Multifunctional garbage transfer and classification integrated management method and system

Abstract

The invention relates to the technical field of image reinforcement learning, in particular to a multifunctional garbage transfer classification integrated management method and system, and the method comprises the steps: analyzing the similarity of color components between adjacent pixel points in each image block, evaluating the similarity and the similarity coefficient of the color component probability distribution condition between each image block and the junk image of the color channel where the image block is located; analyzing the difference between the similarity coefficients of all the reference blocks in the suspected shielding direction of each image block and the similarity coefficients of all the reference blocks in other directions, analyzing the distances from triples of all the reference blocks in the suspected shielding direction to the fitting straight line, and determining a color loss value by combining the color difference; and based on the similarity coefficient and the color difference, garbage is recognized and classified. According to the invention, optimization adjustment is carried out through the preprocessing stage of the garbage image intelligent identification process, so that the garbage identification and classification accuracy and the garbage transfer and classification integrated management efficiency are improved.

Description

Technical Field

[0001] The present application relates to the field of image enhancement learning technology, and in particular to a multifunctional garbage transfer and classification integrated management method and system. Background Art

[0002] As an effective measure, garbage classification is of great significance to ecological environmental protection and economic sustainable development. However, the current total amount of garbage continues to be at a high level, and garbage management is facing tremendous pressure, which urgently needs to be alleviated by practical and effective actions. The integration of multifunctional garbage transfer and classification aims to integrate multiple garbage disposal functions into a single garbage recycling facility or system. With this model, efficient classification, full-type transfer and one-stop disposal of domestic garbage can be achieved, thereby significantly improving the overall efficiency and convenience of garbage disposal.

[0003] At present, garbage sorting mostly relies on residents to do it on their own. However, this method exposes many problems. It is not only inefficient and costly, but also very easy to cause misclassification. In view of this, it is of great practical significance and urgency to explore efficient and intelligent garbage sorting and treatment solutions. In order to deal with the above problems, some areas have introduced intelligent garbage rooms with classification functions for secondary classification and treatment of garbage. The intelligent garbage room integrates a series of intelligent functions such as automatic bag breaking, AI automatic recognition based on deep learning algorithms, automatic sorting, automatic barrel replacement and automatic transmission. However, in the actual operation process, it was found that the accuracy of AI automatic recognition plays a decisive role in the efficiency of garbage sorting and treatment. Due to the variable light conditions in the working environment of the intelligent garbage room, and the possible obstruction of internal mechanical equipment, these interference factors will inevitably introduce noise, resulting in uneven light intensity in the image to be identified. This situation will directly affect the recognition accuracy of the deep learning algorithm for classified garbage, reduce the accuracy of garbage recognition and classification, and reduce the efficiency of integrated management of garbage transfer and classification. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a multifunctional garbage transfer and classification integrated management method and system. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a multifunctional garbage transfer and classification integrated management method, the method comprising the following steps:

[0006] S1: Get the garbage image on each color channel of RGB, and record the component of the pixel value of each pixel in each color channel as the color component of each pixel;

[0007] S2: According to the distribution characteristics of the color components of the pixels in the garbage image under each color channel, the garbage image is enhanced. The specific process is as follows:

[0008] S201: Divide the garbage image of each color channel into multiple image blocks, analyze the similarity of color components between adjacent pixels in each image block under each color channel, determine a first similarity of each image block under each color channel, evaluate the similarity of color component probability distribution between each image block and the garbage image of the color channel where it is located, determine a second similarity of each image block under each color channel, and determine a similarity coefficient of each image block under each color channel in combination with the first similarity;

[0009] S202: In the garbage image of each color channel, each image block is extended in different directions from the starting point, a preset number of image blocks in each direction of each image block are used as reference blocks in each direction of each image block, and the similarity coefficients of all reference blocks in each direction of each image block are combined to screen out suspected occlusion directions from all directions; the difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and those in the remaining directions is analyzed to determine the color difference of each image block in each color channel;

[0010] S203: The coordinates and similarity coefficients of each image block are combined into a triplet, and the triplet of each image block and all reference blocks in the suspected occlusion direction are fitted to obtain a fitting straight line, and the distance from the triplet of all reference blocks in the suspected occlusion direction to the fitting straight line is analyzed, and the color loss value of each image block in each color channel is determined in combination with the color difference;

[0011] S204: analyzing the difference in color loss value between each image block and all other image blocks in the garbage image of each color channel, and performing enhancement processing on the garbage image in combination with the similarity coefficient;

[0012] S3: Identify and classify the enhanced junk images.

