A decoration waste detection method and device

Abstract

The present invention provides a method and device for detecting renovation waste, comprising: obtaining an image to be detected, the image to be detected being captured by a terminal installed at a garbage dump station and photographing vehicles entering or leaving the dump station; inputting the image to be detected into a pre-trained renovation waste detection model, the pre-trained renovation waste detection model comprising a vehicle discrimination sub-model and a renovation waste detection sub-model; using the vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result; and if the vehicle discrimination result indicates that the image to be detected includes a vehicle, inputting the image to be detected into the renovation waste detection sub-model to obtain a renovation waste detection result. The present invention can improve the detection rate and accuracy of renovation waste, facilitating platform managers to standardize renovation waste management based on the renovation waste detection results.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a decoration waste detection method and device. Background Art

[0002] my country's cities have experienced rapid development in recent years, and the removal of urban construction waste has become a crucial component of urban management. Accurately classifying renovation waste can significantly aid in its management. Currently, landfill sites require staff supervision and inspection to identify the types of renovation waste. Each site requires a different staff member to review the waste, a process that consumes significant human resources. Furthermore, staff judgment is subjective, leading to classification errors and negative customer experiences. Therefore, how to more effectively identify renovation waste has become a technical challenge that needs to be addressed. Summary of the Invention

[0003] The present invention provides a decoration waste detection method and device to solve the above problems.

[0004] The present invention provides a decoration waste detection method, comprising: obtaining an image to be detected, wherein the image to be detected is obtained by photographing vehicles entering and leaving the garbage station by a terminal set up at the garbage station;

[0005] Inputting the image to be detected into a pre-trained decoration waste detection model, wherein the pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model;

[0006] Using the vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result;

[0007] When the vehicle identification result indicates that the image to be detected includes a vehicle, the image to be detected is input into a decoration waste detection sub-model to obtain a decoration waste detection result.

[0008] According to a decoration waste detection method provided by the present invention, the decoration waste detection result includes decoration waste category information and decoration waste location information;

[0009] Accordingly, after obtaining the decoration waste detection result, the method further includes:

[0010] The decoration waste detection results are screened according to the decoration waste category information to obtain screened decoration waste detection results.

[0011] According to a decoration waste detection method provided by the present invention, before inputting the image to be detected into a pre-trained decoration waste detection model, the method further includes:

[0012] Scaling the image to be detected to a target size;

[0013] In the case that the scaled image to be detected still does not meet the target size, padding processing is performed on the scaled image to be detected so that the size of the padded image to be detected meets the target size.

[0014] According to a decoration waste detection method provided by the present invention, the pre-trained decoration waste detection model is trained in the following manner:

[0015] Obtain training data and corresponding label data for renovation waste scenarios, and build a renovation waste detection model;

[0016] Inputting the training data into a vehicle discrimination sub-model to predict a vehicle discrimination result during training, and performing a loss calculation based on the vehicle discrimination result during training and the label data to obtain a first loss;

[0017] Inputting the training data into the decoration waste detection sub-model to predict the decoration waste detection results during training, and performing loss calculation based on the decoration waste detection results during training and the label data to obtain a second loss;

[0018] The decoration waste detection model is trained according to the first loss and the second loss to obtain a trained decoration waste detection model.

[0019] According to a decoration waste detection method provided by the present invention, after obtaining training data and corresponding label data in a decoration waste scenario, the method further includes:

[0020] Extracting bounding box coordinates from the training data according to the label data; initializing multiple anchor boxes to obtain an initial anchor box, and obtaining anchor box coordinates corresponding to the initial anchor box;

[0021] Calculate the IoU between the initial anchor box and the bounding box according to the bounding box coordinates and the anchor box coordinates. When the IoU value exceeds the threshold, the corresponding initial anchor box is used as the target anchor box.

[0022] Accordingly, inputting the training data into the decoration waste detection sub-model includes:

[0023] The training data and the target anchor box are input into the decoration waste detection sub-model.

