Garbage treatment method
By identifying and enhancing the low-resolution combustion images in the incinerator, and using neural network model to process, the problem of low image clarity during waste incineration is solved, accurate observation and real-time transmission of combustion state are achieved, and garbage disposal is ensured. Adequacy.
Patent Information
- Application Number
- CN202411257944.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-09
- Publication Date
- 2024-11-22
AI Technical Summary
During the garbage incineration process, due to the different composition and moisture content of the garbage, the collected combustion image has low clarity and cannot effectively observe the combustion state of the combustion substances, which affects the adequacy of garbage disposal.
A garbage disposal method is adopted to collect combustion images of incinerators in real time, use the first model to identify images with low processing resolution, and perform enhanced processing through the second model to ensure image clarity, combining convolutional layer and pooling layer and the neural network model of the fully connected layer uses random inactivation function and pseudo-inverse calculation to determine the enhanced weight coefficients, so as to achieve effective processing of low-resolution images.
It realizes rapid, realistic and effective processing of combustion images during the waste incineration stage, ensures that operators can accurately observe the combustion state, ensures that the sufficiency of garbage disposal and the real-time transmission of image information.
Smart Images

Figure CN119006846A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of garbage incineration, and in particular to a garbage treatment method. Background Art
[0002] At present, the main treatment method for urban domestic waste in my country is incineration. The moisture content and calorific value of domestic garbage vary greatly. During the incineration process, the operator needs to observe the combustion state of the incinerator furnace at all times, and then continuously adjust the incinerator operating parameters to ensure normal and stable incineration. However, due to the different amount of garbage loaded into the incinerator each time, different garbage composition, and large changes in moisture content, when the image acquisition device is used to collect the flame combustion image in the furnace, the picture may be blurred and the clarity is low, and the combustion state of the combustible cannot be effectively observed. However, this phenomenon is intermittent and will not exist for a long time, so it cannot guarantee that the garbage is fully treated. Summary of the invention
[0003] The purpose of the present invention is to provide a garbage disposal method, which can not only determine whether the current combustion image is blurred or low in pixel, but also ensure that when the image clarity collected during the garbage incineration stage is very low, the flame combustion image is effectively processed, so that the operation management personnel can effectively observe the combustion state of the combustible material, and at the same time ensure the real-time transmission of image information, and the combustion image processing is fast, real and effective, so that the garbage can be fully disposed of.
[0004] The present invention provides a garbage disposal method, the method comprising:
[0005] Real-time acquisition of combustion images of the combustion section in the incinerator;
[0006] Performing recognition processing on the collected combustion image to obtain a recognition result, wherein the recognition result includes a low-resolution image and a non-low-resolution image, and the recognition processing is implemented by a first model;
[0007] Performing enhancement processing on the identified low-resolution image to obtain an enhanced image, wherein the enhancement processing is implemented by a second model;
[0008] The combustion conditions of the garbage in the non-low-resolution image and the enhanced image are determined, and when it is determined that the garbage has been burned to ashes, the ashed garbage is discharged, otherwise the garbage continues to burn in the combustion section.
[0009] As a further improvement of the present invention, the first model includes: a first convolutional layer, a pooling layer, a second convolutional layer and a fully connected layer, and the recognition processing is implemented by the first model, including:
[0010] Performing convolution processing on the combustion image through the first convolution layer to extract local features of the combustion image to obtain a first feature vector;
[0011] Performing pooling processing on the first feature vector through the pooling layer to perform feature screening on the extracted local features to obtain a second feature vector;
[0012] Performing convolution processing on the second feature vector through the second convolution layer to extract features of the filtered local features to obtain a third feature vector;
[0013] The third feature vector is fully connected through the fully connected layer to obtain a recognition result.
[0014] As a further improvement of the present invention, the method further includes: training the first model using a first training data set, and the fully connected layer uses a random deactivation function during training.
[0015] As a further improvement of the present invention, the second model includes an input layer, a mapping layer, an enhancement layer and an output layer, and the enhancement processing is implemented by the second model, including:
[0016] Initializing the input layer;
[0017] Mapping the low-resolution image through the mapping layer to obtain a feature vector Z of a feature node;
[0018] Performing enhancement processing on the feature vector of the feature node through the enhancement layer to obtain a feature vector H of the enhanced node;
[0019] The feature vector of the feature node and the feature vector of the enhanced node are cascaded through the output layer S=[Z|H]W O , get the enhanced image, where the W 0 Represents the enhancement weight coefficient.
