Sand shallow buried garbage detection method and system based on intelligent trolley and ir-uwb radar

Abstract

The application provides a sand shallow-buried garbage detection method and system based on an intelligent trolley and an IR-UWB radar, and comprises the following steps: S1: detecting sand by using an IR-UWB radar carried by an intelligent trolley, and collecting radar data that can be used for garbage detection and positioning; S2: pre-processing the collected radar data that can be used for garbage detection and positioning; S3: based on the pre-processed radar data, using an improved algorithm based on Canny edge detection to adaptively detect edge features corresponding to multiple pieces of garbage; S4: extracting garbage features in the corresponding area and constructing the garbage features into a feature vector; and S5: distinguishing garbage from noise based on the constructed feature vector.

Description

Technical Field

[0001] This invention relates to the fields of wireless sensing and ubiquitous computing technology, specifically to a method and system for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar, and more specifically to a method, system, device, and medium for detecting and locating shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar. Background Technology

[0002] Rapid and accurate waste detection is a crucial aspect of waste management. With the continuous advancement of industrialization, more and more waste generated by humans in production, consumption, daily life, and other activities is gradually entering the natural environment, causing irreversible damage over time. Sandy land pollution, such as beach pollution, is becoming increasingly serious. Marine pollution has long been a major component of global waste pollution. Simultaneously, much marine debris is washed ashore due to tides, typhoons, and other natural environmental factors, resulting in large amounts of trash remaining on beaches. Therefore, research and technologies related to sandy land waste detection are of significant practical importance.

[0003] To make garbage disposal more convenient and efficient, many garbage disposal tools have been developed to help humans. At the same time, with the rapid development of computer vision and deep learning, humans can easily identify and locate objects automatically with the help of cameras and computing devices, including all kinds of garbage.

[0004] However, computer vision-based litter detection in sandy areas still has certain limitations. Light, being a high-frequency electromagnetic wave, lacks penetrating power and cannot detect obscured objects. It is also easily affected by environmental factors, such as fog or darkness. Litter in sandy areas, like on beaches, is easily affected by other environmental factors, such as wave erosion, resulting in a significant portion of litter being shallowly buried in the sand, making computer vision-based methods ineffective in detecting this type of litter. Radar, as an early military-grade wireless device for object detection, possesses ranging characteristics and penetrating power, making it a potential candidate for litter detection. Meanwhile, radar technology, already mature in the military field, has been rapidly commercialized in recent years, especially millimeter-wave radar, which has been successfully commercialized and applied to the rapidly developing field of autonomous driving as a target detection method to compensate for the limitations of computer vision.

[0005] This invention aims to design a real-time, accurate, and efficient system for detecting and locating shallowly buried waste in sandy areas, meeting the need for efficient and convenient waste disposal. This invention proposes a method for detecting and locating shallowly buried waste in sandy areas based on a smart vehicle and IR-UWB radar. The smart vehicle's high mobility enables real-time detection, while the IR-UWB radar provides accurate waste detection and location. However, designing such a method faces several challenges. The first challenge is that the raw radar signal is insufficient to extract accurate relevant features for accurate waste detection and location, requiring preprocessing of the raw radar signal to improve the signal-to-noise ratio of the waste signal. The second challenge is that traditional Canny edge detection methods require manual threshold setting, which cannot meet real-time requirements. The third challenge is that waste and noise are easily confused and difficult to distinguish, necessitating the extraction of waste-related features and SVM classification. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar.

[0007] A method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar, according to the present invention, includes:

[0008] Step S1: Use the IR-UWB radar on the smart car to detect the sandy area and collect radar data that can be used for garbage detection and location.

[0009] Step S2: Preprocess the collected radar data that can be used for waste detection and location;

[0010] Step S3: Based on the preprocessed radar data, adaptively detect the edge features corresponding to multiple pieces of trash using an improved algorithm based on Canny edge detection;

[0011] Step S4: Extract the garbage features of the corresponding area and construct them as a feature vector;

[0012] Step S5: Distinguish between garbage and noise based on the constructed feature vectors.

