Method and system for sorting and identifying solid waste for recycling

Abstract

The application relates to the technical field of image recognition, in particular to a solid waste classification and recognition method and system for recycling. The method comprises the following steps: acquiring a garbage image, determining a garbage area through image analysis, determining a characteristic value according to a circular fitting feature and a gray value, determining a mutation amplitude according to the change of the characteristic value, determining a candidate mutation time node of garbage change caused by a conveying belt according to the value of the mutation amplitude, screening a target time node of garbage state mutation on the conveying belt and cutting, determining a time period, determining a state change index according to the change of the overall mutation amplitude in each time period, combining the characteristic value and the state change index to determine a possibility index of the garbage area being a pop can area, and screening the pop can area according to the possibility index of each garbage area. The application can improve the stability of pop can classification and recognition, and the classification and recognition effect of the pop can is better and more accurate.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and more specifically to a method and system for classifying and identifying solid waste for recycling. Background Technology

[0002] Waste recycling is a crucial part of achieving the Sustainable Development Goals. The aluminum recycling process is over 95% more energy-efficient than the primary aluminum production process, thus possessing high recycling value, especially for aluminum cans. However, because aluminum products are non-magnetic, aluminum cans cannot be separated by magnetic separators. Furthermore, the outer shell of aluminum cans is relatively soft and easily deformed, readily flattening them on the conveyor belt, making identification difficult during the screening process. Especially after the initial screening by magnetic separators and air separators, the remaining waste contains not only aluminum cans but also other waste materials, making the secondary screening of flattened aluminum cans even more complex and challenging.

[0003] In related technologies, manual screening or specific image recognition analysis based on the features of the overall image is used. However, the flattened can combined with the bumpy conveyor belt causes the can's state to change, which in turn leads to unstable can feature recognition, resulting in poor can classification and recognition performance and low accuracy. Summary of the Invention

[0004] To address the technical problem in related technologies where the flattened aluminum cans combined with a bumpy conveyor belt cause changes in the can's state, leading to unstable can feature recognition and poor classification accuracy, this invention provides a method and system for classifying and identifying recyclable solid waste. The specific technical solution adopted is as follows:

[0005] This invention proposes a method for classifying and identifying solid waste for recycling, the method comprising:

[0006] Images of waste on the conveyor belt at different sampling times are acquired, and image analysis is performed to determine different waste areas. Based on the matching features of each waste area with the circle and the gray value, the characteristic value of the waste area belonging to the aluminum can is determined.

[0007] Based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined; based on the value of the abrupt change amplitude, the candidate abrupt change time nodes that cause changes in waste due to the conveyor belt are determined.

[0008] Based on the candidate mutation time nodes corresponding to each waste area, the target time nodes that cause mutations in the state of waste on the conveyor belt are selected. Using the target time nodes as the cutting points, the different time periods of each waste area on the conveyor belt are determined. Based on the changes in the overall mutation amplitude in each time period, the state change index of the waste area in each time period is determined.

[0009] By combining the characteristic values ​​and the state change indicators, a probability index is determined that the waste area is an aluminum can area, and aluminum can areas are screened based on the probability index of each waste area.

[0010] Furthermore, the step of determining the characteristic value that a waste area belongs to an aluminum can based on the matching features of each waste area and the fitted circle, as well as the gray value, includes:

[0011] Ellipse fitting is performed on the waste area, and the distance between the two foci of the fitted ellipse is determined as the circle matching distance;

[0012] The negative of the circle matching distance is normalized by taking the maximum and minimum values ​​to obtain the circle matching index;

[0013] Calculate the average grayscale value of all pixels in the garbage area after grayscale conversion to obtain the grayscale feature index;

[0014] Calculate the product of the circle matching index and the grayscale feature index, and normalize the maximum and minimum values ​​to obtain the characteristic value.

[0015] Furthermore, based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined, including:

[0016] Determine any sampling time as the target time, calculate the difference between the characteristic values ​​of the target time and its previous time as the first difference, and calculate the difference between the characteristic values ​​of the target time and its subsequent time as the second difference;

[0017] The normalized value of the sum of the first difference and the second difference is taken as the mutation amplitude at the target time.