[0020] As a further improvement of the present invention, the initialization process of the input layer includes initializing the random weight value W of each feature node i ei and the mapping feature bias value β ei , the mapping process includes:
[0021] According to the random weight value W corresponding to the feature node i ei and the mapping feature bias value β ei , determine the characteristic vector Z of the characteristic node i .
[0022] As a further improvement of the present invention, the initialization process of the input layer includes initializing the enhancement weight value W of each enhancement node j hj and the enhancement bias value β hj , the enhancement processing includes:
[0023] According to the characteristic vector Z of the characteristic node i , the enhanced weight value W corresponding to the enhanced node j hj and the enhanced node bias value β hj , determine the eigenvector H of the enhanced node j .
[0024] As a further improvement of the present invention, the method further includes: training the second model using a second training data set to determine the enhanced weight coefficient W 0 .
[0025] As a further improvement of the present invention, before collecting the combustion image, the method further includes: pushing the waste to be processed into a drying section in an incinerator, injecting combustion-supporting air into the drying section, heating the waste to be processed by the incinerator, evaporating the water in the waste to be processed to obtain dry waste;
[0026] The dry garbage then enters the combustion section of the incinerator for combustion.
[0027] As a further improvement of the present invention, the first convolution layer and the second convolution layer use a 5*5 convolution kernel with a step size of 4;
[0028] The pooling layer adopts maximum pooling.
[0029] As a further improvement of the present invention, during training, the enhancement weight coefficient W is determined by pseudo-inverse calculation. 0 .
[0030] The beneficial effects of the present invention are as follows: it can not only determine whether the current combustion image is blurred or has low pixels, but also ensure that when the clarity of the image collected during the garbage incineration stage is very low, the flame combustion image can be effectively processed, so that the operation management personnel can effectively observe the combustion state of the combustibles, and at the same time it also ensures the real-time transmission of image information, and the combustion image processing is fast, real and effective, so that the garbage can be fully processed. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.
[0032] Figure 1 The present invention is a flowchart of a garbage disposal method according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION
[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0034] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.
[0035] In addition, in the description of the present invention, the terms used are only for illustrative purposes and are not intended to limit the scope of the present invention. The terms "include" and / or "comprise" are used to specify the existence of the elements, steps, operations and / or components, but do not exclude the existence or addition of one or more other elements, steps, operations and / or components. The terms "first", "second" and the like may be used to describe various elements, do not represent the order, and do not limit these elements. In addition, in the description of the present invention, unless otherwise specified, the meaning of "multiple" is two and more than two. These terms are only used to distinguish one element from another element. In conjunction with the following drawings, these and / or other aspects become apparent, and it is easier for a person of ordinary skill in the art to understand the description of the embodiments of the present invention. The accompanying drawings are used to describe the embodiments of the present invention for illustrative purposes only. Those skilled in the art will easily recognize from the following description that, without departing from the principles of the present invention, alternative embodiments of the structures and methods shown in the present invention can be adopted.
[0036] like Figure 1 As shown, a garbage disposal method according to an embodiment of the present invention comprises:
[0037] Real-time acquisition of combustion images of the combustion section in the incinerator;
[0038] Performing recognition processing on the collected combustion image to obtain a recognition result, wherein the recognition result includes a low-resolution image and a non-low-resolution image, and the recognition processing is implemented by a first model;
[0039] Performing enhancement processing on the identified low-resolution image to obtain an enhanced image, wherein the enhancement processing is implemented by a second model;
[0040] The combustion conditions of the garbage in the non-low-resolution image and the enhanced image are determined, and when it is determined that the garbage has been burned to ashes, the ashed garbage is discharged, otherwise the garbage continues to burn in the combustion section.
[0041] In the related art, operators generally directly observe the collected combustion images to determine the burning status of garbage in the incinerator. However, due to the differences in the composition, weight, water content, etc. of the garbage, the collected combustion images may have insufficient clarity, which in turn makes it impossible for operators to accurately determine the burning status of the garbage.