[0024] According to a decoration waste detection method provided by the present invention, the decoration waste detection sub-model is constructed based on a predefined target detection algorithm and is a pre-trained model.

[0025] According to a decoration waste detection method provided by the present invention, the decoration waste category information includes at least one or more of large garbage, hard lumps, residues or wood blocks;

[0026] Accordingly, when the decoration waste category information is residue, filtering the decoration waste detection result according to the decoration waste category information to obtain the filtered decoration waste detection result includes:

[0027] Calculate the area of the debris based on the location information of the renovation waste corresponding to the debris, and determine whether the calculated area exceeds a preset debris area threshold;

[0028] If the calculated area exceeds a preset debris area threshold, reconfirming the decoration waste detection result;

[0029] When the calculated area does not exceed a preset residue area threshold, the decoration waste detection result is output as a final decoration waste detection result.

[0030] The present invention also provides a decoration waste detection device, comprising:

[0031] An image acquisition module is used to acquire images to be detected, wherein the images to be detected are obtained by photographing vehicles entering and leaving the garbage dump station by a terminal installed at the garbage dump station;

[0032] An image input module, configured to input the image to be detected into a pre-trained decoration waste detection model, wherein the pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model;

[0033] A vehicle identification module, configured to use the vehicle identification sub-model to identify the image to be detected and obtain a vehicle identification result;

[0034] The decoration waste detection module is used to input the image to be detected into the decoration waste detection sub-model to obtain a decoration waste detection result when the vehicle identification result indicates that the image to be detected includes a vehicle.

[0035] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, any one of the above-mentioned methods for detecting renovation waste is implemented.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, any of the above-mentioned decoration waste detection methods is implemented.

[0037] The present invention provides a method and device for detecting renovation waste. The method uses a vehicle identification sub-model to identify images to be detected, obtaining a vehicle identification result. Invalid images are filtered out based on the vehicle identification result. Images to be detected, including those of non-enclosed vehicles, are then input into the renovation waste detection sub-model to predict renovation waste detection results. This invention improves the detection rate and accuracy of renovation waste, facilitating standardized management of renovation waste by platform administrators based on the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 This is one of the flow charts of the decoration waste detection method provided by an embodiment of the present invention;

[0040] Figure 2 This is the second flow chart of the decoration waste detection method provided by the embodiment of the present invention;

[0041] Figure 3 A schematic structural diagram of a decoration waste detection device provided by an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0044] With the rapid development of artificial intelligence, intelligent transportation systems have achieved remarkable results. Through digital and visual management, information exchange is achieved, effectively solving many transportation challenges. Renovation waste detection technology is a crucial component of landfill sites. The identification and location of renovation waste is fundamental and crucial, and errors in any step can lead to deviations in the final identification results. Previously, renovation waste was subjectively determined by management personnel, which posed management risks and a waste of human resources. Furthermore, in some scenarios, it's difficult to accurately identify the corresponding category of renovation waste, making standardized management difficult.

[0045] In response to the above-mentioned problems, the present invention proposes a method for detecting renovation waste, which can realize real-time identification of renovation waste categories and provide corresponding feedback, greatly improving the accuracy and applicability of renovation waste detection. AI renovation waste intelligentization can also be applied to more application scenarios that require identification. The renovation waste detection method and device proposed in the present invention are described in detail below with reference to the accompanying drawings.

[0046] Figure 1 This is one of the flow charts of the decoration waste detection method provided by an embodiment of the present invention; Figure 2 This is a second flow chart of the decoration waste detection method provided by an embodiment of the present invention; Figure 1 As shown in FIG2 , the decoration waste detection method includes:

[0047] S101, obtaining an image to be detected.

[0048] The images to be detected are captured by a terminal (such as a camera) installed at the garbage dump station, capturing images of vehicles entering and exiting the dump station. These images are then transmitted to a renovation waste detection platform that is communicatively connected to the terminal. The terminal can directly send videos of vehicles entering and exiting the dump station to the renovation waste detection platform, which then extracts frames for processing. Alternatively, the terminal can convert the video stream into video frames and transmit them to the platform. It should be noted that the garbage dump station here refers to a garbage dump station used to manage construction waste.

