A remote sensing method, apparatus, and treatment equipment for urban solid waste extraction based on improved YOLO V11.

Abstract

This application provides a remote sensing method, apparatus, and processing equipment for urban solid waste extraction based on an improved YOLO V11 model. By making corresponding adaptive optimizations based on the YOLO V11 model, it can effectively solve the contradiction between efficiency and accuracy in solid waste identification of traditional methods, significantly improve the processing capability for large-area and large-scale remote sensing images, meet the high-efficiency and accurate high-quality application requirements in urban solid waste remote sensing monitoring, and thus provide good data support for urban solid waste treatment.

Description

Technical Field

[0001] This application relates to the field of remote sensing technology, specifically to a remote sensing method, apparatus, and processing equipment for urban solid waste extraction based on an improved YOLO V11. Background Technology

[0002] Urban solid waste refers to solid or semi-solid waste materials generated by humans in production, construction, daily life and other activities. It mainly includes three categories: construction waste, domestic waste and industrial waste. It will affect the urban ecological environment and the comfort of residents' lives. Therefore, the dynamic monitoring and on-site investigation mechanism of urban solid waste is particularly important.

[0003] In this context, efficiently and accurately identifying the precise location of urban solid waste is a crucial prerequisite for conducting dynamic monitoring and on-site investigations. Urban solid waste is characterized by its wide distribution, large quantity, and diverse types. Traditional manual on-site survey methods are time-consuming, labor-intensive, lack data objectivity, have limited coverage, and are inefficient, making them unsuitable for meeting the actual needs of urban environmental governance. However, the rapid development of satellite remote sensing technology has provided a new solution for the dynamic monitoring of regional urban solid waste.

[0004] Remote sensing technology has been widely used in fields such as ecological environment monitoring, natural resource detection, urban environmental governance, and natural disaster monitoring due to its advantages such as wide coverage, short revisit period, rich information on ground features, and objective data acquisition. In particular, with the advancement of high-resolution remote sensing image technology, it has now become possible to achieve fine identification and extraction of urban solid waste points on a spatial scale.

[0005] Currently, urban solid waste identification methods based on high-resolution remote sensing imagery can be mainly divided into the following three categories: 1) Manual visual interpretation is a recognition method that combines human-computer interaction. It mainly relies on expert experience, while also referencing high-resolution remote sensing imagery and field survey results to identify and extract solid waste areas. However, although this method can achieve high-precision solid waste area identification, it depends on manual operation, has low efficiency, and is difficult to meet the needs of large-scale monitoring. 2) Semi-automatic interpretation techniques based on classification or segmentation quickly identify urban solid waste patches by designing and training targeted classifiers. This method utilizes the spectral, texture, and geometric features of urban solid waste for target extraction and has a high degree of automation. However, because the algorithms are developed for remote sensing images of specific scenes and time phases, their generalization ability is limited due to algorithm design and applicable scope. In practical applications, this often leads to low accuracy and inaccuracy in solid waste patch extraction. 3) Deep learning-based automatic interpretation technology constructs a recognition model by autonomously learning the detailed features of ground objects in high-resolution remote sensing images. Compared with the previous two methods, this method improves both automation and recognition accuracy. However, it still suffers from problems such as a large requirement for training samples, the tendency to misidentify or miss images when dealing with complex and diverse remote sensing images, and the need to further improve extraction accuracy.

[0006] In summary, existing methods for identifying and extracting urban solid waste each have their own advantages and disadvantages, but they all have certain limitations. In practical applications, further optimization and improvement are still needed to meet the dual requirements of accuracy and efficiency for dynamic monitoring of urban solid waste. Summary of the Invention

[0007] This application provides a remote sensing method, apparatus, and processing equipment for urban solid waste extraction based on an improved YOLO V11 model. By making corresponding adaptive optimizations based on the YOLO V11 model, it can effectively solve the contradiction between efficiency and accuracy in solid waste identification of traditional methods, significantly improve the processing capability for large-area and large-scale remote sensing images, meet the high-efficiency and accurate high-quality application requirements in urban solid waste remote sensing monitoring, and thus provide good data support for urban solid waste treatment.

[0008] Firstly, this application provides a remote sensing method for urban solid waste extraction based on an improved YOLO V11, the method comprising: For the target urban area, acquire 4-band optical satellite remote sensing image data that meet the sample quality requirements; After preprocessing the 4-band optical satellite remote sensing image data, it was further processed into urban solid waste sample data; A remote sensing urban solid waste extraction model was trained based on urban solid waste sample data. The model was trained on an improved YOLO V11 model, in which the original YOLO V11 model was replaced with an LSKNet module as the backbone network. The improved YOLO V11 model consists of an LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making predictions based on the fused features output by the feature fusion layer.

[0009] Secondly, this application provides a remote sensing urban solid waste extraction device based on an improved YOLO V11, the device comprising: The acquisition unit is used to acquire 4-band optical satellite remote sensing image data that meets the sample quality requirements for the target urban area. The production unit is used to produce urban solid waste sample data after preprocessing the 4-band optical satellite remote sensing image data; The training unit is used to train a remote sensing urban solid waste extraction model based on urban solid waste sample data. The remote sensing urban solid waste extraction model is trained on the basis of an improved YOLO V11 model. The improved YOLO V11 model replaces the original YOLO V11 model as the backbone network with an LSKNet module. The improved YOLO V11 model includes an LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making prediction outputs based on the fused features output by the feature fusion layer.