[0042] The method described in the present application divides the combustion images into low-resolution images and non-low-resolution images, wherein the low-resolution images are blurred and have low pixels (the operator sets a standard value, and the pixels are lower than the standard value, which means the pixels are low. The present application does not limit the standard value and can be set according to the actual situation), while other images are non-low-resolution images. For non-low-resolution images, the operator can accurately determine whether the garbage is burned to ashes by direct observation, and for low-resolution images, they are first quickly screened out by the first model, and then enhanced by the second model to improve and strengthen the clarity of the low-resolution images, so that the operator can determine whether the garbage is burned to ashes by direct observation. The coordinated use of the two models can effectively solve the problem of the inability to accurately determine the garbage incineration situation due to the insufficient clarity of the combustion image. At the same time, through the use of the two models, the normal and stable operation of the incinerator can be ensured while the garbage combustion situation in the incinerator can be determined in real time.
[0043] In an optional implementation manner, before acquiring the combustion image, the method further includes:
[0044] Pushing the waste to be processed into the drying section of the incinerator, injecting combustion-supporting air into the drying section, heating the waste to be processed by the incinerator, evaporating the water in the waste to be processed to obtain dry waste;
[0045] The dry garbage then enters the combustion section of the incinerator for combustion.
[0046] Preferably, the garbage to be processed is pushed into the incinerator through a feeder.
[0047] The process of treating the waste to be treated in the waste incinerator is as follows: the feeder first pushes the waste to be treated into the drying section of the incinerator, and the water in the waste is evaporated and precipitated to obtain dry waste through the heat conduction of the incinerator wall, the thermal effect of the high-temperature flue gas in the furnace and the high-temperature combustion-supporting air. After the dry waste enters the combustion section, it burns violently and releases a large amount of heat. The image acquisition device collects the combustion image of the combustion process in the combustion section in real time, and the first model is used to identify whether the combustion image is a non-low-resolution image:
[0048] If it is not a non-low-resolution image, that is, it is recognized as a low-resolution image by the first model, the low-resolution image is input into the second model, an enhanced image is obtained through processing by the second model, and the enhanced image is transmitted to the combustion image management platform;
[0049] If it is a non-low-resolution image, the non-low-resolution image is directly transmitted to the combustion image management platform.
[0050] The combustion image management platform automatically records the time of image transmission. The garbage disposal personnel judge whether the dry garbage is burned to ashing in the combustion stage according to the enhanced image or non-low-resolution image display in the combustion image management platform:
[0051] If it is not burned to ash, continue to burn in the combustion section until it is burned to ash;
[0052] The garbage burned to ashes is discharged from the combustion furnace, and the garbage treatment is completed.
[0053] Preferably, a substantial amount of the heat generated in the combustion stage is used to generate electricity.
[0054] The first model and the second model do not limit the meaning of the model order. The first model represents a model for identifying whether the combustion image is a low-resolution image, and the second model represents a model for enhancing the identified low-resolution image.
[0055] In an optional implementation, the first model includes: a first convolutional layer, a pooling layer, a second convolutional layer and a fully connected layer, and the recognition process is implemented by the first model, including:
[0056] Performing convolution processing on the combustion image through the first convolution layer to extract local features of the combustion image to obtain a first feature vector;
[0057] Performing pooling processing on the first feature vector through the pooling layer to perform feature screening on the extracted local features to obtain a second feature vector;
[0058] Performing convolution processing on the second feature vector through the second convolution layer to extract features of the filtered local features to obtain a third feature vector;
[0059] The third feature vector is fully connected through the fully connected layer to obtain a recognition result.
[0060] The first feature vector, the second feature vector and the third feature vector do not limit the order of the feature vectors. The first feature vector represents the extracted local features, the second feature vector represents the filtered local features, and the third feature vector represents the feature vector for extracting features from the filtered local features.
[0061] The first model of the present application is a neural network model with four basic layers, which has fewer parameters and can quickly extract image details. While realizing rapid recognition of low-resolution images, it greatly reduces the computational complexity of the model, reduces the demand for computing equipment, and correspondingly reduces the size of the model and storage space.
[0062] Preferably, the first convolution layer and the second convolution layer use a 5*5 convolution kernel with a step size of 4;
[0063] The pooling layer adopts maximum pooling.
[0064] The combustion image is input into the first model, and the first model performs recognition processing on the combustion image to recognize whether the combustion image is a low-resolution image or a non-low-resolution image.
[0065] In an optional implementation, the method further includes: training the first model using a first training data set, and using a random dropout function in the fully connected layer during training.