[0018] Further, based on the numerical value of the mutation amplitude, candidate mutation time points for the conveyor belt causing changes in waste are determined, including:

[0019] The time nodes where the mutation magnitude is greater than a preset magnitude threshold are selected as candidate mutation time nodes.

[0020] Furthermore, based on the candidate mutation time nodes corresponding to each waste area, the target time nodes where waste state mutations occur on the conveyor belt are selected, including:

[0021] Determine the frequency of the sampling time corresponding to the candidate mutation time nodes in all waste areas, and select the preset number of candidate mutation time nodes with the highest frequency as the target time nodes.

[0022] Furthermore, based on the changes in the overall magnitude of abrupt changes within each time period, indicators of the state changes of the waste area within each time period are determined, including:

[0023] The stability of mutations within a time period is determined by the dispersion of all mutation amplitudes within the same time period.

[0024] Calculate the negative of the mean of all mutation amplitudes within the same time period, and normalize the maximum and minimum values ​​to use them as numerical feature indicators.

[0025] The product of numerical characteristic indicators and mutation stability is used as an indicator of the state change of the waste area within the corresponding time period.

[0026] Furthermore, based on the dispersion of all mutation amplitudes within the same time period, the mutation stability within that time period is determined, including:

[0027] Calculate the negative of the standard deviation of all mutation amplitudes within the same time period, and normalize the maximum and minimum values ​​to obtain mutation stability.

[0028] Furthermore, by combining the characteristic values ​​and the state change indicators, the probability of the waste area being a can area is determined, including:

[0029] Calculate the mean of the characteristic values ​​at all sampling times within the same time period, and use it as the segment characteristic coefficient;

[0030] Calculate the product of the segment characteristic coefficient and the state change index within the same time period, and use it as the influence index for the time period;

[0031] The average of all influencing indicators over all time periods is used as an indicator of the probability that the corresponding waste area is an area containing aluminum cans.

[0032] Furthermore, can areas are screened based on the probability indicators for each waste area, including:

[0033] Waste areas with a probability index greater than a preset probability threshold are designated as aluminum can areas.

[0034] On the other hand, the present invention also includes a solid waste classification and identification system for recycling, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any of the foregoing.

[0035] The present invention has the following beneficial effects:

[0036] This invention acquires images of waste on a conveyor belt at different sampling times, performs image analysis to determine different waste areas, and determines the characteristic value of each waste area belonging to an aluminum can based on the matching features of each waste area with the circle and the gray value. This characteristic value characterizes the aluminum can feature of the waste area itself, so subsequent numerical analysis and change analysis can be performed based on the characteristic value. Then, based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined. Based on the numerical value of the abrupt change amplitude, candidate abrupt change time nodes causing waste changes due to the conveyor belt are determined. Based on the candidate abrupt change time nodes corresponding to each waste area, target time nodes causing abrupt changes in the waste state on the conveyor belt are selected. Using the target time nodes as cutting points, different time periods for each waste area on the conveyor belt are determined. The division of time periods effectively avoids the "state reset" effect caused by the bumpy rollers, improving the accuracy and reliability of the overall numerical analysis. Based on the changes in the overall abrupt change amplitude in each time period, the state change index of the waste area in each time period is determined. Combining characteristic values ​​and state change indexes, the probability index of the waste area being an aluminum can area is determined. Based on the probability index of each waste area, aluminum can areas are selected. In summary, this invention can combine shape features and grayscale features to perform characteristic value analysis, and further, based on characteristic abrupt changes and the influence of rollers on the conveyor belt, determine more accurate and reliable probability indexes, improving the stability of aluminum can classification and recognition, resulting in better and more accurate aluminum can classification and recognition. Attached Figure Description

[0037] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a method for classifying and identifying solid waste for recycling, provided in one embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of an aluminum can in a parallel state and a state biased towards vertical, according to an embodiment of the present invention.

[0040] Figure 3 This is a schematic diagram of the conveyor belt roller area provided in one embodiment of the present invention. Detailed Implementation

[0041] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a solid waste classification and identification method and system for recycling proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0042] Unless otherwise defined, 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 invention pertains.