[0010] Thirdly, this application provides a processing device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor executes the method provided in the first aspect of this application when it invokes the computer program in the memory.

[0011] Fourthly, this application provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the method provided in the first aspect of this application.

[0012] From the above, it can be concluded that this application has the following beneficial effects: For the dynamic monitoring of urban solid waste, this application uses the YOLO V11 model as a basis and makes corresponding adaptive optimizations to effectively solve the contradiction between efficiency and accuracy in solid waste identification of traditional methods. It significantly improves the ability to process large-area and large-scale remote sensing images, which can meet the high-efficiency and accurate high-quality application requirements in urban solid waste remote sensing monitoring, and thus provide good data support for urban solid waste treatment. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of a remote sensing method for urban solid waste extraction based on the improved YOLO V11 in this application; Figure 2 This is a schematic diagram illustrating an example of the data conversion operation in this application. Figure 3 A logical diagram illustrating the configuration logic for the model in this application; Figure 4 This is a schematic diagram of a remote sensing urban solid waste extraction device based on the improved YOLO V11 in this application. Figure 5 This is a schematic diagram of one type of processing equipment used in this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. The naming or numbering of steps appearing in this application does not imply that the steps in the method flow must be performed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical purpose, as long as the same or similar technical effect is achieved.

[0017] The module division described in this application is a logical division. In practical applications, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual coupling, direct coupling, or communication connections may be through interfaces, and the indirect coupling or communication connections between modules may be electrical or other similar forms, none of which are limited in this application. Moreover, the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed across multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in this application.

[0018] Before introducing the remote sensing method for urban solid waste extraction based on the improved YOLO V11 provided in this application, we will first introduce the background content involved in this application.

[0019] The remote sensing method, apparatus, and computer-readable storage medium for urban solid waste extraction based on the improved YOLO V11 provided in this application can be applied to processing equipment for use in [specific applications].

[0020] The remote sensing urban solid waste extraction method based on the improved YOLO V11 mentioned in this application can be implemented by a remote sensing urban solid waste extraction device based on the improved YOLO V11, or by different types of processing devices such as servers, physical hosts, or user equipment (UE) that integrate the remote sensing urban solid waste extraction device based on the improved YOLO V11. The remote sensing urban solid waste extraction device based on the improved YOLO V11 can be implemented in hardware or software. The UE can be a terminal device such as a smartphone, tablet, laptop, desktop computer, or personal digital assistant (PDA). The processing devices can be configured in a device cluster.

[0021] It is understandable that the solution proposed in this application is usually based on existing data or data that has already been collected. Therefore, the processing equipment that implements the remote sensing urban solid waste extraction method based on the improved YOLO V11 of this application or that is equipped with the corresponding application service of the remote sensing urban solid waste extraction method based on the improved YOLO V11 of this application usually only needs to meet the required data processing capabilities, and its specific equipment type and equipment deployment form are quite flexible.

[0022] If the direct acquisition of existing data mentioned above is also involved, then further hardware and software adaptations are needed for the processing equipment to enable it to acquire data. For example, if real-time acquisition of remote sensing image data is required, the relevant remote sensing image data acquisition system / platform can be incorporated into the processing equipment cluster. Alternatively, the processing equipment itself can be the control unit of the relevant remote sensing image data acquisition system / platform. Or, a third-party call can be used to trigger the relevant remote sensing image data acquisition system / platform outside the processing equipment to perform real-time remote sensing image data acquisition operations.

[0023] In addition, if there is a need to display the processing progress (including the processing results), the processing device itself can be configured with the required display screen (including touch screen) to display the specific content. Of course, the processing device can also display the specific content through an external display device or other devices with a display screen.

[0024] The following section introduces the remote sensing method for urban solid waste extraction based on the improved YOLO V11 provided in this application.

[0025] First, refer to Figure 1 , Figure 1 This paper presents a flowchart illustrating a remote sensing method for urban solid waste extraction based on the improved YOLO V11. The remote sensing method for urban solid waste extraction based on the improved YOLO V11 provided in this application may specifically include the following steps S101 to S103: Step S101: For the target urban area, acquire 4-band optical satellite remote sensing image data that meets the sample quality requirements; Understandably, under the target of dynamic monitoring of urban solid waste, the proposed solution starts from the image recognition level and involves the application of deep learning methods, specifically involving the optimization design based on the YOLO V11 model. In this case, in order to train the remote sensing urban solid waste extraction model, it is necessary to configure the corresponding training samples, which corresponds to the acquisition and processing of 4-band optical satellite remote sensing image data.

[0026] The four-band optical satellite remote sensing image data mentioned above are typically readily available data, such as remote sensing image data products provided by one or more remote sensing platforms. Of course, in practical applications, the possibility of acquiring remote sensing image data in real time as the proposed solution is not excluded.

[0027] Furthermore, it can be noted that the remote sensing image data involved in this application is specifically 4-band optical satellite remote sensing image data, which is an image composed of four spectral bands in a specific wavelength range. These four bands are usually blue, green, red and near-infrared bands.