[0066] Inputting a first training data set into a first model, wherein the first training data set includes a plurality of training combustion images and corresponding recognition results, wherein the recognition results include the low-resolution image and the non-low-resolution image, and inputting the training combustion images into the first model;
[0067] A first convolutional layer in the first model extracts a plurality of local features from the training combustion image;
[0068] Inputting the multiple local features into the pooling layer in the first model for feature screening;
[0069] Inputting the filtered local features into the second convolutional layer in the first model to extract features;
[0070] The extracted features are summarized and input into the fully connected layer in the first model, and a training recognition result is output, wherein the training recognition result includes recognizing the training combustion image as the low-resolution image or the non-low-resolution image, and learning the recognition result in the first training data set with the training recognition result to obtain a trained first model.
[0071] Preferably, the recognition results in the first training data set are compared with the training recognition results for study.
[0072] The first training data set can be divided into a training set and a test set. The first model is trained by the training set. The trained first model is used as a recognition (prediction) model for the combustion image collected in real time, and then the combustion image is identified as a low-resolution image based on the trained first model. The test set is used to test the trained first model to evaluate the performance and generalization ability of the first model. The random inactivation function is used during training, which can effectively alleviate the overfitting problem of the first model when the training set samples are small, play a regularization role, and then improve the generalization ability of the trained first model to achieve accurate recognition of the combustion image.
[0073] In an optional implementation manner, the method includes: the second model includes an input layer, a mapping layer, an enhancement layer and an output layer, and the enhancement processing is implemented by the second model, including:
[0074] Initializing the input layer;
[0075] Mapping the low-resolution image through the mapping layer to obtain a feature vector Z of a feature node;
[0076] Performing enhancement processing on the feature vector of the feature node through the enhancement layer to obtain a feature vector H of the enhanced node;
[0077] The feature vector of the feature node and the feature vector of the enhanced node are cascaded through the output layer S=[Z|H]W O , get the enhanced image, where the W 0 Represents the enhancement weight coefficient.
[0078] The low-resolution image is input to the input layer of the second model, the input layer is initialized, and the low-resolution image is mapped through the mapping layer to obtain the feature vector Z of the feature node, specifically Z=φ(XW e +β e );
[0079] The feature vector of the feature node is enhanced by the enhancement layer to obtain the feature vector H of the enhanced node, specifically H=ξ(ZW h +β h );
[0080] The feature vector of the feature node and the feature vector of the enhanced node are cascaded through the output layer S=[Z|H]W O , get the enhanced image, where the W 0 Represents the enhancement weight coefficient.
[0081] In an optional implementation manner, the initialization process of the input layer includes initializing the random weight value and the mapping feature bias value of each feature node i, and the mapping process includes:
[0082] According to the random weight value W corresponding to the feature node i ei and the mapping feature bias value β ei , determine the characteristic vector Z of the characteristic node i , specifically, Z i =φ(XW ei +β ei ),i=1....n. i represents the serial number of the feature node.
[0083] In an optional implementation, the initialization process of the input layer includes initializing the enhancement weight value W of each enhancement node j hj and the enhancement bias value β hj , the enhancement processing includes:
[0084] According to the characteristic vector Z of the characteristic node i , the enhanced weight value W corresponding to the enhanced node j hj and the enhanced node bias value β hj , determine the eigenvector H of the enhanced node j , specifically, H j =ξ(Z i W hj +β hj ),j=1...n. j represents the serial number of the feature node.
[0085] Where X is the low-resolution image feature matrix, W ei is the mapping feature weight coefficient, β ei is the mapping feature bias value, Z i is the i-th feature node, W hj is the enhancement node weight coefficient, β hj To enhance the node bias value, H j is the jth enhancement node, W O is the output weight value, Z is the feature vector of the feature node, H is the feature vector of the enhanced node, and S is the enhanced image feature matrix.
[0086] In an optional implementation manner, the method further includes: training the second model using a second training data set to determine the enhanced weight coefficient W 0 .
[0087] Inputting a second training data set into a second model, wherein the second training data set includes a plurality of the low-resolution images and a plurality of corresponding high-resolution images, and inputting the low-resolution images into the second model;
[0088] Calculate the i-th feature node Z i and the jth enhancement node H j , the feature matrix of the corresponding high-resolution image is used as the enhanced image feature matrix S, and the output weight value W is calculated by pseudo-inverse calculation O , to obtain the enhanced model after training.
[0089] The enhancement model uses pseudo-inverse calculation during the training process, which can reduce the setting of control parameters, and does not require repeated iterations of gradient descent, thereby improving training efficiency, and can also achieve accurate learning, obtain the optimal output weight value, and improve the enhancement effect of low-resolution images, thereby improving the accuracy of the operator's judgment on the burning situation of garbage in the incinerator. In the description provided herein, a large number of specific details are described. However, it can be understood that embodiments of the present invention can be practiced without these specific details. In some instances, known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.
