overlapping relationship, and calculating the cleaning times timesn=ceil(In / effectn) in combination with the intensity index, high-pollution areas can be processed first and ineffective movement can be reduced.

[0104] In practice, a static path can be planned in a bow-shaped pattern based on the contours of the area to be cleaned, avoiding the problems of repeated cleaning or missed areas that may occur with traditional cleaning methods, thereby completing the cleaning task in the shortest possible time. When encountering an obstacle, the cleaning path (i.e., the dynamic path) is replanned only for the area covered by the static path that intersects the obstacle contour, while the original static path that does not intersect with the obstacle and has not been cleaned is retained. This reduces the number of back-and-forth travel paths of the cleaning robot between different areas and improves cleaning efficiency.

[0105] The cleaning operation module is configured to enable the cleaning robot to perform cleaning operations based on the cleaning path and cleaning frequency, dynamically adjust the cleaning path and cleaning frequency based on the most recently acquired environmental image, and verify the cleaning effect through machine vision. In specific implementations, stains within the cleaning unit can be cleaned step by step from top to bottom and from left to right.

[0106] In practice, the cleaning robot automatically records a map of the environment and determines its own position in real time. In complex indoor environments, such as those with obstacles like furniture and debris, the robot can adjust its cleaning path based on this real-time environmental information to ensure comprehensive coverage of the cleaning area and avoid missing areas due to obstructions. When an obstacle is detected while cleaning along a static path, the robot cleans along its edges to obtain its outline. It then replans a dynamic path for the area covered by the static path that intersects the obstacle outline and does not contain the obstacle. This precise path adjustment ensures the robot effectively cleans the area around the obstacle, improving cleaning quality. By optimizing the cleaning path, the robot reduces ineffective travel and repeated cleaning, thereby reducing energy consumption. During the cleaning process, the robot can allocate power appropriately based on actual needs, avoiding unnecessary energy waste. Proper cleaning path planning ensures even distribution and effective use of detergent, avoiding the problems of overuse or uneven distribution of detergent that can occur with traditional cleaning methods, thereby reducing detergent usage.

[0107] By continuously acquiring new images while cleaning is being performed and dynamically adjusting the path and intensity, it can respond to sudden contamination (such as spilled liquid) and achieve closed-loop control.

[0108] Through multimodal image preprocessing (noise reduction / contrast enhancement / brightness adjustment), environmental stains are accurately identified, and the multi-dimensional feature fusion of color, texture, and shape and the K-means clustering algorithm are combined to adaptively divide the stain types. A stain intensity quantification model based on thickness, adhesion strength, and composition factors is constructed (I=ɑ*T+β*A+γ*C). The unit cleaning path is dynamically planned and the number of cleaning times is intelligently calculated. At the same time, through the real-time image feedback closed-loop adjustment strategy, the coordinated optimization of stain recognition accuracy, cleaning efficiency and resource utilization is achieved. Its parameter scalability can adapt to the needs of multiple scenarios such as home and industry, solving the pain points of traditional robots such as single cleaning intensity, rigid path, and poor environmental adaptability.

[0109] The environmental image acquisition module is specifically used for:

[0110] The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement and brightness adjustment; the noise reduction is based on median filtering or Gaussian filtering; the contrast enhancement is based on histogram equalization or linear contrast stretching; and the brightness adjustment is based on linear transformation or gamma correction.

[0111] By using median filtering / Gaussian filtering for noise reduction, histogram equalization for contrast enhancement, and gamma correction for brightness adjustment, we can eliminate ambient light interference, ensure the robustness of subsequent feature extraction, improve image usability in low-light and high-noise scenarios, and reduce the misjudgment rate.

[0112] During environmental image acquisition, the image may contain some noise due to factors such as sensor noise and ambient light interference. This noise can interfere with the accuracy of feature extraction, so noise reduction is necessary to make the image clearer. Contrast measures the difference in brightness between pixels in an image. In low-contrast images, stains can be difficult to clearly identify. Contrast enhancement algorithms can increase the contrast between stains and the background in the image, making it easier to distinguish them. Different lighting conditions can cause brightness differences in the image. To ensure consistent feature extraction, the brightness of the image needs to be adjusted to an appropriate range.

