Waste charging monitoring system and method based on image processing

Abstract

The invention relates to the technical field of image processing, in particular to a waste charging monitoring system and method based on image processing. The method comprises the following steps: dividing an obtained infrared image after waste enters a furnace to obtain a plurality of sub-regions; performing super-pixel segmentation on the infrared image by taking the center of each sub-region as a seed point to obtain a plurality of super-pixel blocks; calculating the variance of the temperature mean values in all the superpixel blocks; when the variance is larger than a threshold value, the temperature distribution in the furnace is not uniform, and an alarm is given; according to the scheme, the temperature in the furnace can be accurately detected.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a system and method for monitoring waste entering a furnace based on image processing. Background Art

[0002] As the amount of urban domestic garbage and industrial waste increases, traditional garbage disposal methods such as landfill not only occupy a large amount of land resources, but may also cause problems such as groundwater pollution. Therefore, waste incineration technology has emerged and has become an important waste disposal method.

[0003] During the incineration process of waste materials, it is necessary to take infrared images in real time to monitor the temperature distribution in the furnace, so as to find abnormal areas in time, prevent overheating and uneven temperature, and improve the efficiency and safety of incineration. At the same time, the infrared images taken can also help detect whether the equipment and pipelines in the furnace are faulty, and abnormal hot spots or abnormal temperatures can be repaired in time to avoid production interruptions and safety problems.

[0004] Therefore, real-time monitoring of the temperature inside the incinerator through infrared images is helpful to improve the incineration efficiency and ensure the safety of waste incineration.

[0005] However, due to the diffusion and attenuation of infrared radiation energy and the imaging capability limit of the infrared sensor array, the resolution of infrared images is limited, and the imaging results of the target usually show weak characteristics (i.e., few pixels, weak intensity, and low contrast), while the background is complex and blurred, and mixed with more clutter components. Therefore, in the prior art, superpixel segmentation of infrared images is performed to filter out non-target components as much as possible, while identifying potential target components.

[0006] In the related technology, for example, a patent application document with publication number CN116109662A discloses a superpixel segmentation method for infrared images, which discloses that by calculating the image smoothness, the position of the maximum smoothness in the scale of a specified superpixel grid is obtained as the starting position of the superpixel clustering center, clustering is performed by setting segmentation control parameters, and background pixels and target pixels are distinguished, and the clustering results are updated until a superpixel segmentation result is obtained.

[0007] In the above scheme, when performing superpixel segmentation, the number of seed points is set manually; however, there may be large differences in real-time images under different shooting environments. For example, when there are fewer details in the image, the difference of the entire image is small, and a smaller number of seed points can meet the image segmentation effect. However, if the number of seed points is set too much, the computational complexity of the segmentation will increase, and the segmentation effect will not be good. For example, when the image is extremely rich in details, if the number of seed points is set too small, the segmentation result will be relatively rough and cannot accurately reflect the details in the image.

[0008] Therefore, how to reasonably set the number of seed points when performing superpixel segmentation of an image affects the image segmentation effect, and thus affects the problem of temperature monitoring in the incinerator. Summary of the invention

[0009] The purpose of the present invention is to propose a waste incinerator monitoring system and method based on image processing, so as to solve the problem that the infrared image segmentation cannot be accurately performed in the prior art, thereby causing inaccurate temperature monitoring in the incinerator; to this end, the present invention provides solutions in the following two aspects.

[0010] In a first aspect, the present invention provides a method for monitoring waste entering a furnace based on image processing, comprising: Divide the acquired infrared image of the incinerator during the incineration process to obtain multiple sub-areas; Taking the center of each sub-region as a seed point, the infrared image is segmented into superpixels to obtain multiple superpixel blocks; the variance of the temperature mean in all superpixel blocks is calculated; when the variance is greater than a threshold, the temperature distribution in the furnace is uneven, and an alarm is issued; The sub-regions are: obtaining closed edges and non-closed edges in the infrared image; dividing the non-closed edges into important edges and non-important edges; extending all important edges until they intersect with the boundary or any edge of the infrared image and stop extending; the closed edges and the extended important edges divide the infrared image into multiple parent regions; If the unevenness of the parent area is less than the set value, the parent area is used as the child area; if the unevenness of the parent area is greater than or equal to the set value, the remaining area except the dense area is used as the new parent area, and the new unevenness calculation is continued until the new unevenness is less than the set value, and all the child areas of each parent area are obtained; the unevenness for: ; , , are the grayscale entropy, area, and number of edge points of the i-th parent region, respectively. , , are respectively the grayscale entropy, area, and number of edge points of the dense area in the i-th parent area; the dense area is the area with the largest density of non-important edges in the parent area.