[0013] Preferably, step S2 employs:

[0014] Step S2.1: Upsample the radar data using interpolation;

[0015] Step S2.2: Perform range normalization processing on the upsampled radar data;

[0016] Step S2.3: Perform mean filtering on the normalized radar data;

[0017] Step S2.4: Apply Gaussian filtering to the filtered radar data to remove Gaussian noise.

[0018] Preferably, step S3 employs the following methods:

[0019] Step S3.1: Perform Canny edge detection on the preprocessed radar data to obtain an edge set;

[0020] Step S3.2: Perform DBSCAN unsupervised clustering on the obtained edge set, clustering the edge point set into different clusters. Each cluster corresponds to the edge point set of a single hot spot on the heat map. Calculate the specific location of the corresponding hot spot region on the heat map based on the cluster. Calculate the area and average energy intensity of all hot spot regions, and find the maximum value of the average energy intensity and the area of ​​the hot spot region. If the maximum value of the average energy intensity of the hot spot region is greater than the threshold, or the maximum value of the area of ​​the hot spot region is greater than the threshold, then the hot spot corresponding to the detected edge is considered to be caused by garbage; otherwise, it is considered to be caused by noise.

[0021] Step S3.3: When the energy intensity and size of the hot spot region corresponding to the detected edge meet the set threshold, the hot spot region corresponding to the detected edge in the heat map is set to zero, and Canny edge detection is triggered repeatedly. The hot spot region corresponding to the newly detected edge is compared with the set threshold until the set threshold is not met, and the loop ends.

[0022] Preferably, step S4 employs the following methods:

[0023] Step S4.1: Extract the magnitude of the gradient corresponding to the hot spot region as a feature;

[0024] Step S4.2: Extract the side length of the heatmap region as a feature;

[0025] Step S4.3: Extract the kurtosis of the heatmap region as a feature;

[0026] Step S4.4: Extract the peak half-width of the heatmap region as a feature;

[0027] Step S4.5: Construct a feature vector by splicing the magnitude of the gradient corresponding to the extracted hot spot region, the side length of the heat map region, the kurtosis of the heat map region, and the peak half-width of the heat map region.

[0028] Preferably, step S5 involves inputting the constructed feature vector into a support vector machine to classify garbage and noise.

[0029] A sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar, according to the present invention, includes:

[0030] Module M1: Uses the IR-UWB radar on the intelligent vehicle to detect sandy areas and collect radar data that can be used for garbage detection and location;

[0031] Module M2: Preprocesses the collected radar data that can be used for waste detection and location;

[0032] Module M3: Based on preprocessed radar data, it adaptively detects edge features corresponding to multiple pieces of trash using an improved algorithm based on Canny edge detection;

[0033] Module M4: Extracts garbage features from the corresponding area and constructs them as feature vectors;

[0034] Module M5: Distinguishes between garbage and noise based on the constructed feature vectors.

[0035] Preferably, the module M2 adopts:

[0036] Module M2.1: Upsamples radar data through interpolation;

[0037] Module M2.2: Performs range normalization processing on the upsampled radar data;

[0038] Module M2.3: Performs mean filtering on the normalized radar data;

[0039] Module M2.4: Gaussian filtering is used to remove Gaussian noise from the filtered radar data.

[0040] Preferably, the module M3 adopts:

[0041] Module M3.1: Performs Canny edge detection on the preprocessed radar data to obtain an edge set;

[0042] Module M3.2: Performs DBSCAN unsupervised clustering on the obtained edge set, clustering the edge point set into different clusters. Each cluster corresponds to the edge point set of a single hot spot on the heatmap. Calculates the specific location of the corresponding hot spot region on the heatmap based on the cluster. Calculates the area and average energy intensity of all hot spot regions, and finds the maximum value of the average energy intensity and hot spot area. If the maximum value of the average energy intensity or the maximum value of the hot spot area is greater than a threshold, the hot spot corresponding to the detected edge is considered to be caused by garbage; otherwise, it is considered to be caused by noise.

[0043] Module M3.3: When the energy intensity and size of the hot spot region corresponding to the detected edge meet the set threshold, the hot spot region corresponding to the detected edge in the heat map is set to zero, and Canny edge detection is triggered repeatedly. The hot spot region corresponding to the newly detected edge is compared with the set threshold until the set threshold is not met, and the loop ends.

