Intelligent ash removal method of fully-closed driving type sweeper

Abstract

The invention relates to the technical field of image recognition, and provides an intelligent ash removal method of a fully-closed driving type sweeper. The method comprises the following steps: acquiring a smudginess image of an area to be dedusted; performing visual saliency identification analysis on the dirty image to obtain a dirty block; performing dry-wet discriminant analysis on the multi-dimensional image feature information of the dirty block to obtain a discriminant result; when the judgment result shows dry garbage pollution, introducing a predetermined pollution evaluation interference function to perform interference influence analysis on the initial pollution index to obtain an effective pollution index; and matching a corresponding ash removal plan based on the effective dirt index, and executing an ash removal operation by using the target sweeper. According to the method and the device, the technical problems of poor cleaning strategy adaptability and serious resource waste caused by difficulty in accurately distinguishing the dirt type and quantifying the pollution degree based on a single image recognition technology in a dry and wet mixed garbage scene of the existing cleaning equipment are solved, and the technical effect of improving the matching precision and the execution efficiency of the cleaning plan is achieved.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and specifically relates to an intelligent dust cleaning method for a fully enclosed driving type sweeper Background Art

[0002] In closed or semi-closed scenarios such as industrial plants, warehousing and logistics centers, and large parking lots, the ground cleaning work faces complex environmental challenges: on the one hand, wet and dry mixed garbage is widespread in the operation area (such as a mixture of metal debris and oil stains, and an interweaving of dust and condensate). Traditional sweepers rely on single sensors or basic image recognition technology and it is difficult to accurately distinguish the types of pollutants; on the other hand, environmental light changes, ground material differences, and temperature and humidity fluctuations easily lead to dirt recognition errors, directly affecting the adaptability of cleaning strategies, and causing problems such as waste of water resources or incomplete cleaning. Existing solutions mostly adopt fixed cleaning modes or judge pollutants based on simple gray-scale thresholds, which cannot achieve intelligent decision-making in dynamic scenarios and are prone to misjudgment when dealing with complex situations such as reflective floors and shadow interference.

[0003] In this context, there is an urgent need for a cleaning decision-making system that integrates high-precision visual perception and environmental interference correction, and improves the scene adaptability and resource utilization rate of fully enclosed driving type sweepers through multi-dimensional feature fusion and dynamic evaluation mechanisms. Summary of the Invention

[0004] This application provides an intelligent dust cleaning method for a fully enclosed driving type sweeper, aiming to solve the technical problems that existing cleaning equipment is difficult to accurately distinguish the types of dirt and quantify the degree of pollution based on single image recognition technology in the scenario of wet and dry mixed garbage, resulting in poor adaptability of cleaning strategies and serious waste of resources, and achieving the technical effect of accurately positioning dirty blocks and intelligently distinguishing wet and dry through the combination of visual saliency recognition and multi-dimensional image feature analysis, and dynamically correcting the dirt index based on an interference function, thereby improving the matching accuracy and execution efficiency of cleaning plans.

[0005] This application provides an intelligent dust cleaning method for a fully enclosed driving type sweeper. The method includes: dynamically collecting dirty images of the area to be dust-cleaned through an image collector mounted on the target sweeper; performing visual saliency recognition analysis on the dirty images to obtain dirty blocks; performing wet and dry discrimination analysis on the multi-dimensional image feature information of the dirty blocks to obtain a discrimination result; when the discrimination result shows that the area to be dust-cleaned belongs to dry garbage pollution, introducing a predetermined dirt evaluation interference function to perform interference impact analysis on the initial dirt index to obtain an effective dirt index; matching a corresponding dust cleaning plan based on the effective dirt index, and according to the dust cleaning plan, using the target sweeper to perform dust cleaning execution on the area to be dust-cleaned.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] For the intelligent dust cleaning method of the above fully enclosed ride-on sweeper, the method first uses an on-vehicle image collector to capture the dirty images of the area to be dust cleaned in real time, providing a data basis for subsequent analysis. Subsequently, visual saliency recognition is performed on the captured dirty images to extract significant dirty blocks, ensuring that the polluted areas are focused on. Then, combined with multi-dimensional image feature information, wet and dry discrimination is performed on the dirty blocks to clarify the pollution type. If the discrimination result is dry garbage pollution, a predetermined dirty evaluation interference function is introduced to comprehensively consider environmental factors and correct the initial dirty index to obtain a more accurate effective dirty index. Finally, according to the effective dirty index, the corresponding dust cleaning plan is matched to guide the sweeper to perform targeted cleaning operations to ensure the best cleaning effect. This method improves the adaptability and cleaning efficiency of the sweeper in complex environments through intelligent analysis and dynamic decision-making.