[0028] It is also worth noting that the 4-band optical satellite remote sensing image data used in this application is not only the processing range of the target urban area to be addressed by the current scheme, but also high-quality data that has passed the sample quality requirements.

[0029] As an example, for a specific urban area, this application can select 4-band optical satellite remote sensing image data with a spatial resolution better than 0.75 meters, an overall cloud cover of less than 5%, and no clouds in key areas.

[0030] Step S102: After preprocessing the 4-band optical satellite remote sensing image data, continue to produce urban solid waste sample data; Understandably, after initially obtaining 4-band optical satellite remote sensing image data, in order to obtain training samples, namely urban solid waste sample data, in addition to annotation operations, preprocessing may also be involved to improve the data quality, so as to facilitate the subsequent model training work to proceed faster and better.

[0031] In terms of details, as an exemplary embodiment, the preprocessing involved here may include two main aspects: sample amplification processing and data augmentation processing. Specifically, these include: 1) Sample amplification processing includes random cropping, random reversal, random rotation, and random illumination adjustment; Understandably, sample augmentation is a common operation in model configuration. By using a series of sample expansion methods, the sample size can be effectively increased, which can effectively alleviate the problem of insufficient data and help improve the robustness of the model to complex scenarios.

[0032] 2) Data augmentation processing includes radiometric calibration, atmospheric correction, geometric registration, image fusion, and image mosaicking. Image fusion involves the fusion of multispectral and panchromatic images, while image mosaicking involves stitching several images after image fusion through geometric mosaicking, color equalization, feathering, and de-overlapping.

[0033] Understandably, for remote sensing images, radiometric calibration, atmospheric correction, geometric registration, image fusion, and image mosaicking are common operations that can effectively improve image quality. This application specifically builds these operations into a processing architecture suitable for the specific application scenario of this application.

[0034] Specifically, there are: 1) Radiation calibration Radiometric calibration refers to the process of establishing a quantitative relationship between the digital output value (DN) of a remote sensing sensor and the radiance value in its corresponding field of view. Its main function is to eliminate errors generated by the sensor itself, that is, to convert the DN value recorded by the satellite sensor into the absolute radiance value (radiance) in the corresponding field of view.

[0035] 2) Atmospheric correction During transmission, electromagnetic waves are affected by atmospheric and lighting factors, resulting in losses and deviations. To eliminate these errors, the process of restoring the image radiance to the true reflectance of the ground objects is called atmospheric correction.

[0036] 3) Geometric registration The purpose of geometric registration is to accurately match images acquired at different times to a unified geographic coordinate system, eliminating spatial position deviations caused by changes in sensor position, terrain undulations, or differences in image acquisition conditions, so as to ensure that each pixel can accurately correspond to geographic coordinates on the ground.

[0037] 4) Image fusion Image fusion is an important technique in the processing of temporal high-resolution image data. Its purpose is to organically combine multi-source, multi-resolution image data, make full use of the advantages and characteristics of different images, and comprehensively utilize the rich spatial and spectral information of high-resolution remote sensing image data to fuse multispectral and panchromatic images. This results in remote sensing images that contain both rich spectral information and high spatial resolution geometric texture information, thereby improving the spatial detail and spectral information quality of the final data.

[0038] 5) Image mosaic Image mosaicking is a key step in the processing of time-series high-resolution image data. Its purpose is to stitch together multiple images covering different areas into a complete and seamless dataset through processing such as geometric mosaicking, color balancing, feathering, and de-overlapping, so as to obtain complete images of the study area and meet the needs of continuous monitoring of large areas.

[0039] This application also provides a practical solution for the annotation process in configuring urban solid waste sample data.

[0040] Specifically, as an exemplary embodiment, the process of creating urban solid waste sample data may include the following processing steps: 1) Draw the corresponding urban solid waste area vector data for the urban solid waste sample data on the designated GIS platform; GIS stands for Geographic Information System or Geo-Information System.

[0041] Understandably, the GIS platform allows experts or related machines to easily identify and label urban solid waste areas through visual interpretation, thereby constructing corresponding vector data and laying a good foundation for batch cropping of solid waste area images.

[0042] 2) Based on the urban solid waste regional vector data, the urban solid waste sample data is cropped to obtain the urban solid waste image dataset; Understandably, the purpose of cropping is to omit irrelevant urban areas and crop out the image data of urban solid waste areas. This will form an urban solid waste image dataset. The urban solid waste areas in the urban solid waste image dataset correspond to the urban solid waste areas that the subsequent model should theoretically identify. Of course, it should be noted that this is not an equivalent relationship. This is a basic cropping operation. Further data processing is required to obtain urban solid waste extraction results that better match the model's expected output.

[0043] For example, the cropping result here is represented by a square image area (represented by a rectangular detection box composed of horizontal and vertical line segments). However, in reality, urban solid waste is often presented randomly and messily in images. Even regular-shaped urban solid waste is represented by slanted rectangles, which does not match the square image area in the cropping operation here.

[0044] If the GIS platform is equipped with a batch cropping function, the batch cropping function of the GIS platform can be used directly for rapid dataset cropping.