[0043] The following description, in conjunction with the accompanying drawings, details a specific scheme for a method of classifying and identifying solid waste for recycling provided by the present invention.

[0044] Please see Figure 1 The diagram illustrates a flowchart of a method for classifying and identifying solid waste for recycling, according to an embodiment of the present invention. The method includes:

[0045] S101: Acquire images of waste on the conveyor belt at different sampling times, perform image analysis to determine different waste areas; determine the characteristic value of the waste area belonging to the aluminum can based on the matching features of each waste area with the circle fitting and the gray value.

[0046] The specific scenario corresponding to this embodiment of the invention is a waste sorting scenario. This solution mainly focuses on the sorting of aluminum cans. In environments such as waste treatment sites, waste sorting lines, and waste recycling stations, the recycling of aluminum products such as aluminum cans is quite difficult. Traditional recycling methods such as magnetic separation and manual screening are inefficient, and aluminum cans are often misjudged or missed due to changes in shape and different physical states. Therefore, image analysis can be combined to further classify and identify aluminum cans.

[0047] When waste is recycled, the magnetic separator separates the magnetic metal waste that is originally recyclable. However, since aluminum products are not magnetic, they are difficult to separate by the magnetic separator and will remain in the sorted waste for further sorting.

[0048] Because aluminum cans are primarily made of aluminum, their sides are relatively light and soft, making them prone to deformation during recycling. In contrast, the bottom and top of the can are reinforced and therefore less prone to deformation compared to the can body. Consequently, landfill waste typically arrives in one of two states:

[0049] One type is the flattened shape after being crushed from top to bottom, and the other is the intact state with possible partial deformation. Aluminum is a relatively light material with a large volume, making it easy to pass through the waste air separator, and its smaller size makes it easier to pass through. Therefore, intact cans are found in the lighter portion of the waste after passing through the waste air separator, while the crushed cans are relatively heavier and are more likely to be in the same area as the other heavier waste after screening. This solution mainly focuses on analyzing the heavier group of waste after passing through the air separator, that is, specifically analyzing the flattened aluminum cans after being crushed from top to bottom.

[0050] In this embodiment of the invention, a high-definition camera is used to collect image data at different time points to obtain garbage images at different sampling times. It should be noted that the garbage images are specifically image information of garbage above the conveyor belt. Therefore, image preprocessing and other means can be performed during the acquisition process to stably identify the garbage images, and different garbage areas can be determined through image recognition methods.

[0051] It should be noted that a garbage region represents the image region corresponding to a garbage individual. In this embodiment of the invention, garbage individuals can be divided by a pre-set artificial intelligence network model. The specific method is a conventional technique in the art and will not be described in detail. Of course, in other embodiments, other division methods can also be used to determine garbage regions, and there are no restrictions on this.

[0052] For the heavier waste after being sorted by the waste air separator, the aluminum cans (all aluminum cans mentioned below are flattened cans) are mainly flattened cylinders. Because they are smaller in volume and have a lower center of gravity than unflattened aluminum cans, they are not easy to slide or roll on the conveyor belt. However, due to the reduced volume of the flattened aluminum cans, they are easily blocked by other waste. They exist in two states on the conveyor belt: parallel to the conveyor belt and perpendicular to the conveyor belt.

[0053] The region exhibits different image features under different conditions. For image characteristics relevant to identifying this region as belonging to an aluminum can, please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of an aluminum can in a parallel state and a state biased towards vertical, according to an embodiment of the present invention.

[0054] The shape of the aluminum can in the image varies due to being flattened, but the changes in the bottom and lid are relatively small, so they can be the main targets of analysis. On the conveyor belt, when it is parallel to the conveyor belt, the area is close to a circle, and when it changes in the vertical direction, it is similar to an ellipse. Therefore, matching feature analysis can be performed by fitting circles.