[0013] Preferably, the method for determining the first similarity of each image block under each color channel is:

[0014] In the garbage images of each color channel, the gray level co-occurrence matrix of each image block is obtained, and the inverse difference moment of the gray level co-occurrence matrix of each image block is used as the first similarity of each image block under each color channel.

[0015] Preferably, the method for determining the second similarity of each image block in each color channel is:

[0016] Probability distribution statistics are performed on the color components of all pixels in each image block in the garbage image under each color channel, and the color components of all pixels in the garbage image under each color channel, and the results of the probability distribution statistics are fitted to obtain the probability distribution curve of each image block under each color channel and the probability distribution curve of the garbage image under each color channel;

[0017] The similarity of the probability distribution curve between each image block in each color channel and the corresponding garbage image is used as the second similarity of each image block in each color channel.

[0018] Preferably, the similarity coefficient of each image block under each color channel is the product of a first similarity and a second similarity of each image block under each color channel.

[0019] Preferably, the method of screening out suspected occlusion directions from all directions is:

[0020] In the garbage images of each color channel, the cumulative sum of the similarity coefficients of all reference blocks in each direction of each image block is calculated, and the direction corresponding to the maximum cumulative sum is taken as the suspected occlusion direction of each image block.

[0021] Preferably, the method for determining the color difference of each image block is:

[0022] The difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and the other directions is recorded as the color loss difference between the suspected occlusion direction of each image block and the other directions;

[0023] The average of the color loss differences between the suspected occlusion direction of each image block and all other directions is taken as the color difference of each image block.

[0024] Preferably, the method for determining the color loss value of each image block is:

[0025] In the garbage image of each color channel, calculate the mean distance from the triplet of all reference blocks to the fitting straight line in the suspected occlusion direction of each image block, and record it as the mean distance of each image block;

[0026] The color loss value of each image patch is the inverse of the product of the color difference of each image patch and the mean distance.

[0027] Preferably, the enhancing of the junk image comprises:

[0028] In the garbage image of each color channel, the cumulative sum of the differences in color loss values ​​between each image block and all other image blocks is recorded as the detail loss difference of each image block;

[0029] The product of the detail loss difference of each image block and the similarity coefficient is used as the loss factor of each image block;

[0030] The optimized CLF parameter R of image block i under color channel u u,i The expression is: R u,i =S+γ×T u,i ; Wherein, S and γ represent the preset first value and the preset second value respectively; T u,i Represents the normalized value of the loss factor of image block i under color channel u;

[0031] Each image block in the junk image under each color channel is used as the input of the contrast-limited adaptive histogram equalization algorithm, wherein the optimized CLF parameter of each image block is used as the CLF parameter in the contrast-limited adaptive histogram equalization algorithm, and each image block after enhancement is output, and all image blocks in the junk images of all color channels are traversed to obtain the enhanced junk images of all color channels, and the junk images of all color channels are fused to obtain the enhanced junk images.

[0032] Preferably, the identifying and classifying the enhanced junk images includes:

[0033] Obtain a large amount of garbage sample data and its category annotations, train the deep learning model, use the enhanced garbage images as the input of the trained deep learning model, output the category labels of the enhanced garbage images, and transfer the garbage to the corresponding storage area according to the category labels of the garbage images.

[0034] In the second aspect, an embodiment of the present application also provides a multifunctional garbage transfer and classification integrated management system, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, it implements the steps of any one of the multifunctional garbage transfer and classification integrated management methods described above.