[0045] 3) Create urban solid waste sample data from the urban solid waste image dataset.

[0046] It is understandable that this involves specific sample annotation operations. Based on the urban solid waste image dataset obtained through the previous cropping operation, the corresponding rectangular detection boxes that are more closely related to urban solid waste can be annotated by combining the urban solid waste extraction result format that the model is expected to output.

[0047] In terms of details, as an exemplary embodiment, the urban solid waste image dataset is created as urban solid waste sample data, which may specifically include: 3.1) Label the urban solid waste image dataset and perform data augmentation on the labeled data. The data augmentation process includes preprocessing the labeled data to remove noise and redundant information, and optimizing and adjusting the non-standard label boundaries by combining a set threshold range and image augmentation methods. It is understandable that the specific annotation operation can be carried out manually or using automated annotation tools such as roLabelImg. Before using such automated annotation tools, the corresponding automated annotation logic needs to be configured.

[0048] Regarding the annotation logic, taking an automated annotation tool as an example, as a demonstrative implementation, the annotation process can specifically employ an automated annotation tool, and the corresponding annotation strategy can include the following: Based on the spectral reflectance, texture and morphological features of urban solid waste target objects, and using a bounding box containing urban solid waste target objects that reflects the actual boundaries of the objects as the output, automated annotation is performed.

[0049] Among them, the Oriented Bounding Box (OBB), or oriented bounding box, corresponds to the non-square rectangular detection frame mentioned earlier, which is more in line with the corresponding rectangular detection frame for urban solid waste.

[0050] The three features involved here—spectral reflectance, texture, and morphology—are easy to understand. They are specifically the image features that this application focuses on for the task of identifying urban solid waste, and they can also be reflected in the previous work of drawing vector data of urban solid waste areas.

[0051] During the annotation process of this rotating bounding box, the rotation position of the detection frame can be adjusted to further ensure that the actual boundary of the target object is accurately reflected and the saving operation is completed.

[0052] Meanwhile, data augmentation operations applied to the labeled data can further enhance the labeling quality of the samples. For example, by combining a set threshold range with image enhancement methods to optimize and adjust the irregular labeling boundaries, it can effectively ensure that the labels are highly consistent with the actual urban solid waste areas.

[0053] 3.2) Convert the XML format annotation data into a TXT format tag file. The tag file is represented by normalized coordinates. Each line records the information of a target object. The information of the target object includes the category number and the coordinates of the four vertices. The parameters are separated by spaces, and different targets are distinguished by newline characters.

[0054] Understandably, in one application example of this application, the previously generated annotation data is in XML format, while the file format for subsequent improvements to the YOLO V11 model is in TXT format. In this case, this application may further involve file format conversion processing.

[0055] As can be seen, this application has designed a specific conversion strategy for this situation, which is reflected in the form of a txt format tag file, that is, represented by normalized coordinates, such as (x0,y0,x1,y1,x2,y2,x3,y3). Each line records the information of a target object, including the category number and the coordinates of the four vertices. The parameters are separated by spaces, and different targets are distinguished by newline characters. This process ensures the standardization and consistency of the training samples, laying the foundation for subsequent good model training.

[0056] As an example, you can also refer to Figure 2 The following diagram illustrates an example of the data conversion operation of this application to provide a more intuitive understanding. The diagram on the left shows the original XML format annotations, and the diagram on the right shows the TXT format tags.

[0057] Step S103: Train a remote sensing urban solid waste extraction model based on urban solid waste sample data. The remote sensing urban solid waste extraction model is trained on the basis of an improved YOLO V11 model. The improved YOLO V11 model replaces the original YOLO V11 model as the backbone network with an LSKNet module. The improved YOLO V11 model includes an LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making prediction outputs based on the fused features output by the feature fusion layer.

[0058] It is easy to see that the remote sensing urban solid waste extraction model to be trained in this application is trained by an improved YOLO V11 model optimized from the traditional or basic YOLO V11 model (referring to the original YOLO V11 model). The key point is to replace the original backbone network with the LSKNet module to achieve better feature extraction effect, thereby enabling the model to obtain better urban solid waste extraction effect.

[0059] Specifically, the Large Selective Kernel Network (SKNet) module is a backbone network specifically designed for target detection in remote sensing images. Its core lies in dynamically adjusting a large spatial receptive field to adapt to the different contextual information requirements of different objects.

[0060] In terms of details, the construction of the LSKNet module includes: a large kernel selection sub-block and a feedforward network sub-block, corresponding to: Large Kernel Selection (LK Selection) Sub-block: This module is key to achieving adaptive adjustment of the receptive field. It embeds a Large Selective Kernel (LSK) module, which is essentially a multi-scale large kernel feature extraction and spatial dynamic selection fusion. It mainly includes large convolution decomposition, channel blending and compression, spatial selection mechanisms, and feature fusion and recovery. Feed Forward Nets (FFN) Sub-block: Feature Refinement and Channel Enhancement. The feed forward net sub-block follows the large kernel selection sub-block and its core function is to fuse and refine features in the channel dimension, making up for the shortcomings of the large kernel selection sub-block in channel representation.