[0008] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the specific embodiments of this application are specifically given below. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0010] Figure 1 It is a schematic flow chart of an intelligent dust cleaning method for a fully enclosed ride-on sweeper in an embodiment.

[0011] Figure 2 It is a schematic flow chart of obtaining dirty blocks for an intelligent dust cleaning method of a fully enclosed ride-on sweeper in an embodiment. Detailed Embodiments

[0012] By providing an intelligent dust cleaning method for a fully enclosed ride-on sweeper in the embodiments of this application, the technical problem that existing cleaning equipment is difficult to accurately distinguish the type of dirt and quantify the degree of pollution based on a single image recognition technology in the scenario of wet and dry mixed garbage, resulting in poor adaptability of the cleaning strategy and serious waste of resources, is solved. The technical effect of accurately positioning the dirty blocks and intelligently discriminating wet and dry through the combination of visual saliency recognition and multi-dimensional image feature analysis, and dynamically correcting the dirty index based on the interference function, so as to improve the matching accuracy and execution efficiency of the cleaning plan is achieved.

[0013] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0014] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0015] Embodiment, as Figure 1 shown, the present application provides an intelligent dust cleaning method for a fully enclosed driver-operated sweeper, and the method includes:

[0016] Dynamically collect the dirty images of the area to be dust-cleaned through an image collector mounted on the target sweeper.

[0017] In the embodiments of the present application, first, an image collector (such as a high-definition camera or an infrared sensor) installed on the body of the target sweeper is used to capture the ground images of the area to be dust-cleaned in real time. These image collectors can cover all areas on the travel path of the target sweeper in the area to be dust-cleaned in a dynamic scanning manner to ensure dead-angle-free monitoring. This target sweeper is an automatic cleaning device integrating functions of dust collection, sweeping, and water spraying, designed for outdoor environments and suitable for ground cleaning work in various public places. Its application scenarios include outdoor streets, roads, residential property management, indoor and outdoor factory floors, the periphery of large shopping malls, railway stations, bus stations, large property squares, and the periphery of airports, etc. This device has strong dust collection ability, the sweeping width can reach 1900 mm, is equipped with internationally advanced sweeping and throwing technology, and the theoretical utilization rate of the dust box can reach 100%. The images collected by the image collector carried by the target sweeper not only contain visible dirt on the ground (such as dust, oil stains, debris, etc.), but also can capture complex visual information caused by differences in light, shadow, or ground materials. Through this step, high-resolution image data of the area to be dust-cleaned is obtained, and these image data are stored in the dirty images, providing a basis for subsequent dirt identification and analysis. This dynamic collection method can adapt to the movement requirements of the sweeper in different environments, ensuring the real-time and comprehensiveness of the cleaning process.

[0018] Perform visual saliency recognition and analysis on the dirty images to obtain dirty blocks.

[0019] In one embodiment, after obtaining the dirty image, visual saliency recognition analysis is performed on the image to accurately extract the dirty blocks. This process first divides the entire dirty image into multiple small blocks to form a block set, and then extracts features from each block in the block set based on a preset visual saliency index to obtain relevant parameter information. Subsequently, in combination with the correspondence between the blocks and the parameters, a visual parameter map is generated. By analyzing the differences between the center and the surrounding areas of the parameter map, a saliency map for each block is generated. Finally, by superimposing and fusing these saliency maps, a complete dirty saliency map is obtained, and based on this map, the dirty blocks in the image are determined. This process realizes the precise positioning and recognition of the dirty areas through multi-level feature extraction and difference analysis, providing accurate input for subsequent wet-dry discrimination and cleaning decision-making.