[0011] The above scheme first screens out the non-closed edges in the infrared image and analyzes the non-closed edges to obtain important edges, then roughly segments the infrared image by extending the important edges and closed edges to obtain the parent area; then analyzes the non-important edges in each parent area to determine the unevenness of each parent area, so as to achieve a secondary fine segmentation of each parent area, and finally obtain the child area. That is, by dividing the acquired infrared image of the incinerator twice, the divided sub-areas can be finely obtained, and the center of each sub-area is used as the seed point to perform super-pixel segmentation, and accurate super-pixel blocks can be obtained, which provides accurate data for the temperature monitoring of the incinerator during the incineration process.

[0012] Optionally, the specific process of extending all important edges is: Obtain the fitting curve of the edge segment corresponding to the minimum distance between any endpoint on each important edge and each corner point on it; Taking the endpoint on the edge segment as the starting point, expand along the extension direction of the fitting curve.

[0013] Optionally, the dense area is an area with the largest density of non-important edges in the parent area, including: Get all non-significant edges in each parent region; The mean shift clustering algorithm is used to cluster all non-important edges and obtain multiple category clusters; Obtain all outermost edge points in each category cluster, and use the area enclosed by the outermost edge points connected in sequence as a closed area; the outermost edge point is the edge point farthest from the cluster center on the ray where the cluster center and any edge point are connected; The ratio of the number of edge points in the category cluster to the area of ​​the corresponding closed region is taken as the clustering density; The closed area with the largest aggregation density is selected as the dense area corresponding to the parent area.

[0014] The above solution can accurately obtain the dense areas in each parent area.

[0015] Optionally, the process of obtaining the important edges and the unimportant edges is: The K-means algorithm is used to cluster all non-closed edges to obtain the first cluster and the second cluster; the first cluster is the cluster with a larger mean length of non-closed edges, and the second cluster is the cluster with a larger mean length of non-closed edges; All non-closed edges in the first cluster and non-closed edges in the second cluster whose importance is greater than or equal to a set threshold are recorded as important edges; non-closed edges in the second cluster whose importance is less than the set threshold are recorded as unimportant edges; the importance is positively correlated with the length and gradient of each edge in the second cluster.

[0016] The above scheme can accurately screen out important edges from non-closed edges.

[0017] Optionally, the importance is: ; represents the inner edge of the second cluster The importance of For the edge Length, is the minimum length of the edge within the first cluster, represents the inner edge of the second cluster Previous The gradient value of a pixel, The inner edge of the second cluster The total number of pixels on is the maximum value of the average gradient of all edges.

[0018] The above scheme provides a method to accurately calculate the importance.

[0019] Optionally, when performing superpixel segmentation, any pixel point is taken as a target point, and the distance from the target point to the seed point is calculated as: ; in, , , They are the distance, color distance, and spatial distance from the target point to the seed point, respectively. is the maximum color distance within the class, is the maximum spatial distance within the class, is the importance of the jth edge point on the line connecting the target point to the seed point, t is the total number of edge points on the line, represents the spatial distance from the jth edge point on the line to the target point; When the edge where the edge point is located belongs to the first cluster, its importance is 1; when the edge where the edge point is located belongs to the second cluster, its importance is the importance of the edge where the edge point is located.

[0020] In the above scheme, the spatial distance between the target point and the seed point is corrected by introducing the importance and the spatial difference distance between the target point and the seed point, so that the target points crossing multiple edges or crossing edges with greater importance are not assigned to the same area as much as possible. Otherwise, they are assigned to the same area as much as possible.

[0021] Optionally, the process of obtaining the important edges and the unimportant edges is: The length of each non-closed edge is obtained, and the edge whose length is greater than or equal to the first length threshold is recorded as an important edge; the edge whose length is less than the first length threshold is recorded as a non-important edge.