[0113] In the stain feature extraction module, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

[0114] By combining color (grayscale mean / variance), texture (grayscale co-occurrence matrix / LBP), and shape (contour algorithm) features, the physical and chemical property differences of stains are covered, avoiding missed detection of a single feature (such as transparent oil stains through texture recognition and colored stains through color capture), thereby improving the comprehensiveness of stain detection.

[0115] Stains' color is an important visual feature. By analyzing the color distribution of pixels in an environmental image, we can extract color features related to stains. For example, coffee stains typically appear brown, while mud stains may appear darker. The grayscale mean reflects the overall brightness of the stain, while the grayscale variance reflects the uniformity of the stain's color. Histogram analysis can determine the color distribution of stains. Histogram analysis can also be used to extract color distribution features, such as peak location and distribution width.

[0116] The texture of a stain reflects the microscopic structure of its surface, and different types of stains have different textural characteristics. For example, a scratch typically has a linear texture, while a stain may have a mottled texture.

[0117] Gray-level co-occurrence matrix (GLCM) extracts texture features by analyzing the gray-level co-occurrence relationship of pixel pairs in the environmental image. GLCM can calculate a variety of texture features, such as contrast, correlation, and energy. Contrast reflects the clarity of the texture, correlation reflects the directionality of the texture, and energy reflects the uniformity of the texture.

[0118] Local Binary Pattern (LBP) extracts texture features by analyzing the relationship between each pixel and its neighboring pixels. Since LBP has rotation invariance and grayscale invariance, it is very helpful in describing the local features of texture. The histogram of LBP can be used as a descriptor of texture features for subsequent classification.

[0119] The shape of a stain can provide important information about its origin and nature. For example, a circular stain may be caused by a drop of liquid, while an irregularly shaped stain may be caused by dust accumulation. The spatial distribution of a stain in an image can reveal its location and extent on the surface. For example, a stain may be concentrated in a certain area of ​​the object or evenly distributed. Spatial distribution characteristics can be extracted by analyzing information such as the center position and distribution density of the stain.

[0120] In the data matching and identification module, the thickness index is obtained by converting the shadow area in the environmental image;

[0121] In the stain intensity index calculation module, the calculation formula of the stain intensity index is:

[0122] I=ɑ*T+β*A+γ*C;

[0123] Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

[0124] By introducing the thickness index (shadow conversion), adhesion strength index (physical adhesion), and component influencing factor (chemical corrosiveness), a comprehensive strength formula I = ɑ*T + β*A + γ*C is constructed to quantify the difficulty of cleaning different stains and avoid over-cleaning (waste of resources) or under-cleaning (residue) caused by fixed strength.

[0125] Dynamic configuration is supported through parameters such as weight coefficients (ɑ / β / γ) and segmentation size to adapt to different scenarios (home / factory).

[0126] The thickness of a stain is an important indicator of its severity. Thicker stains typically require longer cleaning times or stronger cleaning forces to remove. The thickness index can be calculated by measuring the thickness of the stain using optical or tactile sensors. For example, the thickness of the stain can be estimated by analyzing shadows in optical imaging or changes in the pressure distribution of a tactile sensor. The adhesion strength of a stain reflects its degree of adhesion to the surface. Stains with stronger adhesion are more difficult to remove and may require special cleaning agents or stronger mechanical force. The adhesion strength index can be calculated by measuring the force or energy required during the cleaning process. For example, the adhesion strength of a stain can be assessed by analyzing changes in the friction force detected by a tactile sensor during the cleaning process. The composition of the stain also has a significant impact on its cleaning difficulty. Stains of different compositions may require different cleaning methods and cleaning agents. The compositional impact factor can be determined using near-infrared spectroscopy or other chemical analysis methods. For example, oil stains and water stains require different cleaning methods, and the compositional impact factor can reflect this difference.

