A garbage can overflow lightweight detection method, device and storage device

Abstract

The application discloses a garbage can overflow lightweight detection method, equipment and storage equipment, and the method comprises the steps of constructing a garbage can and a garbage sample data set; labeling the sample data set and dividing it into a training set, a validation set and a test set in proportion; constructing a garbage can overflow detection model, lightening the garbage can overflow detection model, combining an E-ELAN module to enhance the network learning ability, and obtaining a lightened and improved garbage can overflow detection model; training the lightened and improved garbage can overflow detection model to obtain a trained model, and deploying the model to an intelligent robot; the intelligent robot patrols a to-be-detected area to complete real-time detection; if a garbage can is detected, it is judged whether the current garbage can has garbage overflow according to a garbage overflow rule, if yes, a garbage overflow result is reported in real time; otherwise, no disposal is performed. The application has the beneficial effects that the real-time detection and the accuracy can be considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of target detection, and in particular to a garbage can overflow lightweight detection method, device and storage device. BACKGROUND

[0002] At present, garbage collection and transportation mainly relies on artificial periodic and fixed-point inspection and cleaning. Community garbage is related to geographical location, and it is easier to produce garbage in corners and remote locations, and it is related to the flow of people. The production location and total amount are extremely irregular. Therefore, in the development of modern society, the artificial cleaning method of garbage has a large workload and low efficiency. Therefore, to realize intelligent detection of garbage overflow, reduce labor costs and improve the efficiency of intelligent environmental protection is the solution to the current problem.

[0003] At present, the detection of garbage overflow is mainly processed by intelligent hardware and software. The use of intelligent hardware to detect garbage overflow mainly installs various types of sensors on the garbage can, which senses the height, temperature, humidity and odor of the garbage in the can, and when the sensor processing result reaches the set threshold, it triggers an automatic alarm to remind the relevant staff to handle the event. For example, Yihangtong Information Technology Co., Ltd. focuses on the construction of "intelligent environmental protection" information system platform, and designs a garbage overflow detection system combined with Internet of Things technology, installs a sensor in the center of the garbage can, sets an alarm threshold in advance, and when the garbage in the can reaches the threshold, triggers the alarm system to prompt that the garbage is full and needs to be handled in time. The TF series laser radar module of Beixing uses laser radar to automatically detect the amount of garbage, places the device on the lid of the garbage can, and when the amount of garbage reaches 80%, it is transmitted to the data management system to prompt garbage collection. The above detects garbage overflow through sensors, which is limited to fixed-point prompting and increases the cost of garbage cans. If the sensor is damaged, lost, etc., it needs to be replaced, which further increases the cost. The use of intelligent software to detect garbage overflow is mainly based on deep learning, which loads a mature garbage detection system in intelligent equipment, detects garbage overflow in real time through intelligent equipment inspection, and reports the overflow event.

[0004] At present, the detection of garbage overflow through deep learning algorithm can be loaded in intelligent equipment, but it is limited by the size of the model, and the existing industrial convolutional network model cannot balance real-time and accuracy.

[0005] Through comprehensive analysis, the current garbage overflow detection has the following problems:

[0006] (1) The installation of hardware to detect garbage overflow is limited by fixed-point prompting, which lacks flexibility, and the cost needs to be considered, which cannot achieve full coverage;

[0007] (2) The existing network model is not light enough, which is not conducive to deployment in equipment, and cannot balance the accuracy and real-time performance.

[0008] Therefore, an effective method and system for detecting garbage overflow should achieve high detection accuracy, low computational complexity, good robustness, and easy deployment. SUMMARY

[0009] In order to solve the problem that the detection accuracy, detection speed and robustness are difficult to balance in the existing garbage can overflow detection model, the improved convolutional network model YOLOv5 is used to detect garbage overflow. Since the YOLOv5 model is not light enough, it is not easy to deploy in intelligent equipment. Therefore, the YOLOv5 model is improved by combining the ideas of grouped convolution and efficient aggregation network, so as to realize lighter and higher accuracy real-time detection of garbage overflow.