[0022] Optionally, the closed edges and the non-closed edges are obtained by using a Canny edge detection algorithm.

[0023] In a second aspect, a waste furnace monitoring system based on image processing comprises: processor; The memory stores computer instructions for monitoring the entry of waste into a furnace based on image processing. When the computer instructions are executed by the processor, the system executes the above-mentioned method for monitoring the entry of waste into a furnace based on image processing.

[0024] The beneficial effects of the present invention are: The solution of the present invention can make full use of the details and edge information in the image, realize the adaptive setting of the number and position of seed points, enable the superpixel segmentation algorithm to better adapt to the edge characteristics of the image, improve the accuracy and efficiency of the superpixel segmentation algorithm, maintain edge continuity, and reduce the occurrence of over-segmentation and under-segmentation, thereby providing accurate temperature data for subsequent temperature monitoring in the incinerator. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 A flowchart of the steps of a method for monitoring waste entering a furnace based on image processing in this embodiment is schematically shown; Figure 2 The structural block diagram of a waste furnace entry monitoring system based on image processing in this embodiment is schematically shown. DETAILED DESCRIPTION

[0026] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] like Figure 1 As shown, a method for monitoring waste entering a furnace based on image processing in this embodiment includes the following steps: Step S1, obtaining an infrared image of an incinerator during the process of incinerating waste, specifically, an infrared image of the interior of the incinerator taken in real time by an infrared camera.

[0028] After the infrared image is acquired, grayscale processing and denoising processing are performed on the infrared image to obtain a processed infrared image.

[0029] Step S2, dividing the infrared image into regions to obtain a plurality of sub-regions.

[0030] The process of acquiring multiple sub-regions includes steps S21 to S25, which are specifically as follows: Step S21, acquiring multiple edges in the infrared image.

[0031] In this embodiment, the Canny edge detection algorithm is used to perform edge detection to obtain multiple edges, wherein the multiple edges include closed edges and non-closed edges.

[0032] Step S22: Divide all non-closed edges into important edges and unimportant edges.

[0033] Since the closed edges constitute a closed area, only the non-closed edges are analyzed in this embodiment.

[0034] In one embodiment, the method for obtaining important edges and unimportant edges is: The length of each non-closed edge is obtained, and the edge whose length is greater than or equal to the first length threshold is recorded as an important edge; the edge whose length is less than the first length threshold is recorded as a non-important edge.

[0035] The first length threshold mentioned above is the third quartile; of course, it can also be determined according to actual conditions.

[0036] The third quartile is obtained by obtaining the lengths of all non-closed edges, sorting them from large to small, and selecting the 75% value in the sorting as the third quartile.

[0037] The length of the above non-closed edge is the total number of pixels on the edge.

[0038] In another embodiment, the method for obtaining important edges and unimportant edges may also be: First, all non-closed edges are clustered to obtain the first cluster and the second cluster; the first cluster is a cluster with a larger mean edge length, and the second cluster is a cluster with a larger mean edge length.

[0039] Specifically, the lengths of all non-closed edges are clustered using a K-means clustering algorithm. Since the K-means clustering algorithm is a prior art, it will not be described in detail here.

[0040] In order to avoid the situation where the number of edges in each cluster is too sparse during clustering, the number of edges in the first cluster and the second cluster should theoretically be no less than 3.

[0041] It should be noted that the “first” and “second” mentioned above are only used to distinguish two clusters and do not have any special limiting meaning.

[0042] Secondly, the importance of each non-closed edge in the second cluster is calculated.

[0043] Specifically, the importance is: ; represents the inner edge of the second cluster The importance of For the edge Length, is the minimum length of the inner edge of the first cluster, represents the inner edge of the second cluster Previous The gradient value of a pixel, The inner edge of the second cluster The total number of pixels on is the maximum value of the average gradient of all edges.

[0044] Since the length of any open edge in the first cluster is greater than that of any open edge in the second cluster, The value of is less than 1. Length and the average gradient The larger the value, the greater the edge The weaker the isolation, the more marginal The smaller the possibility of noise, the smaller the edge The greater the possibility of real texture or edge in the image, the more important the target edge is. The bigger.