[0127] In the cleaning path and times planning module, the calculation formula for the cleaning times is:

[0128] times n =cei l(I n / effect n );

[0129] Among them, times n Indicates the number of cleaning times of the nth cleaning unit; cei l() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

[0130] In summary, the advantages of the present invention are:

[0131] 1. The cleaning robot collects environmental images in real time and preprocesses them, extracts stain features including at least color, texture and shape from the preprocessed environmental images, locates the stain area based on the stain features, clusters the stain features using the K-means clustering algorithm and then divides the stain categories, matches the corresponding adhesion strength index and component influencing factors based on the stain categories, identifies the thickness index of the stain based on the environmental image, and calculates the stain intensity index based on the thickness index, adhesion strength index and component influencing factors; then the cleaning robot divides the environmental image into several cleaning units based on the preset segmentation size, plans the cleaning path based on the overlapping relationship between each cleaning unit and the stain area, and matches the stain category based on the stain category. It is equipped with a corresponding cleaning intensity index, and the cleaning times of each cleaning unit on the cleaning path are calculated based on the cleaning intensity index and the stain intensity index. Finally, the cleaning operation is performed based on the cleaning path and the cleaning times, and the cleaning path is dynamically adjusted based on the latest collected environmental image; that is, the cleaning robot automatically performs preprocessing, stain feature extraction, cleaning path planning, and cleaning times calculation through the real-time collected environmental image, and dynamically adjusts the cleaning path based on the latest environmental image. It can effectively cope with complex environments or corners, effectively adapt to complex and changing environments, and perform personalized cleaning for different types of stains based on the calculated cleaning times to avoid repeated operations, ultimately greatly improving the cleaning quality, flexibility and efficiency of the cleaning robot.

[0132] 2. By using median filtering / Gaussian filtering for noise reduction, histogram equalization for contrast enhancement, and gamma correction for brightness adjustment, we eliminate ambient light interference, ensure the robustness of subsequent feature extraction, improve image usability in low-light and high-noise scenarios, and reduce the false positive rate.

[0133] 3. By combining color (grayscale mean / variance), texture (grayscale co-occurrence matrix / LBP), and shape (contour algorithm) features, it covers the differences in physical and chemical properties of stains, avoids missed detection of a single feature (such as transparent oil stains through texture recognition and colored stains through color capture), and improves the comprehensiveness of stain detection.

[0134] 4. Classify stains into categories through unsupervised clustering, adapt to dynamic scenarios with unknown stain types, reduce reliance on manual labeling, and achieve data-driven self-learning of stain types.

[0135] 5. By introducing the thickness index (shadow conversion), adhesion strength index (physical adhesion), and component influencing factor (chemical corrosiveness), a comprehensive strength formula I = ɑ*T + β*A + γ*C is constructed to quantify the cleaning difficulty of different stains and avoid over-cleaning (waste of resources) or under-cleaning (residue) caused by fixed strength.

[0136] 6. By dividing the cleaning units by segmentation size, dynamically planning the path based on the overlapping relationship, and combining the intensity index to calculate the number of cleaning timesn=ceil(In / effectn), high-pollution areas can be treated first and ineffective movement can be reduced.

[0137] 7. By continuously acquiring new images while cleaning and dynamically adjusting the path and intensity, it can respond to sudden contamination (such as spilled liquid) and achieve closed-loop control.

[0138] 8. Support dynamic configuration through parameters such as weight coefficient (ɑ / β / γ) and segmentation size to adapt to different scenarios (home / factory).

[0139] 9. Accurately identify environmental stains through multimodal image preprocessing (noise reduction / contrast enhancement / brightness adjustment), combine color, texture, and shape multi-dimensional feature fusion with K-means clustering algorithm to adaptively divide stain types, and build a stain intensity quantification model based on thickness, adhesion strength, and composition factors (I = ɑ*T+β*A+γ*C). Dynamically plan the unit cleaning path and intelligently calculate the number of cleaning times. At the same time, through real-time image feedback closed-loop adjustment strategy, the coordinated optimization of stain recognition accuracy, cleaning efficiency and resource utilization is achieved. Its parameter scalability can adapt to the needs of multiple scenarios such as home and industry, solving the pain points of traditional robots such as single cleaning strength, rigid path, and poor environmental adaptability.

[0140] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A cleaning robot cleaning control method, characterized in that: The steps include: Step S1: The cleaning robot collects environmental images in real time and pre-processes the environmental images; Step S2: The cleaning robot extracts stain features including at least color, texture, and shape from the pre-processed environment image, and locates the stain area based on the stain features; Step S3: The cleaning robot clusters the stain features using a K-means clustering algorithm to further classify the stains. Step S4: The cleaning robot matches the corresponding adhesion strength index and component influencing factor based on the stain category, and identifies the thickness index of the stain based on the environmental image; Step S5: The cleaning robot calculates a stain intensity index based on the thickness index, the adhesion strength index, and the component influencing factor; Step S6: The cleaning robot divides the environment image into a plurality of cleaning units based on a preset segmentation size, plans a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, matches the corresponding cleaning intensity index based on the stain category, and calculates the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index; Step S7: The cleaning robot performs a cleaning operation based on the cleaning path and the number of cleaning times, and dynamically adjusts the cleaning path based on the latest collected environment image.

