Intelligent recycling-based visual data processing method and system

Abstract

The application discloses a visual data processing method and device based on intelligent recycling. The device comprises a control module arranged on a garbage recycling bin cover body, a camera in communication with the control module and used for acquiring newly input recycling objects in the garbage recycling bin, and a recycling object delivery opening arranged on the garbage recycling bin cover body and controlled to be opened, wherein the recycling object delivery opening is connected with the control module. The method comprises the following steps: an image acquisition step, a recycling object detection step, and a new recycling object judgment step. The application provides a visual data processing method and system based on intelligent recycling with low-cost recycling devices and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent waste recycling, and more specifically, to a visual data processing method and system based on intelligent recycling. Background Technology

[0002] Currently, waste recycling systems typically involve placing recycling bins at designated locations in residential areas and office buildings. Residents can dispose of their waste in these bins, which are then collected periodically by waste collection personnel. To facilitate use, these bins are usually placed in multiple locations. However, each bin is designed for different waste categories, leading to frequent instances where residents mistakenly choose the wrong bin.

[0003] Current waste sorting methods largely rely on manual labor, thus failing to achieve true automated waste sorting and recycling. While some developed countries have introduced high-tech smart trash cans, such as solar-powered automatic compression bins and electrically powered compression bins, for plastic bottles, glass bottles, and aluminum cans, these systems require significant energy resources and are not particularly effective at identifying bottle types.

[0004] How to effectively achieve low-cost and efficient bottle recycling and improve bottle classification and identification results has become a problem to be solved. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides a low-cost recycling device and a highly accurate visual data processing method and system based on intelligent recycling.

[0006] The technical solution of the present invention is as follows:

[0007] A visual data processing method based on intelligent recycling is embedded in a visual data processing device based on intelligent recycling. The device includes: a control module mounted on the lid of a waste recycling bin; a camera connected to the control module and used to acquire newly deposited recyclables inside the waste recycling bin; and a controlled-opening recyclable disposal port mounted on the lid of the waste recycling bin, wherein the recyclable disposal port is connected to the control module. The method includes the following steps:

[0008] Image acquisition steps: Acquire an image of the current recyclable load in the trash can using a camera, and use it as the background image P. A ; and obtain an image of the current recyclable load in the recycling bin, as the original image P. B ;

[0009] The steps for detecting recyclable materials are as follows: The background image P is initially detected using the first target detection algorithm. A Original image P BAll detected recyclables;

[0010] Newly arrived recyclable material identification step: The data obtained from the recyclable material detection step is processed by a second-level identification and detection algorithm to locate the background image P. A Each detected object relative to the original image P B The position in the image is used to filter out the original image P. B With background image P A The locations of the detected substances differed;

[0011] The data acquired during the reclaim detection step is processed using a third frame difference algorithm to obtain the locations of reclaimed items with discrepancies. Combining the second-level identification and detection algorithm with the third frame difference algorithm, the location of newly arrived reclaimed items in the original image P is obtained. B The position in the middle;

[0012] Classification Steps: The newly added recyclables confirmed in the new recyclables judgment step are processed by the fourth classification detection algorithm to obtain the classification results of the target detection items.

[0013] Furthermore, after the sensor detects the reclaimed material, the control command activates the camera after a certain time interval to acquire the original image P. B .

[0014] Furthermore, the original image P in the new recyclable material identification step is selected. B With background image P A The locations of the three detected substances showed the greatest differences.

[0015] Furthermore, the second identification and detection algorithm employs Re-Identification technology or a softmax superimposed triplet hard joint training mechanism algorithm.

[0016] Furthermore, the fourth classification detection algorithm adopts the MobileNet algorithm, which includes the SE attention module and the Sa module.

[0017] Furthermore, the third frame difference algorithm processing is applied to the background image P. A Original image P B Frame difference processing is performed to obtain the difference detection objects that are greater than the threshold.

[0018] Furthermore, the first object detection algorithm uses the YOLOv5 algorithm and adds an attention mechanism to each block in its backbone network structure to optimize its resistance to light interference.

[0019] Furthermore, it also includes updating the current background image P after the classification step. A For the original image P B。

[0020] The visual data processing system based on intelligent recycling includes an implementation of any of the above-mentioned methods embedded in a visual data processing method based on intelligent recycling; it also includes a cloud server for processing data collected by a camera, a communication module connecting the control module and the cloud server, and a sensor that triggers the control module to issue a control command that detects the input of recyclable material.

[0021] Furthermore, it also includes terminal devices, which obtain data and instructions from the cloud server through communication modules.

[0022] The advantages of this invention are:

[0023] This invention modifies the lid of existing waste recycling bins by installing a camera on the side of the lid closest to the bin, enabling direct and low-cost detection of the waste loading status inside the bin. The highly accurate image recognition method provides crucial data for subsequent waste recycling bin scheduling.