[0044] Preferably, the module M4 adopts:

[0045] Module M4.1: Extracts the magnitude of the gradient corresponding to the hot spot region as a feature;

[0046] Module M4.2: Extracts the side length of the heatmap region as a feature;

[0047] Module M4.3: Extracts kurtosis from heatmap regions as a feature;

[0048] Module M4.4: Extracts the peak half-width of a heatmap region as a feature;

[0049] Module M4.5: Constructs a feature vector by splicing the modulus of the gradient corresponding to the extracted hot spot region, the side length of the heat map region, the kurtosis of the heat map region, and the peak half-width of the heat map region.

[0050] Preferably, module M5 employs the following method: inputting the constructed feature vectors into a support vector machine to classify garbage and noise.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. Compared with existing camera-based garbage detection and location methods, this invention can detect garbage shallowly buried under sand.

[0053] 2. Compared with existing ground-penetrating radar-based garbage detection and location methods, this invention designs the first mobile sandy shallow-buried garbage detection and location system based on intelligent vehicles and commercial IR-UWB radar, which has the advantages of lower cost and easier deployment.

[0054] 3. This invention achieves the technical effect of detecting garbage shallowly buried under sand by using IR-UWB radar and improved edge detection algorithm. By using commercial IR-UWB radar and a vehicle, it achieves the technical effect of a low-cost and easy-to-deploy mobile sandy shallow-buried garbage detection and positioning system.

[0055] 4. This invention, with the help of an intelligent vehicle and radar, can effectively and quickly detect shallowly buried garbage in sandy areas within a certain region, and has a certain degree of robustness against environmental disturbances.

[0056] 5. The present invention has a reasonable structure and is easy to use, and can overcome the defects of the prior art. Attached Figure Description

[0057] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0058] Figure 1 This is a flowchart of a method for detecting shallowly buried waste in sandy areas based on an intelligent vehicle and IR-UWB radar. Detailed Implementation

[0059] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0060] Example 1

[0061] This invention provides a method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar, such as... Figure 1 As shown, it includes:

[0062] Step S1: Data collection. The IR-UWB radar on the intelligent vehicle is used to detect the sandy area and collect radar data that can be used for garbage detection and location.

[0063] Step S2: Radar signal data preprocessing. A series of data preprocessing operations improve the signal-to-noise ratio of the signal corresponding to the garbage, which facilitates subsequent detection.

[0064] Step S3: An improved algorithm based on Canny edge detection is designed to adaptively detect edge features corresponding to multiple pieces of trash;

[0065] Step S4: Extract the garbage features of the corresponding area to facilitate further differentiation between garbage and noise;

[0066] Step S5: Extract the features to construct a bit feature vector and input it into the SVM for classification to distinguish between garbage and noise.

[0067] Specifically, step S1 includes:

[0068] Step S1.1: To select a suitable radar for sandy waste detection and location on a smart vehicle, common small and lightweight commercial radars were investigated, and their sand penetration capabilities were tested. Specifically, three common commercial radars with center operating frequencies of 77GHz, 24GHz, and 7.29GHz were tested, and preliminary shallow-buried waste experiments were conducted in sandy areas. Based on the experiments, the 7.29GHz IR-UWB radar, which has better penetration capabilities, was selected.

[0069] Step S1.2: Control the car to move along the given direction via Bluetooth communication, and at the same time use a computing device (laptop) to control the radar to scan a certain area of ​​sand in front of the radar. Based on the obtained radar heat map, further processing is performed to detect and locate the buried garbage in the scanned area.

[0070] Specifically, step S2 includes:

[0071] Step S2.1: The signal collected by the IR-UWB radar is essentially a discrete sampled signal point of a continuously received signal. Features are extracted from the Time-of-Flight (ToF) and Peak Amplitude (PAA) of the received radar signal. The accuracy of the ToF of the discrete sampled points determines the accuracy of the extracted features. To obtain a more accurate ToF and PPA of the received signal, it is proposed to upsample the received signal through interpolation.