[0045] At the same time, the edge Length and the average gradient The larger the value, the more it can explain the edge The continuity is strong and the edge The greater the difference between the two sides, the greater the edge The more likely it is to reflect the regional boundary features in the image, the greater the importance of the target edge. Then, all the non-closed edges in the first cluster and the non-closed edges in the second cluster whose importance is greater than or equal to the set threshold are recorded as important edges; the non-closed edges in the second cluster whose importance is greater than or equal to the set threshold are recorded as unimportant edges.

[0046] In this embodiment, all edges in the first cluster are recorded as important edges; and edges in the second cluster whose importance is greater than or equal to a set threshold are also recorded as important edges.

[0047] The threshold value is set as 0.7; of course, it can be set according to the actual situation.

[0048] The purpose of dividing the non-closed edges is that when the length of the non-closed edge is large (i.e., it belongs to the first cluster or the second cluster), the edge may belong to the real texture or edge, and therefore, it is more important; while the non-closed edge with a relatively small length may be noise or other information.

[0049] Step S23, extending each important edge until it intersects with the image boundary or any edge, and stopping the extension. The different extended important edges and closed edges divide the infrared image into multiple closed areas, and each closed area is recorded as a parent area.

[0050] Specifically, the method for extending each important edge is: Get the fitting curve of the edge segment on each important edge, take the upper endpoint of the edge segment as the starting point, and extend it along the direction of the fitting curve. The edge segment is the local edge corresponding to the minimum distance between any endpoint on each important edge and all corner points on it. The corner points are obtained by Harris corner point detection algorithm.

[0051] Exemplarily, when there are two endpoints on an edge, it is necessary to obtain edge segments between the two endpoints and their nearest corner points respectively, that is, the two endpoints correspond to two edge segments.

[0052] In this embodiment, the least squares method is used to perform curve fitting on the edge segment to obtain a fitting curve, and all the acquired endpoints are iteratively extended at the same time. Each iteration process incorporates the next pixel point of the fitting curve into the edge at the endpoint, and the extension step is 1 pixel point. The extended new pixel point is used as the endpoint of the next iteration; when the extended edge reaches the image boundary or intersects with any edge, the extension of the edge is stopped.

[0053] As another embodiment, the condition for stopping the extension can also be that when the extended edge enters the window of any pixel point on any edge, the extension is stopped. At this time, the point closest to the endpoint on any edge is included in the edge, that is, the point closest to the endpoint on any edge is the next pixel point.

[0054] When all important edges stop extending, that is, the entire extension process stops, the entire image is divided into closed regions by the extended important edges and closed edges. Thus, the first region division of the entire image is completed, and multiple parent regions are obtained.

[0055] Step S24, calculating the unevenness of each parent region.

[0056] The process of obtaining the unevenness in this embodiment is as follows: First, obtain the dense regions within each parent region.

[0057] In the dense area, all non-important edges in each parent area are clustered by using the mean shift clustering algorithm to obtain multiple category clusters, and the closed area of ​​the category cluster with the largest clustering density is taken as the dense area.

[0058] When clustering, the centroid of the non-important edges in each parent region is calculated, and all non-important edges are clustered according to the position information of the centroid (the horizontal and vertical coordinate values ​​of the centroid) using the mean shift clustering algorithm. Since the mean shift clustering algorithm is an existing technology, it will not be described in detail here.

[0059] In one embodiment, the enclosed area of ​​the category cluster with the largest clustering density is obtained as follows: Get all the outermost edge points in each category cluster, and take the area enclosed by the outermost edge points connected in sequence as the closed area; take the ratio of the number of edge points in the category cluster to the area of ​​the corresponding closed area as the clustering density; select the closed area with the largest clustering density as the dense one.

[0060] The outermost edge point is the edge point farthest from the cluster center on the ray where the line connecting the cluster center and any edge point is located, with the cluster center as the endpoint. At this time, there is no edge point in the extension direction of the farthest edge point.

[0061] The above-mentioned closed area is actually the maximum circumscribed polygon of the category cluster, that is, the area obtained by the maximum circumscribed polygon surrounding all non-important edges in the category cluster, and combined with the number of all edge points in the category cluster to obtain the clustering density of each category cluster.