[0049] S102: Input the image to be detected into a pre-trained decoration waste detection model.

[0050] The pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model.

[0051] S103: Using the vehicle identification sub-model to identify the image to be detected, to obtain a vehicle identification result.

[0052] In this embodiment, the image to be detected directly obtained by the terminal of the garbage station may show invalid images such as decoration waste only in a corner of the image, or the vehicle that appears is not loaded with decoration waste, etc. Therefore, the image to be detected is filtered before performing decoration waste detection.

[0053] Specifically, a classifier (i.e., the vehicle discrimination sub-model) can be used to distinguish between images containing vehicles and those not containing vehicles. Images containing vehicles are used as input to the renovation waste detection sub-model, while images not containing vehicles are pushed to the administrator of the renovation waste detection platform for confirmation.

[0054] It should be noted that the vehicles in this embodiment are non-enclosed vehicles, and the items loaded in these vehicles can be seen in the image. If the vehicle in the image to be detected is a closed vehicle, the image is sent to the management staff for evaluation. In this embodiment, images that cannot be detected as renovation waste are filtered out.

[0055] S104 : When the vehicle identification result indicates that the image to be detected includes a vehicle, the image to be detected is input into a decoration waste detection sub-model to obtain a decoration waste detection result.

[0056] In this step, after the aforementioned filtering, the image containing the non-enclosed vehicle is fed into the renovation waste detection sub-model, which is built and trained based on conventional object detection algorithms (such as the R-CNN series and Yolo series). This sub-model can identify objects loaded in non-enclosed vehicles and obtain renovation waste detection results that include the type and location of the renovation waste.

[0057] The renovation waste detection method provided by an embodiment of the present invention uses a vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result. Invalid images are filtered out based on the vehicle discrimination result. Images to be detected, including those of non-enclosed vehicles, are then input into the renovation waste detection sub-model to predict the renovation waste detection result. This method can improve the detection rate and accuracy of renovation waste, facilitate platform administrators to standardize renovation waste management based on the detection results, reduce manual intervention, and improve work efficiency.

[0058] Furthermore, the decoration waste detection result includes decoration waste category information and decoration waste location information.

[0059] Accordingly, after obtaining the decoration waste detection result, the method further includes:

[0060] The decoration waste detection results are screened according to the decoration waste category information to obtain screened decoration waste detection results.

[0061] Specifically, the renovation waste category information includes at least one or more of bulky waste, hard lumps, residues or wood blocks.

[0062] Accordingly, when the decoration waste category information is residue, filtering the decoration waste detection result according to the decoration waste category information to obtain the filtered decoration waste detection result includes:

[0063] The area of the debris is calculated based on the renovation waste location information corresponding to the debris, and it is determined whether the calculated area exceeds a preset debris area threshold.

[0064] If the calculated area exceeds the preset debris area threshold, the renovation waste detection result is reconfirmed. In this embodiment, the renovation waste detection platform sends the image to be detected for which confirmation of the renovation waste detection result is required to the corresponding management personnel, who then conduct the renovation waste detection and the result is then output as the final result after the management personnel's confirmation.

[0065] When the calculated area does not exceed a preset residue area threshold, the decoration waste detection result is output as a final decoration waste detection result.

[0066] In addition, the decoration waste detection results also include the empty vehicle situation, that is, the vehicle in the image to be detected is not loaded with decoration waste.

[0067] It should be noted that, in this embodiment, the area threshold is only judged for the case where the decoration waste category is residue to determine whether the corresponding decoration waste detection result can be used as the final output. In other embodiments of the present invention, for decoration waste with uncertain shapes like residue, the size of the area can also be used to determine whether the detection result is abnormal. Under normal circumstances, it is directly output; under abnormal circumstances, it is pushed to relevant management personnel for review, thereby ensuring the accuracy of decoration waste detection.