[0061] For the original YOLO V11 model, we have: YOLO V11, as a first-stage object detection model, is built around improving accuracy and efficiency. Its architecture consists of three parts: backbone network, feature fusion layer and detection head. The backbone network uses C3k2 dual-branch modules, through Convolution and Residual connections combined with convolution efficiently extract multi-scale features and reduce the number of parameters.

[0062] The feature fusion layer is based on an improved Path Aggregation Network (PANet), replacing the old module C3k2, and is paired with an optimized Spatial Pyramid Pooling Fast (SPFF) kernel to balance receptive field and computational efficiency, thereby enhancing the ability to adapt to multi-scale targets.

[0063] The detection head features a decoupled design independent of anchor boxes, integrating a Distribution Focal Loss (DFL) mechanism to improve bounding box regression accuracy. It employs a dual-label allocation strategy (combining one-to-many and one-to-one during training, and using one-to-one for inference to achieve efficient output of Non-Maximum Suppression (NMS)). During the training phase, positive and negative samples are assigned through a Task-Aligned Assigner. The loss function integrates classification cross-entropy and Inner Complete Intersection over Union (Inner-CIoU), while optimizing the weights of hard samples and data augmentation. Ultimately, it achieves a performance breakthrough of 2-5% improvement in mean Average Precision (mAP).

[0064] In this context, this application improves and optimizes the basic network architecture of YOLO V11 for urban solid waste target detection tasks. First, based on the LSKNet model, it serves as the new backbone network framework. Then, the original backbone network in the original YOLO V111 network model is replaced with the improved LSKNet module. This improvement is primarily aimed at enhancing the target detection network's ability to capture contextual information of different objects in satellite remote sensing imagery, especially its feature extraction performance in complex scenes.

[0065] By combining YOLO V11 with the LSKNet module, the performance of the entire network in remote sensing target detection tasks can be significantly improved. This architecture optimization fully considers the characteristics of remote sensing images. While maintaining the high efficiency of YOLO series models, it further improves the recognition accuracy and positioning accuracy of solid waste targets. The improved network structure has a greater advantage in feature representation and is more suitable for the specific application scenario of urban solid waste detection.

[0066] Furthermore, during the iterative training of the improved model, this application may also employ a series of other optimization strategies to enhance model performance and generalization ability.

[0067] For example, to address the diverse size and imbalanced categories of urban solid waste targets, a dynamic weight balancing mechanism and a multi-task loss function optimization method can be introduced to improve the detection accuracy of small targets and alleviate inter-category learning bias.

[0068] Furthermore, by employing an adaptive learning rate adjustment strategy and an early stopping mechanism, the training process converges to the optimal solution more efficiently and effectively avoids model overfitting. During the validation phase, cross-validation is used to evaluate model performance, further ensuring the model's stability and reliability. Finally, after multiple rounds of iterative optimization, the improved model achieves higher detection accuracy and lower inference latency in urban solid waste extraction tasks, providing an efficient and reliable solution for practical applications.

[0069] To improve the detection efficiency and accuracy of the improved model, this application can also systematically optimize key parameters. Through multiple rounds of iterative training, key parameters such as learning rate, weight decay, and momentum coefficient are continuously optimized. The model performance is verified and optimized by combining multi-dimensional evaluation indicators (such as accuracy, recall, mAP, etc.), and finally, a significant improvement in detection accuracy and inference efficiency is achieved in urban solid waste extraction tasks.

[0070] Furthermore, based on the improved model parameters, this application selects high-resolution satellite remote sensing imagery of the target urban area from another time period as a validation dataset to conduct rapid inference experiments. Through actual inference tasks, the model's operational efficiency in real-world scenarios is evaluated, with a focus on whether its inference speed meets the requirements of real-time monitoring. Simultaneously, combined with expert visual interpretation results, the model's detection performance is comprehensively validated, including target detection recall and precision. Through comparative analysis, model parameters and algorithm configurations are further optimized to ensure high accuracy while achieving efficient inference speed and good practical application results.

[0071] The above solutions can also be combined with Figure 3The following is a logical diagram illustrating the configuration logic of the model in this application, for a more intuitive understanding.

[0072] Furthermore, based on the aforementioned model structure, this application also makes further optimizations in terms of specific processing logic.

[0073] Specifically, as an exemplary embodiment, the remote sensing urban solid waste extraction model includes the following processing logic: Within a multi-level buffer zone, different date elements, climate elements, weather elements, terrain elements, and the current specific monitoring time are used as basic influencing factors. The location, type, size, and quantity of key features are used as the first enhancing influencing factor. The orientation and distance relationships between solid waste sites and key features are used as the second enhancing influencing factor. The distance and path relationships between different solid waste sites are used as the third enhancing influencing factor. Fixed site extraction processing is carried out. Key features include greenery, factory buildings, residential buildings, construction land, warehousing land, waste treatment plants, and main roads.

[0074] Among them, the multi-level buffer can be understood as the processing range of different sizes, such as the processing range expanded at intervals of 10m, 20m, 30m, 40m, etc. based on the same center point. Different buffers can have different weights in the determination of urban solid waste, so that a delicate contribution adjustment mechanism can play a better role in the identification of urban solid waste.