[0035] This application has at least the following beneficial effects:

[0036] This application constructs a similarity coefficient by analyzing the similarity of texture features between an image block and a garbage image in its color channel, which can help identify areas of detail loss in garbage images that may be caused by occlusion, thereby providing a basis for subsequent image enhancement processing, thereby improving the efficiency of garbage identification and classification; further, by analyzing the differences in color intensity distribution characteristics of image blocks along different directions, a color loss value is constructed, which can further assist in judging the possibility and degree of occlusion of image blocks, thereby performing image enhancement processing on areas of detail loss caused by occlusion, improving the efficiency of image enhancement while further improving the efficiency of garbage identification and classification; further, the image enhancement algorithm is optimized by combining the color loss value and the similarity coefficient, the image details of the garbage occluded area are amplified, and the garbage is classified and processed based on the enhanced garbage image in combination with a deep learning model, thereby improving the efficiency of garbage identification and classification. This application significantly improves the accuracy of garbage identification and classification and the efficiency of integrated management of garbage transfer and classification by optimizing the preprocessing of garbage images and reducing the impact of uneven lighting and occlusion on garbage identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 A flowchart of the steps of a multifunctional garbage transfer and classification integrated management method provided in one embodiment of the present application;

[0039] Figure 2 A schematic diagram of a color difference extraction process provided by an embodiment of the present application;

[0040] Figure 3 A flowchart of junk image enhancement steps provided for one embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, describes in detail the specific implementation method, structure, features and effects of a multifunctional garbage transfer and classification integrated management method and system proposed in the present application. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0043] The following is a detailed description of a multifunctional integrated management method and system for garbage transfer and classification provided by the present application in conjunction with the accompanying drawings.

[0044] See also Figure 1 , which shows a flowchart of a multifunctional garbage transfer and classification integrated management method provided by an embodiment of the present application, the method comprising the following steps:

[0045] S1: Obtain the garbage image on each color channel of RGB, and record the component of the pixel value of each pixel in each color channel as the color component of each pixel.

[0046] The intelligent garbage room adopted in this embodiment realizes the integrated management of multifunctional transfer and classification. Its management system covers several key modules including garbage collection, intelligent identification, classified transmission, and transfer storage. Each module works closely together to ensure the orderly progress of garbage disposal. Among them, the garbage classification module is mainly responsible for providing residents with convenient and hygienic garbage delivery channels, realizing contactless garbage delivery, minimizing the risk of cross infection, and smoothly transporting the garbage delivered by residents to the subsequent processing links; the intelligent identification module is the core task of accurately classifying and identifying garbage. It uses the powerful ability of the deep learning model to accurately determine the category of garbage, providing a key basis for subsequent classified transmission, and is an important link in realizing fine classification of garbage; the classified transmission module accurately transmits different categories of garbage to the corresponding garbage bin or storage area according to the garbage category information output by the intelligent identification module, realizes efficient classification and collection of garbage, and lays the foundation for subsequent garbage transfer storage and final processing; the transfer storage module temporarily stores the classified garbage, plays the role of buffering and transfer, so that it can be uniformly transferred and processed according to the corresponding garbage disposal process in the future, ensuring the continuity and stability of the entire garbage disposal system.

[0047] Considering that the garbage image may have uneven brightness due to environmental changes during the garbage classification and identification process, which may affect the accuracy of classification, this embodiment obtains the garbage image after the bag is broken from the classification table of the intelligent garbage room.

[0048] Since the brightness state of the garbage image is unstable during the classification process in the intelligent garbage room, and some areas may have low brightness due to ambient light or occlusion, in the actual processing process, this application optimizes and adjusts the image based on the brightness distribution characteristics of the garbage image before classification to improve the accuracy of intelligent garbage recognition; in the image acquisition process in the intelligent garbage room, the acquired garbage image is a color image in the RGB color space. In order to completely retain the color information of the image, this embodiment extracts the image of the garbage image above and below each RGB color channel, and records the component of the pixel value of each pixel in each color channel as the color component of each pixel, and conducts independent analysis on the color intensity change characteristics of the garbage image in each color channel.

[0049] S2: Enhance the garbage image according to the distribution characteristics of the color components of the pixels in the garbage image under each color channel.

[0050] S201: Divide the junk image of each color channel into multiple image blocks, analyze the similarity of color components between adjacent pixels in each image block, determine a first similarity of each image block under each color channel, evaluate the similarity of color component probability distribution between each image block and the junk image of the color channel where it is located, determine a second similarity of each image block under each color channel, and determine a similarity coefficient of each image block under each color channel in combination with the first similarity.