[0068] In some embodiments of the present invention, for renovation waste such as hard blocks and wooden blocks, the present invention uses traditional image algorithms to extract the contour features of objects such as hard blocks and wooden blocks based on the renovation waste detection results, and performs image processing such as corrosion and expansion on the extracted contour features to obtain more detailed features of the objects. Identification is performed based on these detailed features and contour features. For example, hard blocks generally do not have a large difference in length and width, so the corresponding length and width contour features will have a smaller difference; while wooden blocks are generally slender in shape, so their corresponding contour features will be displayed accordingly. Therefore, for renovation waste with a stable shape and a relatively obvious contour, the above-mentioned method can be used to perform a posteriori processing of the renovation waste detection results, thereby improving the accuracy of renovation waste detection.

[0069] In addition to the post-processing mentioned above, in some embodiments of the present invention, the illumination of the image is also taken into consideration. Specifically, the recognition confidence of each category in the identification of renovation waste during the day will be higher, which is higher than the set threshold. However, the confidence level of the identification of renovation waste at night is generally lower. If it is still compared with the previously set threshold, it is easy to fail to accurately identify the renovation waste. Therefore, for the detection of renovation waste in different time periods, different thresholds are set in the renovation waste detection sub-model to improve the recognition accuracy.

[0070] The decoration waste detection method provided in the embodiment of the present invention determines whether the decoration waste detection results are abnormal for different categories of decoration waste through corresponding abnormality judgment methods. In the case of abnormality, the decoration waste detection results are screened to obtain the most accurate decoration waste detection results, further improving the detection accuracy of the decoration waste detection method.

[0071] Furthermore, before inputting the image to be detected into a pre-trained decoration waste detection model, the method further includes:

[0072] The image to be detected is scaled to a target size, wherein the target size is a size pre-set according to the network layer of the decoration waste detection model.

[0073] In the case that the scaled image to be detected still does not meet the target size, padding processing is performed on the scaled image to be detected so that the size of the padded image to be detected meets the target size.

[0074] Specifically, before testing the image to be tested, it is first scaled to the target size. If the image after scaling can only approach the target size infinitely but cannot reach the target size completely (i.e., one side of the image is blank), the scaled image to be tested is padded (with gray or black edges) to ensure that the final image to be tested meets the target size.

[0075] The decoration waste detection method provided by the embodiment of the present invention unifies the size of the image to be detected by scaling and padding, and can effectively utilize the receptive field information of the neural network in the subsequent model detection process. That is, for the side that does not meet the conditions after scaling, it can be filled with a gray edge or a black edge until it can be divided by the receptive field, thereby further improving the detection effect.

[0076] Furthermore, the pre-trained decoration waste detection model is trained in the following way:

[0077] Obtain training data from a renovation waste scenario and its corresponding label data, and build a renovation waste detection model. It should be noted that the training data in this embodiment is images from real scenes, not existing public training sets. Models trained with real data have higher detection accuracy than models trained with public training sets.

[0078] The training data is input into the vehicle discrimination sub-model to predict the vehicle discrimination result during training, and a loss is calculated based on the vehicle discrimination result during training and the label data to obtain a first loss. In this embodiment, the vehicle discrimination sub-model is a conventional classifier.

[0079] The training data is input into the renovation waste detection sub-model to predict the renovation waste detection results during training. A loss is calculated based on the renovation waste detection results during training and the label data to obtain a second loss. In this embodiment, the renovation waste detection sub-model is constructed based on a predefined object detection algorithm. The predefined object detection algorithm is a conventional object detection algorithm, such as the YOLO series.

[0080] The renovation waste detection model is trained based on the first and second losses to obtain a trained renovation waste detection model. Specifically, the sum of the first and second losses is calculated to obtain a total loss. Based on the total loss, the network parameters of the vehicle discrimination sub-model and the renovation waste detection sub-model are updated using a backpropagation algorithm until the two sub-models converge. Updates are then stopped to obtain a trained renovation waste detection model.