[0075] Furthermore, different date elements, climate elements, weather elements, terrain elements, and the current specific monitoring time are objective elements that can cause fluctuations in the urban solid waste identification effect in this application in terms of details; Similarly, this application also believes that in reality, there is a certain correlation between the dumping of urban solid waste and the surrounding environment. This corresponds to whether there is a situation in a certain area that will generate urban solid waste, and also to whether there is a suitable situation in a certain area for placing urban solid waste. At the same time, different land features may also interfere with the identification of nearby urban solid waste. For these three situations, it is understandable that in specific applications, one can be selected or freely combined to form different degrees of auxiliary effects for urban solid waste identification.

[0076] Therefore, the location, type, size, and quantity of key features can be considered as an additional type of influencing factor that needs to be paid attention to, and specifically referred to as the first enhanced influencing factor.

[0077] Furthermore, while taking into account nearby key features, this application can also focus on the orientation and distance relationships between solid waste sites and key features, and introduce them as a second enhancing factor into the urban solid waste identification process as a new auxiliary reference element.

[0078] Furthermore, this application also believes that, regardless of whether the same type of urban solid waste or the same urban solid waste disposal party is involved, there may be a certain degree of correlation between urban solid waste disposal behaviors in actual situations. The unknown relationship can be reflected from the distance and path relationship between different solid waste locations. Therefore, this application can continue to use the distance and path relationship between different solid waste locations as a third enhancing influence factor to assist the model in performing a more refined identification task that is different from the basic urban solid waste identification logic.

[0079] Furthermore, temporal features can be further incorporated into the model's processing logic to perform more nuanced urban solid waste identification from a time perspective.

[0080] After the model training is completed, it can obviously be put into practical applications to perform dynamic monitoring tasks of urban solid waste in the past, present and even future.

[0081] Correspondingly, as an exemplary embodiment, after training a remote sensing urban solid waste extraction model based on urban solid waste sample data, the method of this application may further include: Acquire the current remote sensing image to be processed; The remote sensing image to be processed is fed into the remote sensing urban solid waste extraction model, and the solid waste extraction results output by the remote sensing urban solid waste extraction model are obtained. The solid waste extraction results are displayed through a corresponding interactive interface.

[0082] It is understandable that the remote sensing image to be processed usually follows the same remote sensing image acquisition mechanism / data source as the preceding 4-band optical satellite remote sensing image data. Of course, depending on the model training work and the specific circumstances of actual application, it can also follow a different remote sensing image acquisition mechanism / data source than the preceding 4-band optical satellite remote sensing image data. This can be flexibly adjusted according to the actual situation.

[0083] Thus, once the remote sensing image to be processed is obtained under the triggering conditions of tasks initiated manually, automatically by the system, or received from the outside, it can be handed over to the remote sensing urban solid waste extraction model for remote sensing urban solid waste extraction processing, and the solid waste extraction results output by the remote sensing urban solid waste extraction model can be obtained.

[0084] At this point, it is obvious that the solid waste extraction results can be processed according to the implemented or pre-configured output strategy.

[0085] For example, it can perform local storage, off-site storage, result display, result forwarding, output completion prompts, or further data analysis, etc., which can be adaptively configured according to actual needs.

[0086] Taking the results display as an example, the processing equipment can display the solid waste extraction results through an interactive display interface configured with user-friendly interactive functions. This interactive display interface is easy to understand; it can dynamically adjust the display scale of different dimensions of the results according to the user's interactive operation, thereby meeting the flexible and ever-changing needs for high-quality results presentation and thus obtaining a better user experience.

[0087] In conclusion, regarding the above solutions, this application, based on the YOLO V11 model and through corresponding adaptive optimizations, effectively resolves the contradiction between efficiency and accuracy in solid waste identification using traditional methods. It significantly improves the processing capabilities for large-area, large-scale remote sensing images, meeting the high-efficiency and accurate high-quality application requirements of urban solid waste remote sensing monitoring, and thus providing excellent data support for urban solid waste treatment.

[0088] The above is an introduction to the remote sensing urban solid waste extraction method based on the improved YOLO V11 provided in this application. In order to facilitate better implementation of the remote sensing urban solid waste extraction method based on the improved YOLO V11 provided in this application, this application also provides a remote sensing urban solid waste extraction device based on the improved YOLO V11 from the perspective of functional modules.

[0089] See Figure 4 , Figure 4 This is a schematic diagram of a remote sensing urban solid waste extraction device based on the improved YOLO V11 according to this application. In this application, the remote sensing urban solid waste extraction device 400 based on the improved YOLO V11 may specifically include the following structure: The acquisition unit 401 is used to acquire 4-band optical satellite remote sensing image data that meets the sample quality requirements for the target urban area. Production unit 402 is used to produce urban solid waste sample data after preprocessing the 4-band optical satellite remote sensing image data; Training unit 403 is used to train a remote sensing urban solid waste extraction model based on urban solid waste sample data. The remote sensing urban solid waste extraction model is trained on the basis of an improved YOLO V11 model. The improved YOLO V11 model replaces the original YOLO V11 model as the backbone network with an LSKNet module. The improved YOLO V11 model includes an LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making prediction outputs based on the fused features output by the feature fusion layer.