[0072] Step S2.2: Removing the Influence of Distance. Since the amplitude of the signal received by an ultra-wideband pulse radar (UWBPR) is inversely proportional to the square of the distance, the signal reflected from the target attenuates significantly with the square of the distance, leading to a deterioration in the signal-to-noise ratio (SNR) and making the target difficult to detect. Furthermore, when multiple pieces of trash are present in the sand, the trash closer to the UWBPR experiences relatively weaker signal attenuation, while the trash farther away experiences stronger attenuation. This makes it difficult to detect trash further away from the UWBPR, reducing its detection range. To improve the SNR of the trash signal and increase the detection range of the UWBPR, distance normalization is performed on the received signal to remove the influence of distance.

[0073] Step S2.3: Clutter filtering is performed on the received signal. Mean filtering is used to filter out clutter reflected from the sand surface. That is, the average value of the ultra-wideband pulse radar received signal over a period of time is taken to estimate the background clutter. Then, the estimated background clutter is subtracted from the original received signal data to filter out background noise and obtain the target signal. Next, Gaussian filtering is used to filter out Gaussian noise and further smooth the received signal.

[0074] Step S3 includes selecting edges as features for initial waste detection. After data preprocessing, significant changes can be observed in the heatmap, with waste-corresponding areas exhibiting stronger energy intensity compared to other areas. A large gradient exists at the boundary between waste-corresponding areas and other areas, indicating clear edge information. Therefore, the edge features of the waste are utilized for waste detection.

[0075] This paper employs the Canny edge detection algorithm to perform edge detection on the grayscale image after data preprocessing and transformation, in order to extract edge features related to garbage, facilitating subsequent garbage detection and location. However, Canny edge detection has the problem that the threshold needs to be manually set. Therefore, an improved algorithm based on Canny edge detection is proposed.

[0076] Step S3.1: Initialization. Perform Canny edge detection on the data after the data preprocessing in step S2, setting the low threshold to 0.5 and the high threshold to 0.9. The purpose is to first detect the edges with the largest gradient in the heatmap and suppress noise interference as much as possible.

[0077] Step S3.2: Threshold Determination. First, DBSCAN unsupervised clustering is performed on the edge set obtained in Step S3.1, clustering the edge point set into different clusters. Each cluster corresponds to the edge point set of a single hot spot on the heatmap. Then, the specific location of the corresponding hot spot region on the heatmap can be calculated based on the cluster. The area and average energy intensity of all hot spot regions are calculated, and the maximum value of the average energy intensity and hot spot area is found. If the maximum value of the average energy intensity of the hot spot region is greater than the set threshold, or the maximum value of the hot spot area is greater than the set threshold, then the hot spot corresponding to the detected edge is considered to be highly likely to be caused by garbage; otherwise, it can be considered to be caused by noise.

[0078] Step S3.3: Loop through possible edges. If the energy intensity and size of the hotspot region corresponding to the edge detected by initialization meet the set threshold, then enter the loop. Set the hotspot region corresponding to the detected edge in the heatmap to zero, thus removing the largest edge from the heatmap. Then, the above steps can be repeated continuously to continue Canny edge detection, while the set high and low thresholds remain unchanged. Compare the hotspot region corresponding to the newly detected edge with the set threshold until the set threshold is not met, at which point the loop ends.

[0079] Step S4 includes: Through the previous algorithm, regions with obvious edge information can be identified. Although some noise has been filtered out by setting a threshold, some noise-corresponding regions still cannot be filtered out. In order to further distinguish between noise-corresponding regions and garbage-corresponding regions, corresponding features are further extracted to distinguish between garbage and noise.

[0080] Step S4.1: Extract the magnitude of the gradient corresponding to the hotspot region as a feature. The gradient is essentially a vector, the directional derivative of a function at a point, which reaches its maximum value in the gradient direction. Physically speaking, this means that at that point, the function changes the fastest along the gradient direction, with the largest rate of change. This maximum rate of change is the magnitude of the gradient.

[0081] Step S4.2: Extract the edge length of the heatmap region as a feature. The edge length (length and width) of the edges detected by edge detection is selected as the feature to distinguish between garbage and noise.

[0082] Step S4.3: Extract the kurtosis of the heatmap region as a feature. Kurtosis, also known as the kurtosis coefficient, is a feature number used to characterize the peak value of a probability density distribution curve at its mean. From a physical perspective, kurtosis directly reflects the sharpness of the peak. Kurtosis is a statistical measure compared to a normal distribution. Since noise is Gaussian noise, its kurtosis should be close to a normal distribution. The peaks corresponding to garbage are sharper than those of a normal distribution, and therefore can also be used as one of the features to distinguish noise from garbage.