[0090] In addition, it will be understood by those skilled in the art that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.
[0091] It will be appreciated by those skilled in the art that, although the present invention has been described with reference to exemplary embodiments, various changes may be made and equivalents may be substituted for its elements without departing from the scope of the present invention. In addition, many modifications may be made to adapt specific circumstances or materials to the teachings of the present invention without departing from the essential scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed, but the present invention will include all embodiments falling within the scope of the appended claims.
Claims
1. A method for treating garbage, characterized in that: The method comprises: Real-time acquisition of combustion images of the combustion section in the incinerator; Performing recognition processing on the collected combustion image to obtain a recognition result, wherein the recognition result includes a low-resolution image and a non-low-resolution image, and the recognition processing is implemented by a first model; Performing enhancement processing on the identified low-resolution image to obtain an enhanced image, wherein the enhancement processing is implemented by a second model; The combustion conditions of the garbage in the non-low-resolution image and the enhanced image are determined, and when it is determined that the garbage has been burned to ashes, the ashed garbage is discharged, otherwise the garbage continues to burn in the combustion section.
2. The method according to claim 1, characterized in that The first model includes: a first convolutional layer, a pooling layer, a second convolutional layer and a fully connected layer, and the recognition process is implemented by the first model, including: Performing convolution processing on the combustion image through the first convolution layer to extract local features of the combustion image to obtain a first feature vector; Performing pooling processing on the first feature vector through the pooling layer to perform feature screening on the extracted local features to obtain a second feature vector; Performing convolution processing on the second feature vector through the second convolution layer to extract features of the filtered local features to obtain a third feature vector; The third feature vector is fully connected through the fully connected layer to obtain a recognition result.
3. The method according to claim 2, characterized in that The method also includes: training the first model using a first training data set, and the fully connected layer uses a random deactivation function during training.
4. The method according to claim 1, characterized in that The second model includes an input layer, a mapping layer, an enhancement layer and an output layer, and the enhancement processing is implemented by the second model, including: Initializing the input layer; Mapping the low-resolution image through the mapping layer to obtain a feature vector Z of a feature node; Performing enhancement processing on the feature vector of the feature node through the enhancement layer to obtain a feature vector H of the enhanced node; The feature vector of the feature node and the feature vector of the enhanced node are cascaded through the output layer S=[Z|H]W O , an enhanced image is obtained, where W0 represents the enhancement weight coefficient.
5. The method according to claim 4, characterized in that The initialization process of the input layer includes initializing the random weight value W of each feature node i ei and the mapping feature bias value β ei , the mapping process include: According to the random weight value W corresponding to the feature node i ei and the mapping feature bias value β ei , determine the characteristic vector Z of the characteristic node i .
6. The method according to claim 4, characterized in that The initialization process of the input layer includes initializing the enhanced weight value W of each enhanced node j hj and the enhancement bias value β hj , the enhancement processing includes: According to the characteristic vector Z of the characteristic node i , the enhanced weight value W corresponding to the enhanced node j hj and the enhanced node bias value β hj , determine the eigenvector H of the enhanced node j .
7. The method according to claim 4, characterized in that The method also includes: training the second model using a second training data set to determine the enhanced weight coefficient W0.
8. The method according to claim 1, characterized in that Before acquiring the combustion image, the method further includes: Pushing the waste to be processed into the drying section of the incinerator, injecting combustion-supporting air into the drying section, heating the waste to be processed by the incinerator, evaporating the water in the waste to be processed to obtain dry waste; The dry garbage then enters the combustion section of the incinerator for combustion.
9. The method according to claim 2, characterized in that The first convolution layer and the second convolution layer use a 5*5 convolution kernel with a step size of 4; The pooling layer adopts maximum pooling.
10. The method according to claim 7, characterized in that During training, the enhanced weight coefficient W0 is determined by pseudo-inverse calculation.
Citation Information
Patent Citations
Image background definition detection method based on deep learning
CN105825511A
Urban solid waste incineration process combustion condition recognition method based on flame image color feature extraction
CN110222633A
Flame real-time detection method for inspection robot
CN111860143A
Medical image reconstruction technology method and device, storage medium and electronic equipment
CN111862251A
Incineration state automatic discrimination method and device based on digital image processing
CN113869113A