[0010] Specifically, the application provides a garbage can overflow lightweight detection method, comprising the following steps:

[0011] Step S101, constructing a garbage can and a garbage sample data set;

[0012] Step S102, labeling the sample data set and dividing it into a training set, a validation set and a test set in proportion;

[0013] Step S103, constructing a garbage can overflow detection model, lightening the garbage can overflow detection model, and combining an E-ELAN module to enhance the network learning ability, to obtain a lightened and improved garbage can overflow detection model; in step S103, the lightened and improved garbage can overflow detection model comprises an input module, a Backbone module, a Neck module and a Head module; wherein the input module pre-processes the input image; the Backbone module is used to extract deep features of the image; the Neck module is used to improve the feature extraction capability; and the Head module is used to evaluate the lightened and improved garbage can overflow detection model;

[0014] Step S104, training, validating and detecting the lightened and improved garbage can overflow detection model by using the training set, the validation set and the test set;

[0015] Step S105, outputting the trained garbage can overflow detection model based on the predicted result, and deploying the model to an intelligent robot;

[0016] Step S106, patrolling the area to be detected by the intelligent robot, and using the trained garbage can overflow detection model for real-time detection;

[0017] Step S107, if the garbage can is detected, it is judged according to the established garbage overflow rule whether the current garbage can exists garbage overflow, if yes, the garbage overflow result is reported in real time, otherwise, it is not disposed.

[0018] A storage device stores instructions and data for implementing a garbage can overflow lightweight detection method.

[0019] A garbage can overflow lightweight detection device, comprising: a processor and the storage device; the processor loads and executes the instructions and data in the storage device to implement a garbage can overflow lightweight detection method.

[0020] The beneficial effects provided by the present application are:

[0021] (1) The present application has flexibility and sensitivity, is not limited by location and cost, and has strong practicability;

[0022] (2) The present application is more easily deployed in any intelligent equipment, and the improved garbage overflow detection model provided by the present application is more lightweight, the model calculation complexity is reduced, the detection speed and detection accuracy are better improved, real-time detection is easy, and more accurate information is provided. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is a method flowchart of the present application;

[0024] Figure 2 is a structure diagram of the lightweight improved garbage can overflow detection model in the present application;

[0025] Figure 3 is a hardware device working schematic diagram of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application clearer, the embodiments of the present application will be further described below with reference to the drawings.

[0027] Please refer to Figure 1 is a method flowchart of the present application.

[0028] The present application provides a garbage can overflow lightweight detection method, comprising the following steps:

[0029] Step S101, constructing a garbage can and a garbage sample data set;

[0030] The sample data set in the step S101 includes a certain number of garbage cans and garbage data, which are divided into positive samples and negative samples. The garbage cans and garbage data can be collected according to network pictures and collected in reality.

[0031] In this embodiment, according to network collection and real life collection, 1991 pictures are obtained, pictures containing garbage cans and garbage are taken as positive samples, and pictures not containing garbage cans and garbage are taken as negative samples to form a data set containing 1991 pictures.

[0032] In step S102, the sample data set is labeled and proportionally divided into a training set, a validation set and a test set.

[0033] The step S102 labels the sample data set using a picture labeling tool to generate a corresponding label file, and the label file includes a file format and a file content. The file format can be XML, and the file content is the category and position information of the target, wherein the position information of the target includes the left upper corner and right lower corner coordinate information of the real target. The labeled sample data set is divided into a training set, a validation set and a test set according to a preset proportion.

[0034] Here, the labeling tool can be but is not limited to a picture labeling tool, such as a wizard labeling assistant software; the preset proportion is a given value, such as 8:1:1.

[0035] In this embodiment, the collected garbage can and garbage pictures are labeled as garbage cans (TrashCan) and garbage (trash) respectively, and the wizard labeling assistant software is used to label the position and corresponding category of the garbage can or garbage in the image; the garbage can and garbage pictures and the corresponding label file are divided into the training set, the validation set and the test set according to 8:1:1.

[0036] In step S103, a garbage can overflow detection model is constructed, the garbage can overflow detection model is lightened, and the network learning ability is enhanced in combination with an E-ELAN module to obtain a lightened improved garbage can overflow detection model; in step S103, the lightened improved garbage can overflow detection model includes an input module, a Backbone module, a Neck module and a Head module; wherein the input module pre-processes the input image; the Backbone module is used to extract deep features of the image; the Neck module is used to improve the feature extraction capability; and the Head module is used to evaluate the lightened improved garbage can overflow detection model.