[0090] In one exemplary embodiment, preprocessing includes sample amplification and data augmentation. Sample amplification processing includes random cropping, random inversion, random rotation, and random illumination adjustment; Data augmentation processing includes radiometric calibration, atmospheric correction, geometric registration, image fusion, and image mosaicking. Image fusion involves the fusion of multispectral and panchromatic images, while image mosaicking involves stitching together several images after image fusion using techniques including geometric mosaicking, color balancing, feathering, and de-overlapping.

[0091] In yet another exemplary embodiment, the process of generating urban solid waste sample data includes the following steps: Draw corresponding urban solid waste area vector data for urban solid waste sample data on the designated GIS platform; Based on the urban solid waste regional vector data, the urban solid waste sample data is cropped to obtain the urban solid waste image dataset; The urban solid waste image dataset was used to create urban solid waste sample data.

[0092] In yet another exemplary embodiment, the urban solid waste image dataset is processed into urban solid waste sample data, including: The urban solid waste image dataset was labeled, and the labeled data was augmented. The data augmentation process included preprocessing the labeled data to remove noise and redundant information, and optimizing and adjusting the irregular label boundaries by combining a set threshold range and image augmentation methods. The annotation data in XML format is converted into a tag file in TXT format. The tag file is represented by normalized coordinates. Each line records the information of a target object, including the category number and the coordinates of the four vertices. The parameters are separated by spaces, and different targets are distinguished by newline characters.

[0093] In yet another exemplary embodiment, an automated annotation tool is used during the annotation process, and the corresponding annotation strategy includes the following: Based on the spectral reflectance, texture and morphological features of urban solid waste target objects, and using a bounding box containing urban solid waste target objects that reflects the actual boundaries of the objects as the output, automated annotation is performed.

[0094] In yet another exemplary embodiment, the remote sensing urban solid waste extraction model includes the following processing logic: Within a multi-level buffer zone, different date elements, climate elements, weather elements, terrain elements, and the current specific monitoring time are used as basic influencing factors. The location, type, size, and quantity of key features are used as the first enhancing influencing factor. The orientation and distance relationships between solid waste sites and key features are used as the second enhancing influencing factor. The distance and path relationships between different solid waste sites are used as the third enhancing influencing factor. Fixed site extraction processing is carried out. Key features include greenery, factory buildings, residential buildings, construction land, warehousing land, waste treatment plants, and main roads.

[0095] In yet another exemplary embodiment, the apparatus further includes an application unit 404 for: Acquire the current remote sensing image to be processed; The remote sensing image to be processed is fed into the remote sensing urban solid waste extraction model, and the solid waste extraction results output by the remote sensing urban solid waste extraction model are obtained. The solid waste extraction results are displayed through a corresponding interactive interface.

[0096] This application also provides a processing device from a hardware architecture perspective. As mentioned earlier, in practice, a processing device may exist as a device cluster. In this case, each device in the device cluster can also be referred to as a processing device. See [reference needed]. Figure 5 , Figure 5 This diagram illustrates a structural schematic of the processing device of this application. Specifically, the processing device may include a processor 501, a memory 502, and an input / output device 503. The processor 501 executes the computer program stored in the memory 502 to implement, for example... Figure 1 The corresponding steps of the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the embodiment; or, when the processor 501 executes the computer program stored in the memory 502, it implements as follows: Figure 4 Corresponding to the functions of each unit in the embodiment, the memory 502 is used to store the functions executed by the processor 501 as described above. Figure 1 The corresponding embodiment includes the computer program required for the remote sensing urban solid waste extraction method based on the improved YOLO V11.

[0097] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 502 and executed by processor 501 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a computer device.

[0098] The processing device may include, but is not limited to, processor 501, memory 502, and input / output device 503. Those skilled in the art will understand that the illustrations are merely examples of the processing device and do not constitute a limitation on the processing device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the processing device may also include network access devices, buses, etc., and processor 501, memory 502, input / output device 503, etc., are connected via a bus.

[0099] Processor 501 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the processing device, connecting various parts of the device through various interfaces and lines.

[0100] The memory 502 can be used to store computer programs and / or modules. The processor 501 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 502 and by calling data stored in the memory 502. The memory 502 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function, etc.; the data storage area may store data created according to the use of the processing device, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0101] When processor 501 executes a computer program stored in memory 502, it can specifically perform the following functions: For the target urban area, acquire 4-band optical satellite remote sensing image data that meet the sample quality requirements; After preprocessing the 4-band optical satellite remote sensing image data, it was further processed into urban solid waste sample data; A remote sensing urban solid waste extraction model was trained based on urban solid waste sample data. The model was trained on an improved YOLO V11 model, in which the original YOLO V11 model was replaced with an LSKNet module as the backbone network. The improved YOLO V11 model consists of an LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making predictions based on the fused features output by the feature fusion layer.

[0102] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the remote sensing urban solid waste extraction device, processing equipment, and its corresponding units based on the improved YOLO V11 described above can be found in the following reference: Figure 1 The description of the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the corresponding embodiment will not be repeated here.

[0103] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0104] Therefore, this application provides a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute the present application. Figure 1 The steps of the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the corresponding embodiment can be referred to as follows for specific operations. Figure 1 The description of the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the corresponding embodiment will not be repeated here.