[0081] The decoration waste detection method provided by the embodiment of the present invention obtains a final decoration waste detection model by jointly training two sub-models.

[0082] Furthermore, after obtaining the training data and corresponding label data in the renovation waste scenario, the method further includes:

[0083] The bounding box coordinates are extracted from the training data according to the label data.

[0084] Initialize multiple anchor boxes to obtain the initial anchor box, and obtain the anchor box coordinates corresponding to the initial anchor box.

[0085] The IoU between the initial anchor box and the bounding box is calculated according to the bounding box coordinates and the anchor box coordinates. When the IoU value exceeds the threshold, the corresponding initial anchor box is used as the target anchor box.

[0086] Accordingly, inputting the training data into the decoration waste detection sub-model includes:

[0087] The training data and the target anchor box are input into the decoration waste detection sub-model.

[0088] Specifically, during the model training process, anchor boxes are clustered based on the real data and its corresponding label data to accelerate the convergence and fitting rate of the training process and improve the training rate of the decoration waste detection sub-model. The specific clustering process includes:

[0089] 1) Extract bounding box coordinates based on the data annotation of the renovation waste scene;

[0090] 2) Initialize 9 anchor boxes and calculate the IoU between each bounding box and the anchor box;

[0091] 3) Calculate the accuracy of the anchor box and select the set of anchor boxes with the largest IoU.

[0092] The decoration waste detection method provided by the embodiment of the present invention clusters the anchor boxes to determine the target anchor box closest to the bounding box, thereby accelerating the convergence rate in the subsequent decoration waste detection sub-model training process.

[0093] Furthermore, the decoration waste detection sub-model is constructed based on a predefined target detection algorithm and is a pre-trained model.

[0094] In this embodiment, the decoration waste detection sub-model is a pre-trained model. During the above training process, the parameters of the pre-trained decoration waste detection sub-model are adjusted to further improve the training rate and the detection accuracy of the model.

[0095] In addition, the present invention can also perform data augmentation on the training data to improve the detection accuracy of the model.

[0096] In addition, in the above-mentioned process of screening the decoration waste detection results according to the decoration waste category information, images that require confirmation by management personnel (that is, images that are not accurately detected by the decoration waste detection sub-model) can be used as difficult samples to retrain the decoration waste detection sub-model, thereby further improving the detection accuracy of the model.

[0097] In the label data corresponding to the training data, the categories of the training data can be classified more finely, and then the threshold can be dynamically adjusted according to the results of the algorithm recognition, so as to achieve a soft labeling effect and complete the detailed identification of decoration waste.

[0098] The following describes the decoration waste detection device provided by the present invention. The decoration waste detection device described below and the decoration waste detection method described above can be referenced to each other.

[0099] Figure 3 This is a schematic diagram of the structure of a decoration waste detection device provided by an embodiment of the present invention, as shown in FIG. Figure 3 As shown, the decoration waste detection device includes an image acquisition module 301 , an image input module 302 , a vehicle identification module 303 and a decoration waste detection module 304 .

[0100] The image acquisition module 301 is used to acquire an image to be detected.

[0101] The images to be detected are captured by a terminal (such as a camera) installed at the garbage dump station, capturing images of vehicles entering and exiting the dump station. These images are then transmitted to a renovation waste detection platform that is communicatively connected to the terminal. The terminal can directly send videos of vehicles entering and exiting the dump station to the renovation waste detection platform, which then extracts frames for processing. Alternatively, the terminal can convert the video stream into video frames and transmit them to the platform. It should be noted that the garbage dump station here refers to a garbage dump station used to manage construction waste.

[0102] The image input module 302 is used to input the image to be detected into a pre-trained decoration waste detection model.

[0103] The pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model.

[0104] The vehicle identification module 303 is used to use the vehicle identification sub-model to identify the image to be detected to obtain a vehicle identification result.