[0051] Due to the complex and changeable ambient light in the smart garbage room, interference factors such as garbage stacking or mechanical equipment blocking, the garbage images taken will have uneven brightness, which will cause the contrast of different areas of the garbage image to change, and some areas will lose details and features will be difficult to identify, affecting the deep learning model's extraction of image features. The more serious the uneven brightness problem, the worse the accuracy of intelligent recognition, which in turn reduces the efficiency of garbage classification.

[0052] When the garbage image is in an environment with good lighting conditions, different garbage presents richer colors, which is reflected in the garbage image of each color channel. The distribution of the color components contained in it is more complex, and the edge position of the garbage target is more prominent in the image. Correspondingly, the image contains richer detail information. On the contrary, in areas covered by garbage or areas with low light intensity, the garbage image contains relatively less detail information. Therefore, by analyzing the changes in the color components of pixels in local areas of the garbage image, it is determined whether there is occlusion in the garbage image, and the occluded area is enhanced to further improve the efficiency of garbage classification. Specifically:

[0053] (1) In view of the differences in details contained in different areas of the garbage image, in order to more accurately obtain the features of each local area, the present embodiment divides the garbage image of each color channel into a plurality of continuous and non-overlapping image blocks. It should be noted that the number of divided image blocks is inversely proportional to the size of the image area covered by each image block, that is, the more divided image blocks, the smaller the image area covered by each image block, and conversely, if the number of divided image blocks is smaller, the image area covered by each image block is larger.

[0054] (2) Furthermore, in the junk image of each color channel, the gray level co-occurrence matrix of each image block is obtained, and the inverse difference moment of the gray level co-occurrence matrix of each image block is used as the first similarity of each image block under each color channel; the first similarity reflects the uniformity of the pixel value distribution difference of different image blocks in the junk image. The larger the first similarity, the more uniform the pixel distribution in the local area of ​​the junk image, that is, the local detail features in the junk image will be likely to be weakened, indicating that the possibility of the junk image being blocked is greater.

[0055] Among them, the method for obtaining the gray level co-occurrence matrix and the method for calculating the inverse difference moment are both well-known technologies, and the process of obtaining the gray level co-occurrence matrix and the process of calculating the inverse difference moment will not be repeated.

[0056] (3) Furthermore, considering the difference in the local impact of detail features due to occlusion in the local area after the garbage bag is broken, the second similarity of each image block is constructed by analyzing the similarity between the image block and the garbage image under each color channel to determine the degree of occlusion of the garbage, specifically:

[0057] Probability distribution statistics are performed on the color components of all pixels in each image block in the garbage image under each color channel, and the color components of all pixels in the garbage image under each color channel, and the results of the probability distribution statistics are fitted to obtain the probability distribution curve of each image block under each color channel and the probability distribution curve of the garbage image under each color channel;

[0058] It should be noted that, in this embodiment, the probability distribution histogram of the color components of all pixels in the garbage image under each color channel and the probability distribution histogram of the color components of all pixels in each image block are statistically calculated, and the probability distribution histogram is fitted to obtain the probability distribution curve, wherein the methods for obtaining the histogram and the probability distribution curve are both well-known technologies, and the specific process will not be repeated here.

[0059] In addition, it should be understood that there are many commonly used fitting methods. In this embodiment, the least square method is used to fit the statistical results of the probability distribution. The implementer may also use other fitting methods such as polynomial fitting method. The embodiment does not impose any special restrictions on the selection of the fitting method. Among them, the least square method is a well-known technology, and its specific fitting process will not be repeated.

[0060] (4) Further, the similarity of the probability distribution curve between each image block in each color channel and the corresponding garbage image is used as the second similarity of each image block in each color channel; the larger the second similarity, the more serious the loss of detail features in the garbage image due to environmental occlusion and interference during the garbage processing process.