[0055] Furthermore, in some embodiments of the present invention, the characteristic value of a waste area belonging to an aluminum can is determined based on the matching features of each waste area and the fitted circle and the gray value. This includes: fitting an ellipse to the waste area and determining the distance between the two foci of the fitted ellipse as the circle matching distance; performing maximum and minimum value normalization on the inverse of the circle matching distance to obtain a circle matching index; calculating the mean gray value of all pixels after graying the waste area to obtain a gray-scale feature index; and calculating the product of the circle matching index and the gray-scale feature index, and performing maximum and minimum value normalization to obtain the characteristic value.

[0056] In this study, the closer the distance between the two foci of the ellipse fitting, the more circular the waste area appears. Therefore, by normalizing the inverse of the circular matching distance using a maximum-minimum method, a circular matching index is obtained. A larger circular matching index value indicates that the corresponding waste area more closely matches the characteristics of a parallel conveyor belt for aluminum cans. Conversely, a smaller circular matching index value requires further discussion.

[0057] In this embodiment of the invention, since the magnetic separator filters out the remaining magnetic metals, most of the remaining items on the conveyor belt, except for aluminum cans, lack metallic luster. Therefore, for each waste area, its gloss level can be calculated. The flattened aluminum can surface produces relatively uniform reflected light due to the reflective properties of its metallic material, while other types of waste (such as plastic or paper) often lack this uniformity. Therefore, in each waste area of ​​each image, the greater its reflectivity, the greater the likelihood that it is an aluminum can area. Overlaying the white features corresponding to aluminum cans allows for specific analysis of reflectivity using grayscale values, resulting in superior analytical performance. In this embodiment of the invention, the average grayscale value of all pixels in the waste area after grayscale conversion is calculated to obtain grayscale feature indicators.

[0058] Therefore, the characteristic value is obtained by combining the product of the circle matching index and the grayscale feature index. The larger the characteristic value, the more the corresponding garbage area matches the characteristics of the aluminum can. However, there are still aluminum cans that are biased towards a vertical state, which do not show a circular feature in the image. In addition, aluminum cans are prone to bumps on the conveyor belt, which can cause changes in their state. Therefore, further analysis is required. For the specific analysis process, please refer to the following embodiments.

[0059] S102: Based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, determine the abrupt change amplitude of the same waste area at each sampling time; determine the candidate abrupt change time nodes that cause waste changes due to the conveyor belt based on the value of the abrupt change amplitude.

[0060] In this embodiment of the invention, the same garbage area at different sampling times can be specifically analyzed by optical flow method. Alternatively, since the conveyor belt itself moves at a constant speed, the same garbage area can be matched and identified based on the constant speed movement to determine the same garbage area at different sampling times.

[0061] In the time series of each region, the state of the aluminum can is in multiple states, and its state is constantly changing due to the movement and bumps of the conveyor belt. During the movement and bumps of the conveyor belt, the aluminum can will constantly change between two states: parallel and perpendicular to the conveyor belt. Therefore, its characteristic value will show high and low fluctuations in time. However, since the bumps experienced by the conveyor belt are different at several different positions when it is moving, the fluctuation range of its characteristic value is different.

[0062] Understandably, related technologies analyze specific features using overall characteristics. However, the presence of rollers on the conveyor belt causes bumps at the roller locations, leading to changes in the state of the waste. On a normal conveyor belt, these changes manifest as small vibrations, but when passing through the roller area, the impact is greater. (See [link to relevant documentation]). Figure 3 , Figure 3 This is a schematic diagram of the conveyor belt roller area provided in one embodiment of the present invention. The impact of the state of the waste on the conveyor belt varies. The roller area can cause errors in the overall judgment; therefore, further analysis is needed in conjunction with the roller area.

[0063] Furthermore, in some embodiments of the present invention, the abrupt change amplitude of the same waste area at each sampling time is determined based on the characteristic value changes of the same waste area at different positions on the conveyor belt at different sampling times, including: determining any sampling time as the target time, calculating the difference between the characteristic value of the target time and the previous time as the first difference; calculating the difference between the characteristic value of the target time and the next time as the second difference; and using the normalized value of the sum of the first difference and the second difference as the abrupt change amplitude at the target time.