[0020] Further, as Figure 2 shown, the present application provides visual saliency recognition analysis of the dirty image to obtain dirty blocks, including:

[0021] Extracting the first saliency index in the predefined visual saliency index; performing block processing on the dirty image to obtain a dirty block set, where the dirty block set includes a first block; collecting features of the first block based on the first saliency index to obtain a first index parameter; constructing a first parameter viewable map of the first saliency index in combination with the first correspondence between the first block and the first index parameter; performing central-peripheral difference analysis on the first parameter viewable map to obtain a first saliency map; performing superimposing and fusing on the first saliency map to obtain the dirty saliency map of the dirty image; and determining the dirty blocks according to the dirty saliency map.

[0022] Preferably, when performing visual saliency recognition of the dirty image, first, the first visual index is selected as the first saliency index from the predefined visual saliency indexes (such as brightness, color, direction, etc.), and this first saliency index will be used as the basis for subsequent analysis. Subsequently, the entire dirty image is divided into multiple small blocks, such as the first block, the second block, etc., according to the preset segmentation unit length, where the segmentation unit length is determined by the image resolution and actual application requirements. By adding these segmented blocks to a set one by one, a dirty block set is formed. Each block in the dirty block set is a part of the image, and this block processing helps to perform localized analysis on the image and improve the accuracy of dirty recognition. After that, based on the first saliency index, features of the first block are extracted to obtain a first index parameter. For example, if the first saliency index is brightness, the RGB values of each pixel in the first block are obtained, and by inputting the RGB values into the brightness calculation formula, the brightness value of each pixel is obtained. The brightness calculation formula is specifically as follows: L = wR ·R + w G ·G + w B ·B; where L represents the luminance value, R, G, and B respectively represent the red channel value, green channel value, and blue channel value, and w R , w G , w B represent the weights of the red, green, and blue channels, which are 0.299, 0.587, and 0.114 respectively. The calculated luminance values of these pixels will be used as the first index parameter of the first saliency metric; if the first saliency metric is color, the RGB values of each pixel in the first block will be directly stored as the first index parameter; if the first saliency metric is direction, the Sobel operator will be used to calculate the gradient direction of each pixel, and the calculated gradient direction will be stored as the first index parameter. After obtaining the first index parameter, combining the correspondence between the first block and the first index parameter, that is, mapping each parameter in the first index parameter to the corresponding pixel in the first block, thereby generating a first parameter visualization map, which visually shows the distribution of the visual features of the block and provides visual support for subsequent analysis. Then, with each pixel as the center, determine the central region and surrounding region of each pixel according to the pre-set central region range (such as 3*3, 5*5, etc.) and surrounding region range (such as 7*7, 9*9, etc.), and then calculate the mean value of the first index parameter in each determined region to obtain the mean value of the first index parameter of each region. By calculating the absolute difference between the mean value of the first index parameter of the central region and the mean value of the first index parameter of the surrounding region of each pixel, and then performing normalization processing by the maximum-minimum method, the saliency of each pixel is obtained, and these saliencies will jointly form the first saliency map, indicating the saliency of each pixel in the image. Finally, superimpose and fuse the first saliency map with other saliency maps (such as maps generated based on other visual features), that is, for each pixel position, use the weight of each metric to perform weighted calculation on the saliency of each metric, thereby generating a comprehensive dirt saliency map. This map synthesizes the analysis results of multiple visual features and can more comprehensively reflect the dirty areas in the image. Then, according to this dirt saliency map, by setting a threshold, accurately locate the dirty blocks, providing a basis for subsequent processing. This process realizes the accurate recognition of dirty images through steps such as extracting visual features, block processing, feature collection, visual analysis, difference calculation, map fusion, and block location.