[0083] Step S4.4: Extract the peak half-width of the heatmap region as a feature. The peak half-width is the width of the peak at half its height, and it can also reflect the information of garbage and noise in the time domain to a certain extent.

[0084] Step S4.5: Concatenate the above features to construct a feature vector.

[0085] Step S5 includes: using the classic classification algorithm Support Vector Machine (SVM) for classification. Support Vector Machine is a classic binary classification model. Its basic idea is to find a linear classifier in the feature space that maximizes the interval between data of different categories.

[0086] A sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar, according to the present invention, includes:

[0087] Module M1: Data collection. It uses the IR-UWB radar on the intelligent vehicle to detect sand and collect radar data that can be used for garbage detection and location.

[0088] Module M2: Radar signal data preprocessing. Through a series of data preprocessing operations, the signal-to-noise ratio of the garbage-related signal is improved, which facilitates subsequent detection.

[0089] Module M3: An improved algorithm based on Canny edge detection is designed to adaptively detect edge features corresponding to multiple garbage items.

[0090] Module M4: Extracts the garbage features of the corresponding area, which facilitates further differentiation between garbage and noise.

[0091] Module M5: Extracts features to construct a bit feature vector, which is then input into an SVM for classification, distinguishing between garbage and noise.

[0092] The module M1 includes:

[0093] Module M1.1: To select a suitable radar for sandy waste detection and location on a smart vehicle, commonly available small and lightweight commercial radars were investigated, and their sand penetration capabilities were tested. Specifically, three common commercial radars with center operating frequencies of 77GHz, 24GHz, and 7.29GHz were tested in shallow sandy waste burial experiments. Based on the experiments, the 7.29GHz IR-UWB radar, with its better penetration capability, was selected.

[0094] Module M1.2: Controls the car to move in a given direction via Bluetooth communication, while using a computing device (laptop) to control the radar to scan a certain area of ​​sand in front of the radar. Based on the obtained radar heat map, further processing is performed to detect and locate the buried garbage in the scanned area.

[0095] The module M2 includes:

[0096] Module M2.1: The signal collected from the IR-UWB radar is essentially a discrete sampled signal point of a continuously received signal. Since features need to be extracted from the Time-of-Flight (ToF) and Peak Amplitude (PA) of the received radar signal, the ToF accuracy of the discrete sampled points determines the accuracy of the extracted features. To obtain more accurate ToF and PA of the received signal, an upsampling method using interpolation is proposed.

[0097] Module M2.3: Removing the Influence of Distance. Since the amplitude of the signal received by an ultra-wideband pulse radar (UWBPR) is inversely proportional to the square of the distance, the signal reflected from the target attenuates significantly with the square of the distance, leading to a deterioration in the signal-to-noise ratio (SNR) and making the target difficult to detect. Furthermore, when multiple pieces of trash are present in the sand, the trash closer to the UWBPR experiences relatively weaker signal attenuation, while the trash farther away experiences stronger attenuation, making the distant trash difficult to detect and reducing the detection range of the UWBPR. To improve the SNR of the trash signal and increase the detection range of the UWBPR, distance normalization is performed on the received signal to remove the influence of distance.

[0098] Module M2.4: Performs clutter filtering on the received signal. It uses mean filtering to filter out clutter reflected from the sandy surface. Specifically, it uses the average value of the ultra-wideband pulse radar received signal collected over a period of time to estimate the background clutter. Then, it uses the original received signal data to subtract the estimated background clutter to filter out background noise and obtain the target signal. Next, it uses Gaussian filtering to remove Gaussian noise and further smooth the received signal.

[0099] The module M3 includes: selecting edges as features for initial waste detection. After data preprocessing, it can be observed that the heatmap has changed significantly, with the waste-corresponding area having a stronger energy intensity compared to other areas, and a large gradient at the boundary between the waste-corresponding area and other areas, indicating clear edge information. Therefore, the edge features of the waste can be used for waste detection.