[0064] In other words, the difference between the characteristic values ​​of any sampling time and the two adjacent sampling times is analyzed for abrupt change. The larger the value of the abrupt change, the more obvious the state change at the corresponding target time.

[0065] Furthermore, in some embodiments of the present invention, determining the candidate mutation time nodes that cause changes in waste due to the conveyor belt based on the numerical value of the mutation amplitude includes: taking the time nodes with mutation amplitude greater than a preset amplitude threshold as candidate mutation time nodes.

[0066] In this embodiment of the invention, a preset amplitude threshold is used as a threshold value for the mutation amplitude. Since the larger the value of the mutation amplitude, the more obvious the state mutation occurs at the corresponding target time, threshold analysis is performed directly. Optionally, the preset amplitude threshold can be, for example, 0.7. That is to say, time nodes with mutation amplitude greater than 0.7 are used as candidate mutation time nodes.

[0067] The candidate mutation time point represents the time point when the garbage status changes significantly, but whether the specific cause of the change is due to the rolling mechanism needs further analysis.

[0068] Since every region in the image vibrates at every moment, the vibration caused by the conveyor belt movement is a low-frequency high-frequency vibration, which manifests as changes in each garbage area in the image (i.e., large jolts on the conveyor belt caused by the rollers on the conveyor belt; the amplitude of these jolts is large, but their frequency varies with the distribution of the rollers; the garbage changes significantly at that location, resulting in a reset-like effect in the distribution of garbage on the conveyor belt). Meanwhile, the low-frequency, high-amplitude fluctuations caused by the conveyor belt passing over the rollers (similar to the above, but this is a small-range vibration caused by the movement of the conveyor belt; its amplitude is small, but its frequency is larger than the jolts above) make the classification in each time period calculated based on the candidate mutation nodes unclear. The candidate mutation time nodes may contain mutations caused by non-wheel / axle issues during conveyor belt movement, requiring further screening.

[0069] S103: Based on the candidate mutation time nodes corresponding to each waste area, the target time nodes that cause mutations in the state of waste on the conveyor belt are selected. Using the target time nodes as the cutting points, the different time periods of each waste area on the conveyor belt are determined. Based on the changes in the overall mutation amplitude in each time period, the state change index of the waste area in each time period is determined.

[0070] Since the causes of significant changes on the conveyor belt vary, it is necessary to identify the time points with the largest degree of change and analyze whether the changes between the time points are due to the rollers on the conveyor belt or other relatively soft waste.

[0071] Furthermore, in some embodiments of the present invention, the target time node for the occurrence of a change in the state of waste on the conveyor belt is obtained by screening the candidate mutation time nodes corresponding to each waste area, including: determining the frequency of the sampling time corresponding to the candidate mutation time nodes of all waste areas, and taking the preset number of candidate mutation time nodes with the highest frequency as the target time nodes.

[0072] The preset number can be the same as the number of rollers. For example, if there are 10 rollers on the conveyor belt, the preset number can be set to 10. In this embodiment of the invention, since other locations may also experience significant abrupt changes due to the softness of the waste itself, leading to abrupt changes, but roller abrupt changes are relatively stable, the preset number of candidate abrupt change time nodes with the highest frequency are selected as target time nodes. This allows for direct filtering to obtain the target time nodes caused by roller abrupt changes.

[0073] After determining the target time point, and using that time point as a dividing point, we can determine the different time periods each waste area will experience on the conveyor belt. Within the same time period, other soft waste will undergo frequent changes in state, while hard materials like aluminum cans will show less change. This allows for a detailed analysis of the state changes in each time period.

[0074] Furthermore, in some embodiments of the present invention, the state change index of the waste area in each time period is determined based on the change in the overall mutation amplitude in each time period, including: determining the mutation stability in the time period based on the dispersion of all mutation amplitudes in the same time period; calculating the negative of the mean of all mutation amplitudes in the same time period, and normalizing the maximum and minimum values ​​as numerical feature indicators; and using the product of the numerical feature indicators and the mutation stability as the state change index of the waste area in the corresponding time period.