[0062] Exemplarily, when the ray connecting the cluster center and edge point a on edge A also intersects edge point b on edge B, the distance between edge point b and the cluster center is large, and edge point b is the outermost edge point.

[0063] The above area is the number of all pixels in the closed area.

[0064] In another embodiment, the closed area of ​​the category cluster with the largest clustering density may also be an area enclosed by the circumscribed polygon of the category cluster when the number of edge points in each category cluster is the largest.

[0065] The above-mentioned dense area must contain at least 30% of non-important edges to avoid the dense area being too small and causing over-segmentation.

[0066] Secondly, according to the parent area and the corresponding dense area, the unevenness of the corresponding parent area is obtained.

[0067] In this embodiment, the degree of unevenness in each parent region is calculated based on entropy and edge information density. , the formula is as follows: ; , , are the grayscale entropy, area, and number of edge points of the i-th parent region, respectively. , , are the grayscale entropy, area, and number of edge points of the dense region in the i-th parent region, respectively.

[0068] The above entropy is the entropy of the gray level of the obtained parent region or dense region. The gray level can be divided into 16 gray levels with gray values ​​0 to 255, that is, 16 gray values ​​are divided into 1 gray level, for example, gray level 1 (gray value 0 to 15), gray level 2 (gray value 16 to 31), and so on.

[0069] As another implementation, the grayscale level may also be a grayscale value, that is, the grayscale value of a pixel corresponds to a grayscale level.

[0070] represents the edge information density of the dense area in the i-th parent area, Represents the edge information density of the i-th parent region.

[0071] when When the value is larger, it means that the distribution disorder of gray levels in the dense area and the parent area is quite different, which means that the dense area may have independent characteristics different from the corresponding parent area. The larger the The larger the value, the smaller the area, indicating that the dense area contains more edge information. The edge information density in the dense area is greater than the edge information density of the parent area, indicating the unevenness of the parent area. The bigger.

[0072] Step S25: acquiring a plurality of child regions based on the unevenness of the parent region.

[0073] Specifically, the step of obtaining multiple sub-regions includes steps S251 to S253, which are specifically: Step S251, if the unevenness of the parent region is less than a set value, the parent region is used as a child region; Step S252: if the unevenness of the parent region is greater than or equal to the set value, the dense region is used as the child region, and the remaining region except the dense region is used as the new parent region; and the calculation of the unevenness of the new parent region is continued; Step S253, if the new unevenness is still greater than or equal to the set value, repeat step S252; iterate multiple times until the final unevenness is less than the set value, and then stop to obtain all sub-areas.

[0074] For example, when , it is considered that the corresponding parent area needs to be split again, and the current parent area is split into two parts, namely the dense area and the remaining area.

[0075] The above setting value It is 0.3, and of course it can be set according to actual conditions.

[0076] The above calculation of unevenness is to split the original parent region multiple times until the parent region does not need to be split again, that is, When , the iteration stops, and the segmented sub-region can be obtained. Thus, the division of the sub-region in the parent region is completed, that is, the division of the secondary region of the entire image is completed.

[0077] The sub-areas include the parent area that does not need to be divided and the sub-areas within the divided parent area.

[0078] Step S3, taking the center of each sub-region as a seed point, performing super-pixel segmentation on the infrared image to obtain multiple super-pixel blocks; calculating the variance of the temperature in all super-pixel blocks; when the variance is greater than a threshold, the temperature distribution in the furnace is uneven, and an alarm is issued.

[0079] In this embodiment, the infrared image is segmented into superpixels using the center of each sub-region as a seed point to obtain a plurality of superpixel blocks.

[0080] Specifically, the specific process of performing superpixel segmentation on the infrared image is as follows: First, the center of each sub-region is used as the seed point. A seed point is set in each sub-region, and the position of the seed point is the centroid (center) of the region, but the color information of the seed point is determined by the average color information of all the remaining pixels in the sub-region except the edge points.

[0081] Secondly, for the seed point of each sub-region, the K-means algorithm is used to classify the surrounding pixels into the region.