[0023] Furthermore, the present application provides that the predetermined visual saliency metric at least includes a visual luminance metric, a visual color metric, and a visual direction metric.

[0024] Optionally, the predetermined visual saliency metrics at least include a visual brightness metric, a visual color metric, and a visual orientation metric. Among them, the visual brightness metric reflects the light and dark changes in different regions of the image, and the dirty areas usually have obvious differences in brightness from the surrounding ground; the visual color metric reflects the color differences in different regions of the image, and the dirty areas usually have obvious differences in color from the surrounding ground; the visual orientation metric reflects the directionality of the texture or edges in different regions of the image, and the dirty areas usually have obvious differences in texture or edges from the surrounding ground. By analyzing these metrics, dirty areas such as stains, dust, and oil stains on the ground can be efficiently identified, thereby improving the cleaning efficiency and quality.

[0025] Further, the present application provides for determining the dirty block according to the dirty saliency map, including:

[0026] Obtain any block and denote the any block as the central block; form a set of surrounding blocks of the central block and randomly extract any surrounding block from the set of surrounding blocks; compare to obtain the parameter difference between the central block and the any surrounding block; when the parameter difference is within a predetermined parameter difference threshold, add the any block to the dirty block.

[0027] Optionally, after obtaining the dirty saliency map, a block is randomly selected from the dirty saliency map as the any block, and this any block will be denoted as the central block. Then, the blocks adjacent to the central block are extracted to form a set of surrounding blocks. Subsequently, a block is randomly selected from the set of surrounding blocks and denoted as the any surrounding block. The saliency mean of the central block and the saliency mean of the any surrounding block are calculated, and then the absolute difference between the two means is calculated. If the calculated absolute difference is greater than or equal to the predetermined parameter difference threshold, it indicates that there is a significant difference between the central block and the surrounding block. At this time, the any block corresponding to the central block will be added to the dirty block. Repeat the above process until all blocks in the dirty saliency map have participated in the comparison as the central block. In summary, by comparing the parameter differences between the central block and the surrounding blocks, the dirty areas on the ground can be effectively detected, improving the cleaning accuracy and efficiency of the sweeping robot.

[0028] Perform wet-dry discrimination analysis on the multi-dimensional image feature information of the dirty block to obtain a discrimination result.

[0029] In one embodiment, after obtaining the dirty block, the image texture feature is extracted from the multi-dimensional image features of the dirty block, which is a key index for judging the nature of the dirt. Subsequently, the preset texture limit value is used as the classification standard for dry and wet garbage. By comparing the extracted image texture feature with the texture limit value, it is determined whether the current pollution is dry garbage pollution or wet garbage pollution, and a discrimination result is generated according to the determined category. This process realizes the automatic and intelligent classification of the nature of the dirt by quantifying the image texture feature and combining the preset discrimination criteria, providing an accurate basis for the subsequent ash cleaning operation.

[0030] Furthermore, the present application provides a dry-wet discrimination analysis of the multi-dimensional image feature information of the dirty block to obtain a discrimination result, including:

[0031] Extracting the image texture feature value from the multi-dimensional image feature information; reading the preset texture limit value, and obtaining the discrimination result by combining the preset texture limit value and the image texture feature value; wherein, reading the preset texture limit value and obtaining the discrimination result by combining the preset texture limit value and the image texture feature value includes: when the image texture feature value is greater than the preset texture limit value, the area to be ash-cleaned belongs to dry garbage pollution, and when the image texture feature value is less than the preset texture limit value, the area to be ash-cleaned belongs to wet garbage pollution.

[0032] Preferably, when performing dry-wet discrimination, first extract the image texture feature value from the multi-dimensional image feature information (such as brightness, color, direction, etc.) of the dirty block, such as the gradient magnitude (calculated by the Sobel operator). Subsequently, read the preset texture limit value from the preset database or configuration file, such as the gradient magnitude limit value, which is set based on historical experience and expert decision-making and is a texture feature threshold for distinguishing dry and wet garbage. Then, compare the extracted image texture feature value with the preset texture limit value. If the image texture feature value is greater than the preset texture limit value, it indicates that the dirt in this area has a high granularity, so it is determined as dry garbage pollution. If the image texture feature value is less than the preset texture limit value, it indicates that the dirt in this area has a low granularity, so it is determined as wet garbage pollution. Through this process, the nature of the dirt in the area to be ash-cleaned can be quickly and accurately judged, providing a scientific basis for the subsequent ash cleaning operation.