[0075] In the dispersion analysis, the standard deviation can be specifically used. That is, the negative of the standard deviation of all mutation amplitudes within the same time period is calculated, and the maximum and minimum values ​​are normalized as mutation stability. Mutation stability is a characteristic that represents the stability of the overall mutation amplitude within a time period. The larger the mutation stability value, the more stable it is.

[0076] The negative of the mean of all mutation amplitudes within the same time period, and the normalized maximum and minimum values, are used as numerical characteristic indicators. The larger the numerical characteristic indicator, the smaller the overall mutation amplitude and the smaller the mutation.

[0077] Therefore, the product of the numerical characteristic index and the mutation stability is directly used as the indicator of the state change of the garbage area within the corresponding time period. A larger value for the state defense index indicates more stable state changes in the garbage area within the corresponding time period, and a smaller overall value.

[0078] S104: Combining characteristic values ​​and state change indicators, determine the probability indicators of a waste area being a can area, and screen can areas based on the probability indicators of each waste area.

[0079] Furthermore, in some embodiments of the present invention, determining the probability that a waste area is an aluminum can area by combining characteristic values ​​and state change indicators includes: calculating the average of the characteristic values ​​at all sampling times within the same time period as a segment characteristic coefficient; calculating the product of the segment characteristic coefficient and the state change indicator within the same time period as an influence indicator for the time period; and using the average of the influence indicators for all time periods as an indicator of the probability that the corresponding waste area is an aluminum can area.

[0080] The analysis considers all time periods, combining the impact of each period. Each sampling moment within the same time period has corresponding characteristic values ​​and state change indicators. Therefore, the product of the characteristic coefficient and the state change indicator within the same time period is directly calculated as the impact indicator of the time period. The impact indicator of the time period represents the overall change characteristics and similarity of the waste area to that of aluminum cans within the time period.

[0081] This similarity only represents the characteristics of a single time period. It's necessary to consider all time periods. Therefore, the average of the influencing indicators across all time periods is used as the probability index for a corresponding waste area to be an aluminum can area. This probability index represents the likelihood that a waste area is an aluminum can area. Because it incorporates the overall conveyor belt vibration analysis, this probability index has higher accuracy and objectivity.

[0082] Furthermore, in some embodiments of the present invention, the selection of aluminum can areas based on the probability index of each waste area includes: selecting waste areas with a probability index greater than a preset probability threshold as aluminum can areas.

[0083] The preset probability threshold is a limit value for the probability index. In this embodiment of the invention, the preset probability threshold can be set to 0.7, that is, garbage areas with a probability index greater than 0.7 are considered as aluminum can areas. This achieves accurate identification of aluminum can areas.

[0084] This invention acquires images of waste on a conveyor belt at different sampling times, performs image analysis to determine different waste areas, and determines the characteristic value of each waste area belonging to an aluminum can based on the matching features of each waste area with the circle and the gray value. This characteristic value characterizes the aluminum can feature of the waste area itself, so subsequent numerical analysis and change analysis can be performed based on the characteristic value. Then, based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined. Based on the numerical value of the abrupt change amplitude, candidate abrupt change time nodes causing waste changes due to the conveyor belt are determined. Based on the candidate abrupt change time nodes corresponding to each waste area, target time nodes causing abrupt changes in the waste state on the conveyor belt are selected. Using the target time nodes as cutting points, different time periods for each waste area on the conveyor belt are determined. The division of time periods effectively avoids the "state reset" effect caused by the bumpy rollers, improving the accuracy and reliability of the overall numerical analysis. Based on the changes in the overall abrupt change amplitude in each time period, the state change index of the waste area within each time period is determined. Combining the characteristic values ​​and state change indexes, the probability index of the waste area being an aluminum can area is determined. Based on the probability index of each waste area, aluminum can areas are selected. In summary, this invention can combine shape features and grayscale features to perform characteristic value analysis, and further, based on characteristic abrupt changes and the influence of the rollers on the conveyor belt, determine more accurate and reliable probability indexes, resulting in better and more accurate aluminum can classification and identification.

[0085] The present invention also provides a solid waste classification and identification system for recycling, the system including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned solid waste classification and identification method for recycling.