[0100] This paper employs the Canny edge detection algorithm to perform edge detection on the grayscale image after data preprocessing and transformation, in order to extract edge features related to garbage, facilitating subsequent garbage detection and location. However, Canny edge detection has the problem that the threshold needs to be manually set. Therefore, an improved algorithm based on Canny edge detection is proposed.

[0101] Module M3.1: Initialize. Perform Canny edge detection on the data after the data preprocessing in step S2. Here, the low threshold is set to 0.5 and the high threshold is set to 0.9. The purpose is to first detect the edges with the largest gradient in the heat map and suppress noise interference as much as possible.

[0102] Module M3.2: Threshold Judgment. First, DBSCAN unsupervised clustering is performed on the edge set obtained in step S3.1, clustering the edge points into different clusters. Each cluster corresponds to the edge point set of a single hotspot on the heatmap. Then, the specific location of the corresponding hotspot region on the heatmap can be calculated based on the cluster. The area and average energy intensity of all hotspot regions are calculated, and the maximum value of the average energy intensity and area of ​​the hotspot region is found. If the maximum value of the average energy intensity or the maximum value of the area of ​​the hotspot region is greater than the set threshold, then the hotspot corresponding to the detected edge is considered to be highly likely caused by garbage; otherwise, it can be considered to be caused by noise.

[0103] Module M3.3: Looping to find possible edges. If the energy intensity and size of the hotspot region corresponding to the edge detected by initialization meet the set threshold, then enter the loop. Set the hotspot region corresponding to the detected edge in the heatmap to zero, thus removing the largest edge from the heatmap. Then, the above steps can be repeated continuously to continue Canny edge detection, while the set high and low thresholds remain unchanged. Compare the hotspot region corresponding to the newly detected edge with the set threshold until the set threshold is not met, at which point the loop ends.

[0104] The module M4 includes:

[0105] The algorithm above has identified regions with clear edge information. Although some noise has been filtered out by setting a threshold, some noise-corresponding regions still cannot be filtered out. In order to further distinguish between noise-corresponding regions and garbage-corresponding regions, corresponding features were extracted to differentiate between garbage and noise.

[0106] Module M4.1: Extract the magnitude of the gradient corresponding to the hotspot region as a feature. The gradient is essentially a vector; it's the directional derivative of a function at a point, reaching its maximum value along the gradient direction. Physically speaking, this means the function changes most rapidly and has the largest rate of change along the gradient direction at that point. This maximum rate of change is the magnitude of the gradient. Step S4.2: Extract the edge length of the heatmap region as a feature. The edge length (length and width) detected by edge detection is used as a feature to distinguish between trash and noise.

[0107] Module M4.3: Extracting kurtosis as a feature from heatmap regions. Kurtosis, also known as the kurtosis coefficient, is a feature used to characterize the peak value of a probability density distribution curve at its mean. Physically, kurtosis directly reflects the sharpness of the peak. Kurtosis is a statistical measure compared to a normal distribution. Since noise is Gaussian noise, its kurtosis should be close to a normal distribution. The peaks corresponding to garbage are sharper than those of a normal distribution, and therefore can also be used as one of the features to distinguish noise from garbage.

[0108] Module M4.4: Extracts the peak half-width as a feature from the heatmap region. Peak half-width is the width of the peak at half its height, and it can also reflect the information of garbage and noise in the time domain to some extent.

[0109] Step S4.5: Concatenate the above features to construct a feature vector.

[0110] The module M5 includes:

[0111] The classic classification algorithm Support Vector Machine (SVM) was used for classification. SVM is a classic binary classification model. Its basic idea is to find a linear classifier in the feature space that maximizes the margin between data of different classes.

[0112] This invention provides a method, system, device, and medium for detecting and locating shallowly buried waste in sandy areas based on an intelligent vehicle. It employs a framework of data collection, feature extraction, edge detection, SVM classification, and waste detection and location. This framework accurately extracts relevant features from radar signals, eliminates the influence of ground clutter and Gaussian noise on radar signals, and precisely estimates the amount of different shallowly buried waste within a certain range of sandy areas.

[0113] Thirdly, a computer-readable storage medium is provided storing a computer program, which, when executed by a processor, implements the steps of the method for detecting and locating shallowly buried garbage in sandy areas using a drone.

[0114] Fourthly, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the UAV swarm intelligence perception and scheduling method.