[0033] When the discrimination result shows that the area to be ash-cleaned belongs to dry garbage pollution, a preset dirty evaluation interference function is introduced to analyze the interference effect on the initial dirt index, and an effective dirt index is obtained.

[0034] In one embodiment, when it is determined through feature analysis that the area to be cleaned of ash belongs to dry garbage pollution, a predetermined dirt evaluation interference function is introduced to correct the initial dirt index, which is determined based on multi-dimensional image feature information and the regional material of the area to be cleaned of ash and is used to preliminarily evaluate the degree of dirt. Subsequently, the initial dirt index is input into the predetermined dirt evaluation interference function for dirt index correction to balance the influence of light intensity, light direction, environmental temperature, etc. on the evaluation of the pollution degree, and an effective dirt index is obtained, so as to better support the evaluation of dry garbage pollution and the cleaning decision-making.

[0035] Furthermore, when the discrimination result shows that the area to be cleaned of ash belongs to dry garbage pollution, the present application provides an analysis of the interference influence of a predetermined dirt evaluation interference function on the initial dirt index to obtain an effective dirt index, including:

[0036] Extracting the image hue feature value from the multi-dimensional image feature information; performing weighted analysis on the image hue feature value and the image texture feature value to obtain an image feature value; obtaining the regional material of the area to be cleaned of ash and matching the material image feature value of the regional material; comparing the image feature value with the material image feature value to obtain a comparison feature difference value, and recording the comparison feature difference value as the initial dirt index.

[0037] Preferably, color features are extracted from the multi-dimensional image feature information, and the color features are converted from the RGB color space to the HSV color space to separate the hue, saturation, and value information. Among them, the color space conversion is carried out through mathematical formulas, including steps such as normalizing the RGB values, calculating the maximum and minimum values, and calculating the hue. Subsequently, the calculated hue value is used as the image hue feature value, and the image hue feature value and the image texture feature value are normalized by the maximum-minimum value method to make them in the same dimension. Then, the image hue feature value and the image texture feature value are weighted and calculated according to the weights determined based on actual needs and expert decisions to obtain an image feature value. After that, the regional image of the area to be cleaned of ash is input into the material classification channel for annotation to obtain the material type of the area to be cleaned of ash, such as cement, asphalt, metal, wood, etc. Then, from the predefined material feature library, the standard hue feature value and texture feature value of the material are obtained, and the standard hue feature value and texture feature value are weighted and calculated through the above-determined weights to obtain the material image feature value. Then, by comparing the image feature value and the material image feature value, the difference between the image feature value and the material image feature value is calculated, and this difference value is used as the initial dirt index, which reflects the difference between the area to be cleaned of ash and the standard material and provides a quantitative basis for subsequent dirt evaluation and cleaning decision-making.

[0038] For the material classification channel, first collect the image data of the sample area. This sample area image data has regional material identification. By dividing the sample area image data, a training set and a validation set are obtained. Subsequently, a convolutional neural network (CNN) is used to construct the material classification channel, including an input layer, a convolutional layer, an activation function, a pooling layer, a fully connected layer, and an output layer. Then, use a random initialization method (such as Xavier initialization or He initialization) to assign initial values to the weights of the convolutional layer and the fully connected layer, and initialize the bias term to 0 or a small random number. After that, input the regional images of the training set into the CNN, and successively pass through the convolutional layer, the activation function, the pooling layer, the fully connected layer, and the output layer to calculate the predicted probability of each category. Then, use the cross-entropy loss function to calculate the difference between the prediction result and the true label. Then, calculate the gradient of the loss with respect to the weights layer by layer through the backpropagation algorithm, and use the Adam optimizer to update the weights to minimize the value of the loss function. Repeat the processes of forward propagation, loss calculation, backpropagation, and weight update until the maximum number of iterations is reached or the loss value converges. Finally, use the validation set to test the channel performance and calculate the accuracy of material classification. If the accuracy meets the expected requirements, save the parameters of the current material classification channel. If the accuracy does not meet the expected requirements, adjust the hyperparameters such as the learning rate, batch size, and network depth, and retrain the model.