[0082] Among them, each pixel point whose distance to a seed point is being calculated is called a target point, and the spatial distance is corrected according to the cross-edge degree of the line connecting the target point to a seed point. That is, the calculation formula for the distance from the target point to the seed point is as follows: ; in, , , Respectively represent the distance, color distance, and spatial distance from the target point to the seed point. represents the maximum color distance within the class, represents the maximum spatial distance within a class, is the importance of the jth edge point on the line connecting the target point to the seed point, t is the total number of edge points on the line, represents the spatial distance from the jth edge point on the line to the target point, and min() is the minimum function.

[0083] in, is the distance metric formula in the existing superpixel segmentation algorithm. , , , The calculation method and meaning of are the same as those in the existing superpixel segmentation algorithm and will not be described in detail here.

[0084] When the edge where the edge point is located belongs to the first cluster, its importance is 1; when the edge where the edge point is located belongs to the second cluster, its importance is the importance of the edge where the edge point is located.

[0085] In the above distance metric formula, the cross-edge degree is added The spatial distance is corrected. When there are multiple edge pixels on the line connecting the target point to the seed point, or there are edge points with greater importance on the line, it means that there are multiple edges or edges with greater importance between the target point and the seed point. That is, if the cost of dividing the target point into the area of ​​the seed point is high, it is necessary to cross multiple edges or cross edges with greater importance, indicating that it is likely that the target point and the seed point do not belong to the same area. At this time, the spatial distance from the target point to the seed point is increased, making the distance from the target point to the seed point larger, thereby reducing the possibility of the target point being assigned to the area of ​​the seed point.

[0086] Then, superpixels are merged. Specifically, superpixels are merged based on the distance from the pixel to the nearest seed point and the similarity between adjacent superpixels.

[0087] Furthermore, iterative optimization is performed to obtain multiple super pixel blocks. The above steps are repeated until a preset number of iterations is reached or convergence is achieved, thereby obtaining multiple super pixel blocks after division.

[0088] Thus, multiple iterations of superpixel segmentation are performed to complete the segmentation of the entire image.

[0089] In this embodiment, the temperature mean of each area obtained after segmentation is statistically calculated, and the variance of the temperature mean of all areas is used to determine whether the temperature distribution in the furnace is uniform. The formula is as follows: ; is the variance of the mean temperature of all superpixel blocks, For the The average temperature of superpixel blocks, is the average temperature value of all superpixel blocks, is the number of superpixel blocks.

[0090] After obtaining the variance of the temperature mean of the infrared image, the variance is compared with the temperature threshold to determine whether the temperature distribution in the furnace is uniform. Specifically, when the variance is greater than the temperature threshold, it proves that the temperature distribution in the furnace is uneven.

[0091] The temperature threshold may be set to 300°C, and may also be set according to actual conditions.

[0092] In this embodiment, when the temperature distribution in the furnace is uneven, an alarm is also required to promptly remind the staff that abnormal conditions may occur in the furnace and to perform maintenance.

[0093] The solution of the present invention divides the infrared image twice to obtain multiple regions, and then uses the cross-edge degree of the line connecting the target point to the seed point in each region to correct the spatial distance, complete the calculation of the distance from the target point to the seed point, and iterates multiple times to complete the entire superpixel segmentation. According to the variance of the temperature of all regions, it is judged whether the temperature distribution in the furnace is uniform.

[0094] The present invention also provides a waste furnace monitoring system based on image processing. Figure 2 As shown, the monitoring system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned method for monitoring waste entering a furnace based on image processing according to the present invention is implemented.

[0095] The monitoring system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0096] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented by computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0097] In the description of this specification, “plurality” means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0098] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for monitoring waste entering a furnace based on image processing, characterized in that: include: Divide the acquired infrared image of the incinerator during the incineration process to obtain multiple sub-areas; Taking the center of each sub-region as a seed point, the infrared image is segmented into superpixels to obtain a plurality of superpixel blocks; the variance of the mean temperature in all superpixel blocks is calculated; when the variance is greater than a threshold, the temperature distribution in the furnace is uneven, and an alarm is issued; The sub-regions are: obtaining closed edges and non-closed edges in the infrared image; dividing the non-closed edges into important edges and non-important edges; extending all important edges until they intersect with the boundary or any edge of the infrared image; the closed edges and the extended important edges divide the infrared image into multiple parent regions; If the unevenness of the parent area is less than the set value, the parent area is used as the child area; if the unevenness of the parent area is greater than or equal to the set value, the remaining area except the dense area is used as the new parent area, and the new unevenness calculation is continued until the new unevenness is less than the set value, and all the child areas of each parent area are obtained; the unevenness for: ; , , are the grayscale entropy, area, and number of edge points of the i-th parent region, respectively. , , are respectively the grayscale entropy, area, and number of edge points of the dense area in the i-th parent area; the dense area is the area with the largest density of non-important edges in the parent area.