[0061] It should be noted that there are many methods for measuring the similarity between curves. In this embodiment, the KL divergence of the probability distribution curve between each image block and the corresponding garbage image under each color channel is used as the similarity between the probability distribution curve between each image block and the corresponding garbage image under each color channel. In actual application, as other implementation methods, the implementer may also use other methods for measuring the similarity between curves such as cosine similarity. Regarding the selection of the method for measuring the similarity between curves, this embodiment does not impose any special restrictions. Among them, the calculation method of KL divergence is a well-known technology, and its specific calculation process will not be repeated.

[0062] (5) Further combining the first similarity and the second similarity to determine the similarity coefficient, and determine the degree to which the garbage is blocked and interfered, specifically:

[0063] The product of the first similarity and the second similarity of each image block under each color channel is used as the similarity coefficient of each image block under each color channel.

[0064] According to the similarity of each image block under each color channel, it can be understood that if the first similarity of the current image block is larger and the second similarity is larger, the similarity coefficient of the current image block is larger, indicating that the image of the corresponding area of ​​the image block in the garbage image has caused serious detail loss due to environmental occlusion interference. Therefore, the image of the corresponding area in the garbage image where the image block is located is enhanced, so that the garbage can be more accurately identified and classified according to the enhanced texture details, thereby improving the garbage classification efficiency; conversely, if the first similarity of the current image block is smaller and the second similarity is smaller, the similarity coefficient of the current image block is larger, indicating that the image of the corresponding area of ​​the image block in the garbage image is less affected by environmental occlusion interference, which can speed up the image enhancement processing progress, thereby improving the garbage classification efficiency.

[0065] S202: In the garbage image of each color channel, each image block is taken as a starting point and extended in different directions, and a preset number of image blocks in each direction of each image block are used as reference blocks in each direction of each image block, and the similarity coefficients of all reference blocks in each direction of each image block are combined to screen out suspected occlusion directions from all directions; the difference between the similarity coefficients of the suspected occlusion direction of each image block and all reference blocks in other directions is analyzed to determine the color difference of each image block.

[0066] Considering the randomness of garbage bag breaking during the garbage disposal process, the detail loss characteristics of garbage disposal images collected at different times due to occlusion have extension characteristics in different directions, that is, due to different occlusion angles, the detail loss in different directions caused by light occlusion weakens as the light extends. Therefore, based on the above analysis, this embodiment performs extension comparison in different directions for each image block, specifically:

[0067] In the garbage images of each color channel, each image block is taken as the starting point, and extended comparisons are performed in the directions of 0 degree, 45 degree, 90 degree, 135 degree, 180 degree, 225 degree, 270 degree, and 315 degree, respectively, wherein a preset number of image blocks in each direction of each image block are used as reference blocks in each direction of each image block.

[0068] It should be noted that the value of the preset number is artificially set. In this embodiment, the value of the preset number is 5. The implementer can also set it by himself according to the specific situation. This embodiment does not impose any special restrictions.

[0069] Furthermore, if the similarity coefficient between the reference block extending in a certain angle direction of the image block and the image block is relatively close, and the similarity coefficient gradually decreases in this direction, but the similarity coefficient difference between the reference block extending in other directions and the image block at the corresponding starting point is relatively large, and does not show a gradually decreasing trend, then it means that the occlusion extension feature of the image block is more obvious. Therefore, the similarity coefficients of all reference blocks in each direction of each image block are combined to screen out the suspected occlusion direction from all directions; the difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and the other directions is analyzed to determine the color difference of each image block, specifically:

[0070] In the garbage images of each color channel, the cumulative sum of the similarity coefficients of all reference blocks in each direction of each image block is calculated, and the direction corresponding to the maximum cumulative sum is taken as the suspected occlusion direction of each image block.

[0071] The difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and the other directions is recorded as the color loss difference between the suspected occlusion direction and the other directions of each image block; if the color loss difference of the current image block is larger, it means that the color difference between the suspected occlusion direction and other directions is more obvious, which means that the current image block is more likely to be disturbed by environmental occlusion, and it is more necessary to enhance the area corresponding to the image block in the garbage image; conversely, if the color loss difference of the current image block is smaller, it means that the current image block is less likely to be disturbed by environmental occlusion.