[0039] Furthermore, the present application provides the expression of the predetermined dirt evaluation interference function as follows:

[0040] where, I eff refers to the effective dirt index, I0 refers to the initial dirt index, F interf refers to the dirt evaluation interference coefficient, μ i is the weight of the i-th interference factor in the predetermined interference factor set, satisfying f i is the interference coefficient of the i-th interference factor, and its value range is from 0 to 1, indicating the degree of interference of this factor on the dirt index.

[0041] Preferably, the predetermined dirt evaluation interference function is used to adjust the initial dirt index to obtain the effective dirt index, so as to more accurately evaluate the degree of dirt that the sweeping robot needs to clean. The specific form of the predetermined dirt evaluation interference function is as follows: where, I eff refers to the effective dirt index, which is the dirt index finally used for the dust cleaning decision, and is the value after being corrected by the interference factors; I0 refers to the initial dirt index, which is the originally calculated dirt index and has not been corrected by the environmental factors; F interf refers to the dirt evaluation interference coefficient, representing the comprehensive influence degree of all environmental interference factors; μ iis the weight of the i-th interference factor in the set of predetermined interference factors, representing the importance of this factor relative to other factors, and the sum of all weights is 1; f i is the interference coefficient of the i-th interference factor, with a value range of 0 to 1, representing the degree of interference of this factor on the dirt index, and is obtained by calculating the ratio of the value of the interference factor to the standard value; n is the number of all interference factors, for example, light, temperature, humidity, etc. The size of the effective dirt index is determined by the weights and influence degrees of environmental factors. If some environmental factors (such as light intensity, humidity, etc.) have a greater impact on the dirt index, it will reduce the effective dirt index and make the cleaning strategy more accurate.

[0042] Further, the present application provides that the set of predetermined interference factors includes light intensity, light direction, environmental temperature, environmental humidity, and environmental pollution.

[0043] Optionally, the set of predetermined interference factors at least includes light intensity, light direction, environmental temperature, environmental humidity, and environmental pollution. These factors will all have a significant impact on the dirt assessment result. Among them, light intensity refers to the brightness of the light irradiating on the object surface. Strong light may make the dirt more obvious, while weak light may cover up some dirt, and it can be directly measured by a light sensor (such as a photometer); light direction refers to the angle at which the light irradiates on the object surface. Different light directions will affect the visual effect of the dirt. For example, it will produce shadows or reflections, and the light direction can be directly measured by a light direction sensor or a multi-angle light measurement device; environmental temperature refers to the temperature of the surrounding air. Temperature changes will affect the physical state of the stains. For example, high temperature makes the grease spread, and low temperature makes the stains solidify, and it can be directly measured by a temperature sensor; environmental humidity refers to the content of water vapor in the air. Humidity changes will affect the adhesion and cleaning difficulty of the stains. For example, high humidity makes the dust easier to adhere, and it can be directly measured by a humidity sensor; environmental pollution refers to the concentration of pollutants existing in the air. The degree of pollution directly affects the accumulation speed and severity of the dirt, and it can be directly obtained through an environmental monitoring station or an air quality monitoring network. By comprehensively considering these interference factors and their weights and coefficients, the dirt degree can be evaluated more accurately, so as to formulate a more effective cleaning strategy.

[0044] Further, the present application provides that when the discrimination result shows that the area to be cleaned of ash belongs to wet waste pollution, the water-oil ratio is added to the set of predetermined interference factors.