2. The method for monitoring waste entering a furnace based on image processing according to claim 1, characterized in that: The specific process of extending all important edges is: Obtain the fitting curve of the edge segment corresponding to the minimum distance between any endpoint on each important edge and each corner point on it; Taking the endpoint on the edge segment as the starting point, expand along the extension direction of the fitting curve.

3. The method for monitoring waste entering a furnace based on image processing according to claim 1, characterized in that: The dense area is the area with the largest density of non-important edges in the parent area, including: Get all non-significant edges in each parent region; The mean shift clustering algorithm is used to cluster all non-important edges and obtain multiple category clusters; Obtain all outermost edge points in each category cluster, and use the area enclosed by the outermost edge points connected in sequence as a closed area; the outermost edge point is the edge point farthest from the cluster center on the ray where the line connecting the cluster center and any edge point is located with the cluster center as the endpoint; The ratio of the number of edge points in the category cluster to the area of ​​the corresponding closed region is taken as the clustering density; The closed area with the largest aggregation density is selected as the dense area corresponding to the parent area.

4. The method for monitoring waste entering a furnace based on image processing according to claim 1, characterized in that: The process of obtaining the important edges and the unimportant edges is as follows: The K-means algorithm is used to cluster all non-closed edges to obtain the first cluster and the second cluster; the first cluster is the cluster with a larger mean length of non-closed edges, and the second cluster is the cluster with a larger mean length of non-closed edges; All non-closed edges in the first cluster and non-closed edges in the second cluster whose importance is greater than or equal to a set threshold are recorded as important edges; The non-closed edges in the second cluster whose importance is less than a set threshold are recorded as non-important edges; the importance is positively correlated with the length and gradient of each edge in the second cluster.

5. The method for monitoring waste entering a furnace based on image processing according to claim 4, characterized in that: The importance levels are: ; represents the inner edge of the second cluster The importance of For the edge Length, is the minimum length of the edge within the first cluster, represents the inner edge of the second cluster Previous The gradient value of a pixel, The inner edge of the second cluster The total number of pixels on is the maximum value of the average gradient of all edges.

6. The method for monitoring waste entering a furnace based on image processing according to claim 5, characterized in that: It also includes that when performing superpixel segmentation, any pixel point is taken as the target point, and the distance from the target point to the seed point is calculated as: ； in, , , They are the distance, color distance, and spatial distance from the target point to the seed point, respectively. is the maximum color distance within the class, is the maximum spatial distance within the class, is the importance of the jth edge point on the line connecting the target point to the seed point, t is the total number of edge points on the line, represents the spatial distance from the jth edge point on the line to the target point; When the edge where the edge point is located belongs to the first cluster, its importance is 1; when the edge where the edge point is located belongs to the second cluster, its importance is the importance of the edge where the edge point is located.

7. The method for monitoring waste entering a furnace based on image processing according to claim 1, characterized in that: The process of obtaining the important edges and the unimportant edges is as follows: The length of each non-closed edge is obtained, and the edge whose length is greater than or equal to the first length threshold is recorded as an important edge; the edge whose length is less than the first length threshold is recorded as a non-important edge.

8. The method for monitoring waste entering a furnace based on image processing according to claim 1, characterized in that: The closed edges and the non-closed edges are obtained by using the Canny edge detection algorithm.

9. A waste furnace monitoring system based on image processing, characterized in that: include: processor; A memory storing computer instructions for monitoring waste entering a furnace based on image processing. When the computer instructions are executed by the processor, the system executes a method for monitoring waste entering a furnace based on image processing according to any one of claims 1 to 8.

Images