[0037] The lightened improved garbage can overflow detection model is built based on a deep learning framework, and the deep learning framework is not limited to a Pytorch framework.

[0038] In this embodiment, the Pytorch framework is used for building. Please refer to Figure 2 , Figure 2 is a structure diagram of the lightened improved garbage can overflow detection model in the present application.

[0039] It should be noted that the input module is used for pre-processing of the image; specifically, the input module scales each image according to a preset grid size of 640x640, and then normalizes it. At the same time, data enhancement processing is adopted, such as the Mosaic method, to improve the precision and training speed of the model.

[0040] It should be noted that the Backbone module replaces part of the C3 module and part of the Conv module in the traditional Backbone module with the grouping convolution IGCV3 and E-ELAN module during construction, forming a new sixteen-layer network structure of the Backbone module, as follows:

[0041] First layer: Conv module, 32 convolution kernels with a size of 6x6, step size of 2, to obtain data with a feature of 320x320x32;

[0042] Second layer: Conv module, 64 convolution kernels with a size of 3x3, step size of 2, to obtain data with a feature of 160x160x64;

[0043] Third layer: repeat 3 C3 modules, 64 convolution kernels with a size of 3x3, step size of 1, to obtain data with a feature of 160x160x64;

[0044] Fourth layer: Conv module, 128 convolution kernels with a size of 3x3, step size of 2, to obtain data with a feature of 80x80x128;

[0045] Fifth layer: repeat 6 C3 modules, 128 convolution kernels with a size of 3x3, step size of 1, to obtain data with a feature of 80x80x128;

[0046] Sixth layer: IGCV3 module, 128 convolution kernels, step size of 2, expansion coefficient of 2, to obtain data with a feature of 40x40x128;

[0047] Seventh layer: Conv module, 64 convolution kernels with a size of 1x1, step size of 1, to obtain data with a feature of 40x40x64;

[0048] Eighth layer: Conv module, 64 convolution kernels with a size of 1x1, step size of 1, to obtain data with a feature of 40x40x64;

[0049] Ninth layer: Conv module, 64 convolution kernels with a size of 3x3, step size of 1, to obtain data with a feature of 40x40x64;

[0050] Tenth layer: Conv module, 64 convolution kernels with a size of 3x3, step size of 1, to obtain data with a feature of 40x40x64;

[0051] Eleventh layer: Conv module, 64 3x3 convolution kernels, step 1, get data with feature 40x40x64;

[0052] Twelfth layer: Conv module, 64 3x3 convolution kernels, step 1, get data with feature 40x40x64;

[0053] Thirteenth layer: connect the outputs of the seventh, eighth, tenth and twelfth layers of the network structure of the new Backbone module, output data with feature 40x40x256

[0054] Fourteenth layer: Conv module, 256 1x1 convolution kernels, step 1, get data with feature 40x40x256;

[0055] Fifteenth layer: IGCV3 module, 512 convolution kernels, step 2, expansion coefficient 2, get data with feature 20x20x512;

[0056] Sixteenth layer: SPPF module, 3x3 pooling window, step 1, get data with feature 20x20x512.

[0057] Here, the SPPF module is a spatial pyramid pooling, which solves the problem of inconsistent input image size.

[0058] It can be understood that, in the lightweight improved garbage can overflow detection model, the Backbone module is used for feature extraction, and the original lightweight improved garbage can overflow detection model is composed of convolution Conv and C3 modules; the improved lightweight improved garbage can overflow detection model is more lightweight and has higher precision, specifically, the grouped convolution IGCV3 and E-ELAN modules are used to replace part of the modules in the Backbone module; the grouped convolution IGCV3 module is used for lightweight network results, and its structure is first 1x1 point convolution, which is divided into two groups for dimensionality expansion; then 3x3 deep grouped convolution is performed, which is divided into several groups for feature extraction; finally, 1x1 point convolution is performed, which is divided into two groups for dimensionality reduction. The E-ELAN module is used to improve the model performance and improve the network learning ability, and the result is divided into two branches, the first branch is changed in channel number through a 1x1 convolution, and the second branch is changed in channel number through a 1x1 convolution module, and then four 3x3 convolution modules are used for feature extraction.