[0105] In this embodiment, the image to be detected directly obtained by the terminal of the garbage station may show invalid images such as decoration waste only in a corner of the image, or the vehicle that appears is not loaded with decoration waste, etc. Therefore, the image to be detected is filtered before performing decoration waste detection.

[0106] Specifically, a classifier (i.e., the vehicle discrimination sub-model) can be used to distinguish between images containing vehicles and those not containing vehicles. Images containing vehicles are used as input to the renovation waste detection sub-model, while images not containing vehicles are pushed to the administrator of the renovation waste detection platform for confirmation.

[0107] It should be noted that the vehicles in this embodiment are non-enclosed vehicles, and the items loaded in these vehicles can be seen in the image. If the vehicle in the image to be detected is a closed vehicle, the image is sent to the management staff for evaluation. In this embodiment, images that cannot be detected as renovation waste are filtered out.

[0108] The decoration waste detection module 304 is configured to input the image to be detected into a decoration waste detection sub-model to obtain a decoration waste detection result when the vehicle identification result indicates that the image to be detected includes a vehicle.

[0109] In this module, after the aforementioned filtering, images containing non-enclosed vehicles are fed into the renovation waste detection sub-model, which is built and trained based on conventional object detection algorithms (such as the R-CNN and Yolo algorithms). This sub-model can identify objects loaded in non-enclosed vehicles and obtain renovation waste detection results, including the type and location of the waste.

[0110] The renovation waste detection device provided by an embodiment of the present invention uses a vehicle identification sub-model to identify the image to be detected, obtain a vehicle identification result, and filter out invalid images based on the vehicle identification result. Images to be detected, including those of non-enclosed vehicles, are then input into the renovation waste detection sub-model to predict the renovation waste detection result. This invention can improve the detection rate and accuracy of renovation waste, facilitate platform administrators to standardize renovation waste management based on the renovation waste detection results, reduce manual intervention, and improve work efficiency.

[0111] Figure 4 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute a decoration waste detection method, which includes: obtaining an image to be detected, wherein the image to be detected is obtained by photographing vehicles entering and leaving the garbage dump by a terminal installed at the garbage dump; inputting the image to be detected into a pre-trained decoration waste detection model, wherein the pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model; using the vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result; and when the vehicle discrimination result indicates that the image to be detected includes a vehicle, inputting the image to be detected into the decoration waste detection sub-model to obtain a decoration waste detection result.

[0112] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0113] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the decoration waste detection method provided by the above-mentioned method, the decoration waste detection method comprising: obtaining an image to be detected, the image to be detected being obtained by photographing vehicles entering and leaving the garbage dump by a terminal set up at the garbage dump; inputting the image to be detected into a pre-trained decoration waste detection model, the pre-trained decoration waste detection model comprising a vehicle discrimination sub-model and a decoration waste detection sub-model; using the vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result; when the vehicle discrimination result is characterized as including a vehicle in the image to be detected, inputting the image to be detected into the decoration waste detection sub-model to obtain a decoration waste detection result.

[0114] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. That is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0115] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or certain parts of the embodiment.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A decoration waste detection method, characterized in that: include: Acquiring an image to be detected, wherein the image to be detected is obtained by photographing vehicles entering or leaving the garbage dump by a terminal installed at the garbage dump; Inputting the image to be detected into a pre-trained decoration waste detection model, wherein the pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model; Using the vehicle discrimination sub-model to discriminate the image to be detected to obtain a vehicle discrimination result; If the vehicle identification result indicates that the image to be detected includes a vehicle, the image to be detected is input into a decoration waste detection sub-model to obtain a decoration waste detection result; The decoration waste detection result includes decoration waste category information and decoration waste location information; Accordingly, after obtaining the decoration waste detection result, the method further includes: Filtering the decoration waste detection results according to the decoration waste category information to obtain filtered decoration waste detection results; The renovation waste category information includes at least one or more of bulky waste, hard lumps, residues or wood blocks; Accordingly, when the decoration waste category information is residue, filtering the decoration waste detection result according to the decoration waste category information to obtain the filtered decoration waste detection result includes: Calculate the area of the debris based on the location information of the renovation waste corresponding to the debris, and determine whether the calculated area exceeds a preset debris area threshold; If the calculated area exceeds a preset debris area threshold, reconfirming the decoration waste detection result; When the calculated area does not exceed a preset residue area threshold, the decoration waste detection result is output as a final decoration waste detection result.