[0024] This invention addresses the practical problem of recyclable material detection. By optimizing relevant algorithms, it achieves more accurate identification and classification of newly disposed recyclables. Specifically, after a user deposits waste, the system quickly identifies the type and location of the item. This technology eliminates the need for additional physical structures to buffer the falling object, requiring only two images of the waste bin before and after disposal. This increases the robustness of the identification process and effectively solves the problem of inaccurate identification caused by significant changes in the posture of newly placed objects. Furthermore, the smart waste bin lid features a specially designed dispensing opening for bottles, particularly plastic, glass, aluminum cans, and paper bottles. This standardizes the types of recyclables and establishes a paid recycling mechanism. The control module can utilize a low-end chip, reducing lid modification costs. Cloud services handle the identification process, minimizing waste collection and management costs as well as equipment hardware costs. Attached Figure Description

[0025] Figure 1a This is a flowchart of the visual data processing method based on intelligent recycling of the present invention;

[0026] Figure 1b The present invention relates to a visual data processing method based on intelligent recycling;

[0027] Figure 2 This is an image processing flowchart of the visual data processing method based on intelligent recycling of the present invention;

[0028] Figure 3 This is a schematic diagram of the visual data processing system based on intelligent recycling according to the present invention.

[0029] Figure 4 This is a schematic diagram of the recyclable material delivery port of the present invention. Detailed Implementation

[0030] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0031] As shown in Figure 1 to Figure 4 As shown, a visual data processing method based on intelligent recycling is embedded in a visual data processing device based on intelligent recycling. This device includes: a control module mounted on the lid of a waste recycling bin; a camera connected to the control module and used to acquire newly deposited recyclables inside the waste recycling bin; and a controlled-opening recyclable disposal port mounted on the lid of the waste recycling bin, wherein the recyclable disposal port is connected to the control module. The method specifically includes the following steps:

[0032] S11: Image Acquisition Step: Acquire an image of the current recyclable load in the trash can using a camera, which will serve as the background image P. A ; and obtain an image of the current recyclable load in the recycling bin, as the original image P. B That is, after the sensor detects the recyclable material, the control command activates the camera at certain intervals to acquire the raw image P. B It has achieved its function of monitoring the current situation inside the waste recycling bins, effectively detecting whether new recyclables have entered, so as to initiate subsequent recyclable detection.

[0033] S12: Recyclable material detection step: Initial detection of background image P using the first target detection algorithm. A Original image P B All detected recyclables.

[0034] Preferably, the first object detection algorithm uses the YOLOv5 algorithm, which is optimized by adding an attention mechanism to each block in its backbone network structure. This makes the object detection box localization more accurate and enhances its resistance to light interference, making it better suited to the complex environment inside current garbage bins. Alternatively, the SSD or FASTRCNN algorithms can also be used.

[0035] S13: New Recycling Item Judgment Step: The data obtained from the recycling item detection step is processed by a second-level identification and detection algorithm to locate the background image P. A Each detected object relative to the original image P B The position in the image is used to filter out the original image P. B With background image P A The locations of the detected substances differ.

[0036] Among these steps, the location of the detected object is selected, and the original image P is used to determine the location of the newly recovered object. B With background image P A The locations of the top three samples with the greatest differences are sufficient to confirm the location of newly recovered samples.

[0037] Preferably, the second recognition and detection algorithm can use Re-Identification technology, or ReID technology for short, or further adopt a softmax plus triplet hard joint training mechanism to adapt to the needs of accurate re-recognition in current application scenarios.

[0038] refer to Figure 2 Simultaneously, the data acquired in the reclaimed material detection step is processed using the third frame difference algorithm to obtain the locations of reclaimed materials with discrepancies. Combining the second-level recognition and detection algorithm with the third frame difference algorithm, the location of newly arrived reclaimed materials in the original image P is obtained. B The position within the image. The third frame difference algorithm processes the background image P. A Original image P B By performing frame difference processing, we can obtain the difference detection objects that are greater than the threshold, thereby effectively confirming the location of newly added recyclables.

[0039] S14: Classification Step: The newly added recyclables confirmed in the new recyclables judgment step are processed by the fourth classification detection algorithm to obtain the classification result of the target detection object. The fourth classification detection algorithm uses the MobileNet algorithm, with added SE attention and Sa modules to improve classification accuracy. Alternatively, ShuffleNet or EfficientNet techniques can be used more effectively.

[0040] Preferably, an update can be performed after the classification step to update the current background image P. A For the original image P B In order to process the next new input, repeat steps S11-S14 above.

[0041] refer to Figure 3 In another aspect, the present invention discloses a visual data processing system based on intelligent recycling, including the aforementioned implementation of a visual data processing method embedded in an intelligent recycling method; it also includes a cloud server for processing data collected by a camera, a communication module connecting the control module and the cloud server, and a sensor that triggers the control module to issue a control command for detecting the input of recyclable material.

[0042] Preferably, it may also include a terminal device, which obtains data and instructions from the cloud server through a communication module.