[0059] In this way, the present application can replace part of the Backbone module in the lightweight improved garbage can overflow detection model with the grouped convolution IGCV3 and E-ELAN modules, which can lightweight the network while ensuring the accuracy, has fewer calculation parameter amounts, and thus improves the detection efficiency and sensitivity.

[0060] It should be noted that the Neck module. The grouping convolution IGCV3 is fused into the lightweight improved garbage can overflow detection model to construct the lighter garbage overflow detection model, including: replacing part of the C3 module in the Neck module with the grouping convolution IGCV3 module to constitute a new Neck module.

[0061] Here, the Neck module is followed by the Backbone module, that is, the final output of the Backbone module is taken as the input of the Neck module. The IGCV3 model is used to replace all C3 modules in the Neck module to constitute a fifteen-layer network structure of the new Neck module, as follows:

[0062] The first layer: Conv module, 256 convolution kernels with a size of 1*1 and a step of 1, to obtain data with a feature of 20*20*256;

[0063] The second layer: 2 times up-sampling, to obtain data with a feature of 40*40*256;

[0064] The third layer: connecting the second layer of the network structure of the new Neck module with the output of the fourteenth layer of the network structure of the new Backbone module, to output data with a feature of 40*40*512;

[0065] The fourth layer: IGCV3 module, 256 convolution kernels, a step of 1 and an expansion coefficient of 2, to obtain data with a feature of 40*40*256;

[0066] The fifth layer: Conv module, 128 convolution kernels with a size of 1*1 and a step of 1, and an expansion coefficient of 2, to obtain data with a feature of 40*40*128;

[0067] The sixth layer: 2 times up-sampling, to obtain data with a feature of 80*80*128;

[0068] The seventh layer: connecting the sixth layer of the network structure of the new Neck module with the output of the fifth layer of the network structure of the new Backbone module, to output data with a feature of 80*80*256;

[0069] The eighth layer: IGCV3 module, 128 convolution kernels, a step of 1 and an expansion coefficient of 2, to obtain data with a feature of 80*80*128;

[0070] The ninth layer: Conv module, 128 convolution kernels with a size of 3*3 and a step of 2, to obtain data with a feature of 40*40*128;

[0071] The tenth layer: the outputs of the ninth layer and the fifth layer of the network structure of the new Neck module are connected, and the output feature is 40*40*256 data;

[0072] The eleventh layer: IGCV3 module, 256 convolution kernels, step 1, expansion coefficient 2, and the feature is 40*40*256 data;

[0073] The twelfth layer: Cnov module, 256 convolution kernels with a size of 3*3, step 2, and the feature is 20*20*256 data;

[0074] The thirteenth layer: the outputs of the twelfth layer and the first layer of the network structure of the new Neck module are connected, and the output feature is 20*20*512 data;

[0075] The fourteenth layer: IGCV3 module, 512 convolution kernels, step 1, expansion coefficient 2, and the feature is 20*20*512 data;

[0076] The fifteenth layer: Detect module, and the features are 80*80, 40*40 and 20*20 data respectively, which are used for detecting targets of different sizes.

[0077] It can be understood that through multi-layer convolution feature extraction of the target, the feature size of the output of the Neck module is 80*80, 40*40 and 20*20 data respectively, and the grid feature data of different sizes are used for detecting targets of different sizes. Specifically, the 20*20 feature map predicts large-size targets, the 20*20 and 40*40 feature maps jointly predict medium-size targets, and the 20*20, 40*40 and 80*80 feature maps predict small-size targets.

[0078] It can be understood that the Neck module is used to improve the feature extraction capability, and in the lightweight improved garbage can overflow detection model, it includes SPPF structure and PANet structure; the grouping convolution IGCV3 module replaces all C3 modules in the Neck module of the lightweight improved garbage can overflow detection model. In this way, the present application can replace the C3 module in the Neck module of the lightweight improved garbage can overflow detection model with the grouping convolution IGCV3 module, reduce the calculation amount while maintaining the model accuracy, meet the real-time requirement, and be suitable for deployment in intelligent robots for real-time detection.