[0105] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0106] Because of the instructions stored in the computer-readable storage medium, the present application can be executed as described above. Figure 1The steps of the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the corresponding embodiment can therefore achieve the results of this application. Figure 1 The beneficial effects that can be achieved by the remote sensing urban solid waste extraction method based on the improved YOLO V11 in the corresponding embodiments are detailed in the preceding description and will not be repeated here.

[0107] The above provides a detailed description of the remote sensing urban solid waste extraction method, apparatus, processing equipment, and computer-readable storage medium based on the improved YOLO V11 provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the core ideas of this application; furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A remote sensing method for urban solid waste extraction based on an improved YOLO V11, characterized in that, The method includes: For the target urban area, acquire 4-band optical satellite remote sensing image data that meet the sample quality requirements; After preprocessing the four-band optical satellite remote sensing image data, it is further processed into urban solid waste sample data; A remote sensing urban solid waste extraction model is trained based on the urban solid waste sample data. The remote sensing urban solid waste extraction model is trained on an improved YOLO V11 model. The improved YOLO V11 model replaces the original YOLO V11 model as the backbone network with an LSKNet module. The improved YOLO V11 model includes the LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing images input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making prediction outputs based on the fused features output by the feature fusion layer.

2. The method according to claim 1, characterized in that, The preprocessing includes sample amplification and data augmentation. The sample amplification process includes random cropping, random reversal, random rotation, and random illumination adjustment; The data augmentation process includes radiometric calibration, atmospheric correction, geometric registration, image fusion, and image mosaicking. The image fusion involves the fusion of multispectral and panchromatic images, and the image mosaicking involves stitching together several images after image fusion using geometric mosaicking, color balancing, feathering, and de-overlapping.

3. The method according to claim 1, characterized in that, The process of creating the urban solid waste sample data includes the following steps: On the designated GIS platform, draw the corresponding urban solid waste area vector data for the urban solid waste sample data; Based on the urban solid waste area vector data, the urban solid waste sample data is cropped to obtain an urban solid waste image dataset; The urban solid waste image dataset is used to create urban solid waste sample data.

4. The method according to claim 3, characterized in that, The step of creating urban solid waste sample data from the urban solid waste image dataset includes: The urban solid waste image dataset is labeled, and the labeled data is augmented. The data augmentation process includes preprocessing the labeled data to remove noise and redundant information, and optimizing and adjusting the irregular label boundaries by combining a set threshold range and image augmentation methods. The annotation data in XML format is converted into a tag file in TXT format. The tag file is represented by normalized coordinates, and each line records the information of a target object. The information of the target object includes a category number and the coordinates of four vertices. The parameters are separated by spaces, and different targets are distinguished by newline characters.

5. The method according to claim 4, characterized in that, Automated annotation tools are used during the annotation process, and the corresponding annotation strategies include the following: Based on the three features of urban solid waste target objects—spectral reflectance, texture, and morphology—automatic annotation is performed using a bounding box containing the urban solid waste target objects and reflecting the actual boundaries of the objects as the output.

6. The method according to claim 1, characterized in that, The remote sensing urban solid waste extraction model includes the following processing logic: Within a multi-level buffer zone, different date elements, climate elements, weather elements, terrain elements, and the current specific monitoring time are used as basic influencing factors. The location, type, size, and quantity of key features are used as the first enhancing influencing factor. The orientation and distance relationships between solid waste sites and the key features are used as the second enhancing influencing factor. The distance and path relationships between different solid waste sites are used as the third enhancing influencing factor. Fixed site extraction processing is carried out. The key features include greenery, factory buildings, residential buildings, construction land, warehousing land, waste treatment plants, and main roads.

7. The method according to claim 1, characterized in that, After training the remote sensing urban solid waste extraction model based on the urban solid waste sample data, the method further includes: Acquire the current remote sensing image to be processed; The remote sensing image to be processed is sent into the remote sensing urban solid waste extraction model, and the solid waste extraction result output by the remote sensing urban solid waste extraction model is obtained. The solid waste extraction results are displayed through a corresponding interactive interface.

8. A remote sensing urban solid waste extraction device based on an improved YOLO V11, characterized in that, The device includes: The acquisition unit is used to acquire 4-band optical satellite remote sensing image data that meets the sample quality requirements for the target urban area. The production unit is used to produce urban solid waste sample data after preprocessing the 4-band optical satellite remote sensing image data; The training unit is used to train a remote sensing urban solid waste extraction model based on the urban solid waste sample data. The remote sensing urban solid waste extraction model is trained on an improved YOLO V11 model. The improved YOLO V11 model replaces the original YOLO V11 model with an LSKNet module as the backbone network. The improved YOLO V11 model includes the LSKNet module, a feature fusion layer, and a detection head. The LSKNet module is responsible for extracting features from the remote sensing image input to the model. The feature fusion layer is responsible for fusing the multi-level features output by the LSKNet module. The detection head is responsible for making prediction outputs based on the fused features output by the feature fusion layer.

9. A processing device, characterized in that, The method includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the method as described in any one of claims 1 to 7 when it invokes the computer program in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the method of any one of claims 1 to 7.

Images