[0086] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0087] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for classifying and identifying solid waste for recycling, characterized in that, The method includes: Images of waste on the conveyor belt at different sampling times are acquired, and image analysis is performed to determine different waste areas. Based on the matching features of each waste area with the circle and the gray value, the characteristic value of the waste area belonging to the aluminum can is determined. Based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined; based on the value of the abrupt change amplitude, the candidate abrupt change time nodes that cause changes in waste due to the conveyor belt are determined. Based on the candidate mutation time nodes corresponding to each waste area, the target time nodes that cause mutations in the state of waste on the conveyor belt are selected. Using the target time nodes as the cutting points, the different time periods of each waste area on the conveyor belt are determined. Based on the changes in the overall mutation amplitude in each time period, the state change index of the waste area in each time period is determined. By combining the characteristic values ​​and the state change indicators, a probability index is determined that the waste area is an aluminum can area, and aluminum can areas are screened based on the probability index of each waste area.

2. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, The step of determining the characteristic value of a waste area belonging to an aluminum can based on the matching features of each waste area and the fitted circle, as well as the gray value, includes: Ellipse fitting is performed on the waste area, and the distance between the two foci of the fitted ellipse is determined as the circle matching distance; The negative of the circle matching distance is normalized by taking the maximum and minimum values ​​to obtain the circle matching index; Calculate the average grayscale value of all pixels in the garbage area after grayscale conversion to obtain the grayscale feature index; Calculate the product of the circle matching index and the grayscale feature index, and normalize the maximum and minimum values ​​to obtain the characteristic value.

3. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the changes in characteristic values ​​of the same waste area at different positions on the conveyor belt at different sampling times, the abrupt change amplitude of the same waste area at each sampling time is determined, including: Determine any sampling time as the target time, calculate the difference between the characteristic values ​​of the target time and its previous time as the first difference, and calculate the difference between the characteristic values ​​of the target time and its subsequent time as the second difference; The normalized value of the sum of the first difference and the second difference is taken as the mutation amplitude at the target time.

4. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the numerical value of the mutation amplitude, candidate mutation time points for the conveyor belt-induced waste changes are determined, including: The time nodes where the mutation magnitude is greater than a preset magnitude threshold are selected as candidate mutation time nodes.

5. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the candidate mutation time nodes corresponding to each waste area, the target time nodes where waste state mutations occur on the conveyor belt are selected, including: Determine the frequency of the sampling time corresponding to the candidate mutation time nodes in all waste areas, and select the preset number of candidate mutation time nodes with the highest frequency as the target time nodes.

6. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the changes in the overall magnitude of abrupt changes within each time period, the indicators of the state changes of the waste area within each time period are determined, including: The stability of mutations within a time period is determined by the dispersion of all mutation amplitudes within the same time period. Calculate the negative of the mean of all mutation amplitudes within the same time period, and normalize the maximum and minimum values ​​to use them as numerical feature indicators. The product of numerical characteristic indicators and mutation stability is used as an indicator of the state change of the waste area within the corresponding time period.

7. A method for classifying and identifying solid waste for recycling as described in claim 6, characterized in that, Based on the dispersion of all mutation amplitudes within the same time period, the mutation stability within that time period is determined, including: Calculate the negative of the standard deviation of all mutation amplitudes within the same time period, and normalize the maximum and minimum values ​​to obtain mutation stability.

8. The method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the aforementioned characteristic values ​​and the aforementioned state change indicators, the probability of the waste area being a can area is determined, including: Calculate the mean of the characteristic values ​​at all sampling times within the same time period, and use it as the segment characteristic coefficient; Calculate the product of the segment characteristic coefficient and the state change index within the same time period, and use it as the influence index for the time period; The average of all influencing indicators over all time periods is used as an indicator of the probability that the corresponding waste area is an area containing aluminum cans.

9. A method for classifying and identifying solid waste for recycling as described in claim 1, characterized in that, Based on the probability indicators for each waste area, can areas are screened, including: Waste areas with a probability index greater than a preset probability threshold are designated as aluminum can areas.

10. A solid waste sorting and identification system for recycling, the system 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 the method as described in any one of claims 1 to 9.

Images