[0043] Specifically, the recycling outlet includes an outlet body 1 and an outlet cover 2. The outlet cover 2 is connected to the outlet body 1 via a normally open hinge 3, keeping the outlet in a normally open state. Traditional trash cans use a similar hinge 3, meaning the trash can is either always open or closed for extended periods. Users frequently need to manually open the lid, significantly increasing the burden on them. After prolonged use, the trash can becomes dirty, and many users are unwilling to manually open the lid, instead placing the trash next to the bin. This is detrimental to waste recycling and environmental pollution. Therefore, using a normally open hinge 3 to keep the outlet properly open will encourage users to actively participate in waste recycling.

[0044] Specifically, the normally open hinge 3 uses a spring to keep the feed port body 1 and the feed port cover 2 in a normally open state. Using a spring reduces costs, facilitates maintenance and replacement, and promotes wider adoption.

[0045] A detection device can also be installed at the recycling inlet of the delivery port body 1. This device detects the number of delivered items and is unaffected by weather or temperature. This allows for more efficient detection of the quantity of delivered items. The detection device uses an infrared detector, which is unaffected by weather or temperature, achieves accurate detection, and is relatively low in cost.

[0046] Around the recycling inlet of the disposal port 1, AIoT-based cameras can be installed to acquire information about the recyclables. These cameras can assist the detection device in counting the number of recyclable items. Alternatively, they can acquire information about the items; when the detection device detects an item, the camera activates to photograph it, and the content of the item can be identified from the image. A mechanism that triggers the camera's operation based on the detection device can be implemented for more efficient operation and energy conservation.

[0047] The communication module typically employs AIoT technology to better ensure system operation. To standardize the separate disposal of such items by residents, a paid recycling mechanism can be established to encourage resident use. This solution automatically opens the lid after the user scans the code, and the sensor is triggered in real-time for identification after the user deposits the item. The camera requirements are not stringent, and a high-definition industrial camera is not necessary. Image compression reduces the image size to 1 / 50 to 1 / 30 of a typical high-definition image, effectively improving the user experience.

[0048] In summary, this design is easy to implement and portable. It can be easily upgraded to various sizes of standard trash cans with a single click, has a small footprint, is very easy to promote, and is convenient and cost-effective for use in public areas. The recyclable material disposal opening is rationally designed; users simply drop the recyclables from top to bottom and earn points instantly upon closing the lid. The data processing employs cascaded model fusion technology to increase the robustness of recognition, effectively solving problems such as inaccurate recognition caused by significant changes in the posture of objects after being hit.

[0049] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the concept of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A visual data processing method based on intelligent recycling, characterized in that, An embedded visual data processing device based on intelligent recycling is provided. The device includes: a control module mounted on the lid of a waste recycling bin; a camera connected to the control module and used to acquire newly deposited recyclables inside the waste recycling bin; and a controllably open recyclable disposal port mounted on the lid of the waste recycling bin, the disposal port being connected to the control module. The method includes the following steps: Image acquisition steps: Acquire an image of the recyclable material load inside the recycling bin using a camera, which will serve as the background image P. A After the sensor detects the recyclable material, the control command activates the camera at regular intervals to acquire the raw image P. B ; The steps for detecting recyclable materials are as follows: The background image P is initially detected using the first target detection algorithm. A Original image P B All detected recyclables; Newly arrived recyclable material identification step: The data obtained from the recyclable material detection step is processed by a second-level identification and detection algorithm to locate the background image P. A Each detected object relative to the original image P B The position in the image is used to filter out the original image P. B With background image P A The locations of the detected substances differed; The data acquired during the reclaim detection step is processed using a third frame difference algorithm to obtain the locations of reclaimed items with discrepancies. Combining the second-level identification and detection algorithm with the third frame difference algorithm, the location of newly arrived reclaimed items in the original image P is obtained. B The position in the middle; Classification Steps: The newly added recyclables confirmed in the new recyclables judgment step are processed by the fourth classification detection algorithm to obtain the classification results of the target detection items.

2. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: After the sensor detects the recyclable material, the control command activates the camera after a certain time interval to acquire the original image P. B .

3. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: The second identification and detection algorithm uses Re-Identification technology or a softmax plus triplet hard joint training mechanism algorithm.

4. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: The fourth classification detection algorithm adopts the MobileNet algorithm, which includes the SE attention module and the Sa module.

5. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: The third frame difference algorithm processing is applied to the background image P. A Original image P B Frame difference processing is performed to obtain the difference detection objects that are greater than the threshold.

6. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: The first object detection algorithm uses the YOLOv5 algorithm and adds an attention mechanism to each block in its backbone network structure to optimize its resistance to light interference.

7. The visual data processing method based on intelligent recycling according to claim 1, characterized in that: This also includes updating the current background image P after the classification step. A For the original image P B。 8. A visual data processing system based on intelligent recycling, characterized in that: The method includes an implementation of any one of claims 1 to 7, which is embedded in a visual data processing method based on an intelligent recycling method; it also includes a cloud server for processing data collected by a camera, a communication module that connects the control module to the cloud server, and a sensor that triggers the control module to issue a control command that detects the input of recyclable material.

9. The visual data processing system based on intelligent recycling according to claim 8, characterized in that: It also includes terminal devices, which obtain data and instructions from the cloud server through the communication module.

Images