[0079] It should be noted that the Head module is used to measure the pros and cons of the garbage overflow detection model prediction result, and the module includes a classification loss function, a positioning loss function and a confidence loss function; here, the classification loss function is used to calculate the classification loss of the anchor frame and the target frame; the positioning loss function is used to calculate the error between the predicted frame and the target frame, and the mean average precision (mAP) of the garbage overflow detection model can be determined; the confidence loss function is used to calculate the confidence of the garbage overflow detection model.

[0080] Step S104, using the training set, the validation set and the test set, training, verifying and detecting the light-weight improved garbage can overflow detection model;

[0081] The step S104 includes: training the light-weight improved garbage overflow detection model using the training set; verifying the trained garbage overflow detection model using the validation set; and detecting the garbage overflow detection model using the test set. In this embodiment, the specific steps are as follows:

[0082] Step S104-1, training and verifying the model. Training can be set in advance for a number of rounds, and the light-weight improved garbage can overflow detection model is trained based on the training set. Generally, 300 rounds or 600 rounds are set, and convergence can be achieved. It can be understood that after each round of training and verification, the mAP of the current garbage overflow detection model is obtained, so that the final mAP of the light-weight improved garbage can overflow detection model is obtained after the training is completed, and the final mAP is the best mAP of the light-weight improved garbage can overflow detection model on the validation set.

[0083] Step S104-2, testing the model. The light-weight improved garbage can overflow detection model is used to predict the test set to be identified, and the prediction result on the test set is obtained. An accuracy threshold is set in advance, and if the accuracy is less than the preset accuracy threshold, the training is stopped and the light-weight improved garbage can overflow detection model is output. Here, the preset accuracy threshold is a given value, for example, 80%.

[0084] Step S105, outputting the trained garbage can overflow detection model based on the predicted result, and deploying the model to the intelligent robot;

[0085] The step S105 includes deploying the output trained garbage can overflow detection model to the intelligent robot. Given a prediction accuracy threshold, if the final prediction result reaches the threshold, the model can be used. The model is deployed to the intelligent robot to ensure that the model can run.

[0086] Step S106, the intelligent robot patrols the area to be detected, and uses the trained garbage can overflow detection model for real-time detection.

[0087] The step S106 comprises: setting the area to be detected in advance, and using the intelligent robot to patrol the area to be detected, and using the trained garbage can overflow detection model for real-time detection; the set area to be detected is a range, which can be a community or an intersection, etc.

[0088] In this embodiment, a user plans a community as the area to be detected, and uses the intelligent robot to patrol the community, and the model analyzes the video stream in real time and detects each frame in real time to obtain the garbage can and the garbage position.

[0089] Step S107, if the garbage can is detected, whether the garbage can exists garbage overflow is judged according to the formulated garbage overflow rule, if yes, the garbage overflow result is reported in real time; otherwise, no disposal.

[0090] The step S107 comprises formulating the garbage overflow rule. When the garbage can and the garbage position are detected, whether the garbage overflow exists is determined through the garbage overflow rule, and if yes, it is reported to the community management center to inform the relevant staff to handle in time.

[0091] In this embodiment, the garbage overflow rule can be that the garbage is detected in the limited area around the garbage can, that is, it is judged that the garbage is overflowed, wherein the height of the garbage can is h, the width is w, and the limited area is within the range of 1 / 2h above and below the garbage can and w to the left and right of the garbage can.

[0092] Please refer to Figure 3 , Figure 3 is a hardware device working schematic diagram of the embodiment of the application, and the hardware device specifically comprises: a garbage can overflow lightweight detection device 401, a processor 402 and a storage device 403.

[0093] The garbage can overflow lightweight detection device 401: the garbage can overflow lightweight detection device 401 realizes the garbage can overflow lightweight detection method.

[0094] The processor 402: the processor 402 loads and executes the instructions and data in the storage device 403 to realize the garbage can overflow lightweight detection method.

[0095] The storage device 403: the storage device 403 stores instructions and data; the storage device 403 is used to realize the garbage can overflow lightweight detection method.

[0096] The beneficial effects of the application are:

[0097] (1) The application has flexibility and sensitivity, is not limited by place and cost, and has strong practicability;

[0098] (2) The application is easier to deploy in any intelligent equipment, the improved garbage overflow detection model provided in the application is more lightweight, the detection speed and detection accuracy are better improved while reducing the model calculation complexity, real-time detection is easy, and more accurate information is provided.