[0045] Optionally, when the discrimination result shows that the area to be cleaned of ash belongs to wet waste pollution, the characteristics of wet waste (such as moisture and grease content) will have a significant impact on the dirt assessment and cleaning strategy. Therefore, the water-oil ratio needs to be added to the set of predetermined interference factors to more accurately evaluate the degree of dirt and formulate a cleaning plan. The water-oil ratio refers to the relative content of moisture and grease in wet waste. A high moisture content may cause stains to spread more easily and increase the cleaning difficulty. Especially when moisture is mixed with dust or other pollutants, a high grease content may make the stains more difficult to remove. Especially in a low-temperature environment, grease is easy to solidify and form stubborn stains, which can be determined by the saturation distribution of the image, that is, comparing the saturation of each pixel with the water-oil discrimination threshold to determine the pixel ratio of the low-saturation area (presumed water stain) to the high-saturation area (presumed oil stain), and this pixel ratio is the water-oil ratio.

[0046] Match the corresponding ash cleaning plan based on the effective dirt index, and according to the ash cleaning plan, use the target floor sweeper to perform ash cleaning in the area to be cleaned of ash.

[0047] In one embodiment, after obtaining the effective dirt index, a corresponding ash cleaning plan will be matched according to this index. Among them, the ash cleaning plan is different cleaning strategies divided according to the effective dirt index, usually divided into multiple levels. For example, for mild pollution (I eff ≤0.3), simple cleaning, using the ordinary cleaning mode; for moderate pollution (0.3 < I eff ≤0.6), deep cleaning, using the enhanced cleaning mode, and local key cleaning may be required; for severe pollution (I eff > 0.6), comprehensive cleaning, using the high-intensity cleaning mode, and multiple cleanings or special cleaning agents may be required. Subsequently, according to the matched ash cleaning plan, the target floor sweeper is adjusted to the corresponding working mode for cleaning, ensuring that the floor sweeper can select the most appropriate ash cleaning method according to the type and degree of pollution, and improving the cleaning efficiency and effect.

[0048] In summary, the embodiments of the present application at least have the following technical effects:

[0049] In the embodiments of the present application, first, a dirty image of the area to be cleaned is dynamically collected by an image collector mounted on a target floor sweeper; the dirty image is subjected to visual saliency recognition and analysis to obtain dirty blocks; multi-dimensional image feature information of the dirty blocks is subjected to dry-wet discrimination analysis to obtain a discrimination result; when the discrimination result shows that the area to be cleaned is contaminated by dry garbage, a predetermined dirty evaluation interference function is introduced to perform interference impact analysis on the initial dirty index to obtain an effective dirty index; based on the effective dirty index, a corresponding cleaning plan is matched, and according to the cleaning plan, the target floor sweeper is used to perform the cleaning execution of the area to be cleaned. These technical effects jointly solve the technical problem that in the scenario of dry-wet mixed garbage, it is difficult to accurately distinguish the types of dirt and quantify the degree of contamination based on a single image recognition technology for existing cleaning equipment, resulting in poor adaptability of cleaning strategies and serious waste of resources, and achieve the technical effect of combining visual saliency recognition with multi-dimensional image feature analysis to realize accurate positioning of dirty blocks and intelligent dry-wet discrimination, and dynamically correcting the dirty index based on the interference function, thereby improving the matching accuracy and execution efficiency of the cleaning plan.

[0050] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired result. In some embodiments, multi-tasking and parallel processing are also possible or may be advantageous.

[0051] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0052] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. An intelligent dust cleaning method for a fully enclosed driving sweeper, characterized in that: include: The dirty image of the area to be cleaned is obtained by dynamically collecting the dirty image of the area to be cleaned through the image collector mounted on the target sweeper; Performing visual saliency recognition analysis on the dirty image to obtain a dirty block; Performing dry-wet discrimination analysis on the multi-dimensional image feature information of the dirty block to obtain a discrimination result; When the discrimination result shows that the area to be cleaned is contaminated by dry garbage, a predetermined contamination assessment interference function is introduced to perform interference influence analysis on the initial contamination index to obtain an effective contamination index; A corresponding cleaning plan is matched based on the effective dirtiness index, and according to the cleaning plan, the target sweeper is used to clean the area to be cleaned.