2. A cleaning robot cleaning control method according to claim 1, characterized in that: The step S1 is specifically as follows: The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement, and brightness adjustment; the noise reduction is based on a median filter or a Gaussian filter; the contrast enhancement is based on a histogram equalization or a linear contrast stretch; The brightness adjustment is based on linear transformation or gamma correction.

3. A cleaning robot cleaning control method according to claim 1, characterized in that: In step S2, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

4. A cleaning robot cleaning control method according to claim 1, characterized in that: In step S4, the thickness index is obtained by converting the shadow area in the environment image; In step S5, the calculation formula of the stain intensity index is: I=ɑ*T+β*A+γ*C; Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

5. The cleaning control method of a cleaning robot according to claim 1, characterized in that: In step S6, the calculation formula for the number of cleaning times is: times n =those l(I n / effect n ); Among them, times n Indicates the number of cleaning times of the nth cleaning unit; ceil() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

6. A cleaning robot cleaning control system, characterized by: Includes the following modules: An environmental image acquisition module is used by the cleaning robot to collect environmental images in real time and pre-process the environmental images; A stain feature extraction module is used for the cleaning robot to extract stain features including at least color, texture and shape from the pre-processed environment image, and locate the stain area based on the stain features; A stain classification module is used for the cleaning robot to cluster the stain features using a K-means clustering algorithm and then classify the stains into categories; A data matching and identification module is used for the cleaning robot to match the corresponding adhesion strength index and component influencing factor based on the stain category and identify the thickness index of the stain based on the environmental image; A stain intensity index calculation module, used for the cleaning robot to calculate the stain intensity index based on the thickness index, adhesion strength index and component influencing factor; a cleaning path and number planning module, configured for the cleaning robot to divide the environment image into a plurality of cleaning units based on a preset segmentation size, plan a cleaning path based on the overlapping relationship between each cleaning unit and the stain area, match the corresponding cleaning intensity index based on the stain category, and calculate the number of cleaning times for each cleaning unit on the cleaning path based on the cleaning intensity index and the stain intensity index; The cleaning operation module is used for the cleaning robot to perform a cleaning operation based on the cleaning path and the number of cleaning times, and dynamically adjust the cleaning path based on the latest collected environment image.

7. A cleaning robot cleaning control system according to claim 6, characterized in that: The environmental image acquisition module is specifically used for: The cleaning robot collects environmental images in real time through a camera and performs preprocessing on the environmental images, including at least noise reduction, contrast enhancement, and brightness adjustment; the noise reduction is based on a median filter or a Gaussian filter; the contrast enhancement is based on a histogram equalization or a linear contrast stretch; The brightness adjustment is based on linear transformation or gamma correction.

8. The cleaning robot cleaning control system according to claim 6, characterized in that: In the stain feature extraction module, the color stain feature is extracted based on grayscale mean, grayscale variance and histogram; the texture stain feature is extracted based on grayscale co-occurrence matrix or local binary pattern; and the shape stain feature is extracted based on contour algorithm.

9. The cleaning robot cleaning control system according to claim 6, characterized in that: In the data matching and identification module, the thickness index is obtained by converting the shadow area in the environmental image; In the stain intensity index calculation module, the calculation formula of the stain intensity index is: I=ɑ*T+β*A+γ*C; Among them, I represents the stain intensity index; T represents the thickness index; A represents the adhesion strength index; C represents the component influence factor; ɑ, β, and γ all represent weight coefficients.

10. The cleaning robot cleaning control system according to claim 6, characterized in that: In the cleaning path and times planning module, the calculation formula for the cleaning times is: times n =those l(I n / effect n ); Among them, times n Indicates the number of cleaning times of the nth cleaning unit; cei l() indicates the upper bound function; I n Indicates the stain intensity index of the nth cleaning unit; effect n Represents the cleaning intensity index of the nth cleaning unit.

Images