[0072] It should be noted that there are many methods for measuring the differences between data groups. In this embodiment, the DTW distance between the similarity coefficients of each image block in the suspected occlusion direction and all the reference blocks in the other directions is taken as the difference between the similarity coefficients of each image block in the suspected occlusion direction and all the reference blocks in the other directions. In actual application, as other implementation methods, the implementer may also adopt other methods for measuring the differences between data groups, such as Euclidean distance or Manhattan distance. This embodiment does not impose any special restrictions on the selection of methods for measuring the differences between data groups.

[0073] The calculation method of the DTW distance is a well-known technology, and its specific calculation process will not be described in detail.

[0074] Furthermore, the mean of the color loss difference between the suspected occlusion direction and all other directions of each image block is taken as the color difference of each image block; if the color difference of the current image block is smaller, it means that the detail loss effect of the current image block along the suspected occlusion direction is gradually weakened, which means that the current image block is more likely to be occluded by the environment; conversely, if the color difference of the current image block is larger, it means that the distribution difference of the color components of the pixels in the current image block along the suspected occlusion direction is larger, and the current image block is less likely to be occluded by the environment.

[0075] Preferably, the color difference extraction process schematic diagram provided in this embodiment is as follows Figure 2 shown.

[0076] S203: The coordinates and similarity coefficients of each image block are grouped into a triplet, and the triplet of each image block and all reference blocks in the suspected occlusion direction is fitted to obtain a fitting straight line, and the distance from the triplet of all reference blocks in the suspected occlusion direction to the fitting straight line is analyzed, and the color loss value of each image block is determined in combination with the color difference.

[0077] In the garbage image, the occluded part of the garbage usually occupies a small part of the entire garbage image. If the impact of uneven illumination caused by occlusion is more serious, the color difference of the detail loss extension feature between each image block will be smaller. Therefore, by analyzing the change trend of the color component of the reference block along the suspected occlusion direction of each image block, the color loss value of each image block is constructed, which is:

[0078] The coordinates and similarity coefficients of each image block are formed into a triplet, and the triplet of each image block and all reference blocks in the suspected occlusion direction is fitted to obtain a fitting straight line. There are many commonly used fitting methods. In this embodiment, the multivariate linear regression method is used to fit the triplet to obtain a fitting straight line. In actual application, as other implementation methods, the implementer may also use other fitting methods such as multivariate polynomial regression. This embodiment does not impose any special restrictions on the selection of the fitting method. Multivariate linear regression is a well-known technology, and its specific principle process will not be repeated.

[0079] Furthermore, in the garbage images of each color channel, the mean distance from the triplets of all reference blocks to the fitting straight line in the suspected occlusion direction of each image block is calculated, and recorded as the mean distance of each image block;

[0080] The color loss value of each image patch is the inverse of the product of the color difference of each image patch and the mean distance.

[0081] According to the color loss value of each image block, it can be understood that if the color difference of the current image block is smaller and the mean distance of the image block is smaller, the color loss value of the current image block is larger, which means that the color difference between the image block and the reference block in the suspected occlusion direction is smaller, indicating that the possibility that the current image block is blocked is greater; conversely, if the color difference of the current image block is larger and the mean distance of the image block is larger, the color loss value of the current image block is smaller, which means that the color difference between the image block and the reference block in the suspected occlusion direction is larger, indicating that the possibility that the current image block is blocked is smaller.

[0082] S204: In the garbage images of each color channel, the difference in color loss value between each image block and all other image blocks is analyzed, and the garbage images are enhanced in combination with the similarity coefficient.

[0083] In order to solve the problem of uneven brightness in garbage images due to environmental changes, the color loss value of each image block is obtained from the above analysis. This value reflects the degree of occlusion of each image block. Therefore, according to the degree of occlusion of the image block, the image enhancement processing is performed on the corresponding area of ​​the image block in the garbage image, specifically:

[0084] In the garbage image of each color channel, the cumulative sum of the differences in color loss values ​​between each image block and all other image blocks is recorded as the detail loss difference of each image block;

[0085] Furthermore, the product of the detail loss difference of each image block and the similarity coefficient is used as the loss factor of each image block;

[0086] According to the loss factor of each image block, it can be understood that if the difference in detail loss of the image block is greater and the similarity coefficient is greater, the loss factor is greater, which means that the detail loss of the image block is more serious and the possibility of the image block being occluded is greater, and the image block should be enhanced; conversely, if the difference in detail loss of the image block is smaller and the similarity coefficient is smaller, the loss factor is smaller, which means that the possibility of the image block being occluded is smaller.