2. The decoration waste detection method according to claim 1, characterized in that: Before inputting the image to be detected into a pre-trained decoration waste detection model, the method further includes: Scaling the image to be detected to a target size; In the case that the scaled image to be detected still does not meet the target size, padding processing is performed on the scaled image to be detected so that the size of the padded image to be detected meets the target size.

3. The decoration waste detection method according to claim 1, characterized in that: The pre-trained decoration waste detection model is trained in the following way: Obtain training data and corresponding label data for renovation waste scenarios, and build a renovation waste detection model; Inputting the training data into a vehicle discrimination sub-model to predict a vehicle discrimination result during training, and performing a loss calculation based on the vehicle discrimination result during training and the label data to obtain a first loss; Inputting the training data into the decoration waste detection sub-model to predict the decoration waste detection results during training, and performing loss calculation based on the decoration waste detection results during training and the label data to obtain a second loss; The decoration waste detection model is trained according to the first loss and the second loss to obtain a trained decoration waste detection model.

4. The decoration waste detection method according to claim 3, characterized in that: After obtaining the training data and corresponding label data in the renovation waste scenario, the method further includes: Extracting bounding box coordinates from the training data according to the label data; Initialize multiple anchor boxes to obtain the initial anchor box, and obtain the anchor box coordinates corresponding to the initial anchor box; Calculate the IoU between the initial anchor box and the bounding box according to the bounding box coordinates and the anchor box coordinates. If the IoU value exceeds the threshold, the corresponding initial anchor box is used as the target anchor box. Accordingly, inputting the training data into the decoration waste detection sub-model includes: The training data and the target anchor box are input into the decoration waste detection sub-model.

5. The decoration waste detection method according to claim 3, characterized in that: The decoration waste detection sub-model is constructed based on a predefined target detection algorithm and is a pre-trained model.

6. A decoration waste detection device, characterized in that: include: An image acquisition module is used to acquire images to be detected, wherein the images to be detected are obtained by photographing vehicles entering and leaving the garbage dump station by a terminal installed at the garbage dump station; An image input module, configured to input the image to be detected into a pre-trained decoration waste detection model, wherein the pre-trained decoration waste detection model includes a vehicle discrimination sub-model and a decoration waste detection sub-model; A vehicle identification module, configured to use the vehicle identification sub-model to identify the image to be detected and obtain a vehicle identification result; A decoration waste detection module is configured to input the image to be detected into a decoration waste detection sub-model to obtain a decoration waste detection result when a vehicle identification result indicates that the image to be detected includes a vehicle; The decoration waste detection result includes decoration waste category information and decoration waste location information; Accordingly, after obtaining the decoration waste detection result, the method further includes: Filtering the decoration waste detection results according to the decoration waste category information to obtain filtered decoration waste detection results; The renovation waste category information includes at least one or more of bulky waste, hard lumps, residues or wood blocks; Accordingly, when the decoration waste category information is residue, filtering the decoration waste detection result according to the decoration waste category information to obtain the filtered decoration waste detection result includes: Calculate the area of the debris based on the location information of the renovation waste corresponding to the debris, and determine whether the calculated area exceeds a preset debris area threshold; If the calculated area exceeds a preset debris area threshold, reconfirming the decoration waste detection result; When the calculated area does not exceed a preset residue area threshold, the decoration waste detection result is output as a final decoration waste detection result.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the decoration waste detection method according to any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the decoration waste detection method according to any one of claims 1 to 5 is implemented.

Images