[0099] The above merely describes preferred embodiments of the application and is not intended to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A lightweight method for detecting overflowing trash cans, characterized in that: The method comprises the following steps: Step S101, constructing a garbage can and a garbage sample data set; Step S102, labeling the sample data set and dividing it into a training set, a validation set and a test set in proportion; Step S103, constructing a garbage can overflow detection model, lightening the garbage can overflow detection model, and combining an E-ELAN module to enhance the network learning ability, to obtain a lightened and improved garbage can overflow detection model; In step S103, the lightened and improved garbage can overflow detection model comprises an input module, a Backbone module, a Neck module and a Head module; wherein the input module pre-processes the input image; the Backbone module is used for extracting deep image features; the Neck module is used for improving the feature extraction capability; and the Head module is used for evaluating the lightened and improved garbage can overflow detection model; Step S104, training, validating and detecting the lightened and improved garbage can overflow detection model by using the training set, the validation set and the test set; Step S105, outputting the trained lightened and improved garbage can overflow detection model based on the predicted result, and deploying the trained lightened and improved garbage can overflow detection model to an intelligent robot; Step S106, patrolling the detection area by the intelligent robot, and detecting in real time by using the trained lightened and improved garbage can overflow detection model; Step S107, if a garbage can is detected, judging whether the current garbage can has garbage overflow according to the formulated garbage overflow rule, and if yes, reporting the garbage overflow result in real time; otherwise, not disposing; In step S103, the garbage can overflow detection model adopts an improved YOLOv5 model; In step S103, the garbage can overflow detection model is lightened, specifically, grouping convolution and an E-ELAN module are fused into the YOLOv5 model to lighten it; The construction process of the Backbone module is as follows: a grouping convolution IGCV3 and an E-ELAN module are used to replace part of C3 modules and Conv modules in a traditional Backbone module, to obtain the Backbone module.

2. The method of claim 1, wherein: The structure of the Backbone module comprises: a first layer: a Conv module; a second layer: a Conv module; a third layer: three repeated C3 modules; a fourth layer: a Conv module; a fifth layer: six repeated C3 modules; a sixth layer: an IGCV3 module; a seventh layer: a Conv module; an eighth layer: a Conv module; a ninth layer: a Conv module; a tenth layer: a Conv module; an eleventh layer: a Conv module; a twelfth layer: a Conv module; a thirteenth layer: connecting the outputs of the seventh, eighth, tenth and twelfth layers; a fourteenth layer: a Conv module; a fifteenth layer: an IGCV3 module; and a sixteenth layer: an SPPF module.

3. The method of claim 1, wherein: The construction process of the Neck module is as follows: a grouping convolution IGCV3 module is used to replace part of C3 modules in a traditional Neck module, to constitute the Neck module.

4. The method of claim 3, wherein: The structure of the Neck module comprises: a first layer: a Conv module; a second layer: 2 times up-sampling; a third layer: connecting the second layer with the output of the fourteenth layer of the network structure of the Backbone module; a fourth layer: an IGCV3 module; a fifth layer: a Conv module; a sixth layer: 2 times up-sampling; a seventh layer: connecting the sixth layer with the output of the fifth layer of the network structure of the Backbone module; an eighth layer: an IGCV3 module; a ninth layer: a Conv module; a tenth layer: connecting the output of the ninth layer and the fifth layer; an eleventh layer: an IGCV3 module; a twelfth layer: a Cnov module; a thirteenth layer: connecting the output of the twelfth layer and the first layer; a fourteenth layer: an IGCV3 module; and a fifteenth layer: a Detect module.

5. The method of claim 1, wherein: In step S107, the garbage overflow rule is specifically: detecting garbage in a limited garbage can surrounding area, that is, judging as garbage overflow, wherein the garbage can height is h, the width is w, and the limited area is within 1 / 2h above and below the garbage can and w range to the left and right of the garbage can, wherein h and w are both preset values.

6. A storage device, characterized by: The storage device stores instructions and data for implementing any one of the garbage can overflow lightweight detection methods of claims 1-5.

7. A trash can overflow lightweight detection device, characterized in that: It comprises: A processor and a storage device; the processor loads and executes the instructions and data in the storage device to implement any one of the garbage can overflow lightweight detection methods of claims 1-5.

Images