[0087] Furthermore, based on the loss factor of each image block, the contrast-limited adaptive histogram equalization algorithm is improved. Specifically:

[0088] The optimized CLF parameter R of image block i under color channel u u,i The expression is: R u,i =S+γ×T u,i ; Wherein, S and γ represent the preset first value and the preset second value respectively; T u,i Represents the normalized value of the loss factor of image patch i under color channel u.

[0089] It should be noted that, since the value range of the CLF parameter in the contrast-limited adaptive histogram equalization algorithm is generally between [2, 5], the values ​​of the preset first value S and the preset second value γ in this embodiment are 2 and 3 respectively, and there is no limit on the size between the preset first value and the preset second value. The implementer can also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0090] Furthermore, each image block in the junk image under each color channel is used as the input of the contrast-constrained adaptive histogram equalization algorithm, wherein the optimized CLF parameter of each image block is used as the CLF parameter in the contrast-constrained adaptive histogram equalization algorithm, and each image block after enhancement is output, and all image blocks in the junk images of all color channels are traversed to obtain enhanced junk images of all color channels, and the junk images of all color channels are fused to obtain enhanced junk images.

[0091] The process of fusing the garbage images of all color channels is a well-known technology, and its specific principle will not be described in detail.

[0092] Preferably, the flowchart of the garbage image enhancement steps provided in this embodiment is as follows: Figure 3 shown.

[0093] S3: Identify and classify the enhanced junk images.

[0094] In this embodiment, a large amount of garbage sample data is obtained and the garbage sample data is labeled by category, all the garbage sample data and their category standards are used as a convolutional neural network for training to obtain a trained convolutional neural network, and the enhanced garbage image is used as the trained convolutional neural network to output the category label of the enhanced garbage image, and the garbage is transmitted to the corresponding storage area according to the category label of its garbage image.

[0095] Among them, convolutional neural network is a well-known technology, and its specific principle will not be repeated here.

[0096] So far, this embodiment has solved the problem of low intelligent recognition accuracy in the actual processing process caused by uneven image illumination due to variable light and occlusion in the intelligent recognition module. By integrating the detail-rich features and color distribution characteristics of each image block area in the garbage image, the preprocessing stage of the image intelligent recognition process is optimized and adjusted, thereby reducing the impact of uneven illumination on the accuracy of garbage recognition and classification, and effectively improving the integrated management efficiency of garbage transfer and classification.

[0097] Based on the same inventive concept as the above method, an embodiment of the present application also provides a multifunctional garbage transfer and classification integrated management system, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, it implements the steps of any one of the above-mentioned multifunctional garbage transfer and classification integrated management methods.

[0098] It should be noted that the above sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0099] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0100] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application should be included in the protection scope of the present application.

Claims

1. A multifunctional integrated management method for waste transfer and classification, characterized in that: The method comprises the following steps: S1: Get the garbage image on each color channel of RGB, and record the component of the pixel value of each pixel in each color channel as the color component of each pixel; S2: According to the distribution characteristics of the color components of the pixels in the garbage image under each color channel, the garbage image is enhanced. The specific process is as follows: S201: Divide the garbage image of each color channel into multiple image blocks, analyze the similarity of color components between adjacent pixels in each image block under each color channel, determine a first similarity of each image block under each color channel, evaluate the similarity of color component probability distribution between each image block and the garbage image of the color channel where it is located, determine a second similarity of each image block under each color channel, and determine a similarity coefficient of each image block under each color channel in combination with the first similarity; S202: In the garbage image of each color channel, each image block is extended in different directions from the starting point, a preset number of image blocks in each direction of each image block are used as reference blocks in each direction of each image block, and the similarity coefficients of all reference blocks in each direction of each image block are combined to screen out suspected occlusion directions from all directions; the difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and those in the remaining directions is analyzed to determine the color difference of each image block in each color channel; S203: The coordinates and similarity coefficients of each image block are combined into a triplet, and the triplet of each image block and all reference blocks in the suspected occlusion direction are fitted to obtain a fitting straight line, and the distance from the triplet of all reference blocks in the suspected occlusion direction to the fitting straight line is analyzed, and the color loss value of each image block in each color channel is determined in combination with the color difference; S204: analyzing the difference in color loss value between each image block and all other image blocks in the garbage image of each color channel, and performing enhancement processing on the garbage image in combination with the similarity coefficient; S3: Identify and classify the enhanced junk images.

2. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The method for determining the first similarity of each image block under each color channel is: In the garbage images of each color channel, the gray level co-occurrence matrix of each image block is obtained, and the inverse difference moment of the gray level co-occurrence matrix of each image block is used as the first similarity of each image block under each color channel.

3. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The method for determining the second similarity of each image block under each color channel is: Probability distribution statistics are performed on the color components of all pixels in each image block in the garbage image under each color channel, and the color components of all pixels in the garbage image under each color channel, and the results of the probability distribution statistics are fitted to obtain the probability distribution curve of each image block under each color channel and the probability distribution curve of the garbage image under each color channel; The similarity of the probability distribution curve between each image block in each color channel and the corresponding garbage image is used as the second similarity of each image block in each color channel.

4. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The similarity coefficient of each image block under each color channel is the product of the first similarity and the second similarity of each image block under each color channel.

5. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The method for screening out suspected occlusion directions from all directions is: In the garbage images of each color channel, the cumulative sum of the similarity coefficients of all reference blocks in each direction of each image block is calculated, and the direction corresponding to the maximum cumulative sum is taken as the suspected occlusion direction of each image block.

6. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The method for determining the color difference of each image block is: The difference between the similarity coefficients of all reference blocks in the suspected occlusion direction of each image block and the other directions is recorded as the color loss difference between the suspected occlusion direction of each image block and the other directions; The average of the color loss differences between the suspected occlusion direction of each image block and all other directions is taken as the color difference of each image block.

7. A multifunctional integrated management method for waste transfer and classification as claimed in claim 1, characterized in that: The method for determining the color loss value of each image block is: In the garbage image of each color channel, calculate the mean distance from the triplet of all reference blocks to the fitting straight line in the suspected occlusion direction of each image block, and record it as the mean distance of each image block; The color loss value of each image patch is the inverse of the product of the color difference of each image patch and the mean distance.

8. A multifunctional integrated management method for waste transfer and classification as claimed in claim 1, characterized in that: The step of enhancing the junk image comprises: In the garbage image of each color channel, the cumulative sum of the differences in color loss values ​​between each image block and all other image blocks is recorded as the detail loss difference of each image block; The product of the detail loss difference of each image block and the similarity coefficient is used as the loss factor of each image block; The optimized CLF parameter R of image block i under color channel u u,i The expression is: R u,i =S+γ×T u,i ; Wherein, S and γ represent the preset first value and the preset second value respectively; T u,i Represents the normalized value of the loss factor of image block i under color channel u; Each image block in the junk image under each color channel is used as the input of the contrast-limited adaptive histogram equalization algorithm, wherein the optimized CLF parameter of each image block is used as the CLF parameter in the contrast-limited adaptive histogram equalization algorithm, and each image block after enhancement is output, and all image blocks in the junk images of all color channels are traversed to obtain the enhanced junk images of all color channels, and the junk images of all color channels are fused to obtain the enhanced junk images.

9. A multifunctional garbage transfer and classification integrated management method as claimed in claim 1, characterized in that: The identifying and classifying the enhanced junk images includes: Obtain a large amount of garbage sample data and its category annotations, train the deep learning model, use the enhanced garbage images as the input of the trained deep learning model, output the category labels of the enhanced garbage images, and transfer the garbage to the corresponding storage area according to the category labels of the garbage images.

10. A multifunctional integrated management system for waste transfer and classification, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, it implements the steps of a multifunctional garbage transfer and classification integrated management method as described in any one of claims 1-9.

Images