2. According to claim 1, the intelligent dust cleaning method of a fully enclosed driving sweeper is characterized in that: Perform visual saliency recognition analysis on the dirty image to obtain dirty blocks, including: Extracting a first saliency index from among predetermined visual saliency indexes; Performing block processing on the dirty image to obtain a dirty block set, wherein the dirty block set includes a first block; Collecting features of the first block based on the first significance indicator to obtain a first indicator parameter; constructing a first parameter visual graph of the first significance indicator in combination with a first corresponding relationship between the first block and the first indicator parameter; Performing a center-periphery difference analysis on the first parameter visual map to obtain a first significance map; Superimposing and fusing the first saliency maps to obtain a dirt saliency map of the dirt image; The dirty block is determined according to the dirt saliency map.

3. According to claim 2, the intelligent dust cleaning method of a fully enclosed driving sweeper is characterized in that: The predetermined visual saliency index includes at least a visual brightness index, a visual color index and a visual direction index.

4. According to claim 2, the intelligent dust cleaning method for a fully enclosed driving sweeper is characterized in that: Determining the dirty block according to the dirt saliency map includes: Obtain an arbitrary block, and record the arbitrary block as a central block; Forming a surrounding block set of the central block, and randomly extracting any surrounding block from the surrounding block set; Comparing and obtaining the parameter difference between the central block and any surrounding blocks; When the parameter difference value is within a predetermined parameter difference threshold, the arbitrary block is added to the dirty block.

5. According to claim 1, the intelligent dust cleaning method for a fully enclosed driving sweeper is characterized in that: Performing dry-wet discrimination analysis on the multi-dimensional image feature information of the dirty block to obtain a discrimination result, including: Extracting image texture feature values ​​from the multi-dimensional image feature information; Reading a predetermined texture limit value, and combining the predetermined texture limit value with the image texture feature value to obtain the discrimination result; Among them, the predetermined texture limit is read, and the judgment result is obtained by combining the predetermined texture limit and the image texture feature value, including: when the image texture feature value is greater than the predetermined texture limit, the area to be cleaned is contaminated by dry garbage; when the image texture feature value is less than the predetermined texture limit, the area to be cleaned is contaminated by wet garbage.

6. According to claim 5, the intelligent dust cleaning method for a fully enclosed driving sweeper is characterized in that: When the discrimination result shows that the area to be cleaned is contaminated by dry garbage, a predetermined contamination assessment interference function is introduced to perform interference impact analysis on the initial contamination index to obtain an effective contamination index, including: Extracting image tone feature values ​​from the multi-dimensional image feature information; Performing weighted analysis on the image tone feature value and the image texture feature value to obtain an image feature value; Acquire the regional material of the area to be cleaned, and match the material image feature value of the regional material; The image feature value is compared with the material image feature value to obtain a comparison feature difference value, and the comparison feature difference value is recorded as the initial dirtiness index.

7. According to claim 1, the intelligent dust cleaning method of a fully enclosed driving sweeper is characterized in that: The expression of the predetermined dirt assessment interference function is as follows: Among them, I eff refers to the effective dirtiness index, I0 refers to the initial dirtiness index, F interf Refers to the dirt assessment interference coefficient, μ i is the weight of the i-th interference factor in the predetermined interference factor set, satisfying f i is the interference coefficient of the i-th interference factor, ranging from 0 to 1, indicating the degree of interference of this factor on the dirtiness index.

8. The intelligent dust cleaning method for a fully enclosed driving sweeper according to claim 7, characterized in that: The predetermined interference factor set includes light intensity, light direction, ambient temperature, ambient humidity and ambient pollution.

9. The intelligent dust cleaning method for a fully enclosed driving sweeper according to claim 8, characterized in that: When the discrimination result shows that the area to be cleaned is contaminated by wet garbage, the water-oil ratio is added to the predetermined interference factor set.

Images