[0115] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0116] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar, characterized in that, include: Step S1: Use the IR-UWB radar on the smart car to detect the sandy area and collect radar data that can be used for garbage detection and location. Step S2: Preprocess the collected radar data that can be used for waste detection and location; Step S3: Based on the preprocessed radar data, adaptively detect the edge features corresponding to multiple pieces of trash using an improved algorithm based on Canny edge detection; Step S4: Extract the garbage features of the corresponding area and construct them as a feature vector; Step S5: Distinguish between garbage and noise based on the constructed feature vectors; Step S3 employs the following: Step S3.1: Perform Canny edge detection on the preprocessed radar data to obtain an edge set; Step S3.2: Perform DBSCAN unsupervised clustering on the obtained edge set, clustering the edge point set into different clusters. Each cluster corresponds to the edge point set of a single hot spot on the heat map. Calculate the specific location of the corresponding hot spot region on the heat map based on the cluster. Calculate the area and average energy intensity of all hot spot regions, and find the maximum value of the average energy intensity and the area of ​​the hot spot region. If the maximum value of the average energy intensity of the hot spot region is greater than the threshold, or the maximum value of the area of ​​the hot spot region is greater than the threshold, then the hot spot corresponding to the detected edge is considered to be caused by garbage; otherwise, it is considered to be caused by noise. Step S3.3: When the energy intensity and size of the hot spot region corresponding to the detected edge meet the set threshold, the hot spot region corresponding to the detected edge in the heat map is set to zero, and Canny edge detection is triggered repeatedly. The hot spot region corresponding to the newly detected edge is compared with the set threshold until the set threshold is not met, and the loop ends.

2. The method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar according to claim 1, characterized in that, Step S2 employs the following: Step S2.1: Upsample the radar data using interpolation; Step S2.2: Perform range normalization processing on the upsampled radar data; Step S2.3: Perform mean filtering on the normalized radar data; Step S2.4: Apply Gaussian filtering to the filtered radar data to remove Gaussian noise.

3. The method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar according to claim 1, characterized in that, Step S4 employs the following: Step S4.1: Extract the magnitude of the gradient corresponding to the hot spot region as a feature; Step S4.2: Extract the side length of the heatmap region as a feature; Step S4.3: Extract the kurtosis of the heatmap region as a feature; Step S4.4: Extract the peak half-width of the heatmap region as a feature; Step S4.5: Construct a feature vector by splicing the magnitude of the gradient corresponding to the extracted hot spot region, the side length of the heat map region, the kurtosis of the heat map region, and the peak half-width of the heat map region.

4. The method for detecting shallowly buried garbage in sandy areas based on an intelligent vehicle and IR-UWB radar according to claim 1, characterized in that, Step S5 involves inputting the constructed feature vectors into a support vector machine to classify garbage and noise.

5. A sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar, characterized in that, include: Module M1: Uses the IR-UWB radar on the intelligent vehicle to detect sandy areas and collect radar data that can be used for garbage detection and location; Module M2: Preprocesses the collected radar data that can be used for waste detection and location; Module M3: Based on preprocessed radar data, it adaptively detects edge features corresponding to multiple pieces of trash using an improved algorithm based on Canny edge detection; Module M4: Extracts garbage features from the corresponding area and constructs them as feature vectors; Module M5: Distinguishes between garbage and noise based on the constructed feature vectors; The module M3 adopts: Module M3.1: Performs Canny edge detection on the preprocessed radar data to obtain an edge set; Module M3.2: Performs DBSCAN unsupervised clustering on the obtained edge set, clustering the edge point set into different clusters. Each cluster corresponds to the edge point set of a single hot spot on the heatmap. Calculates the specific location of the corresponding hot spot region on the heatmap based on the cluster. Calculates the area and average energy intensity of all hot spot regions, and finds the maximum value of the average energy intensity and hot spot area. If the maximum value of the average energy intensity or the maximum value of the hot spot area is greater than a threshold, the hot spot corresponding to the detected edge is considered to be caused by garbage; otherwise, it is considered to be caused by noise. Module M3.3: When the energy intensity and size of the hot spot region corresponding to the detected edge meet the set threshold, the hot spot region corresponding to the detected edge in the heat map is set to zero, and Canny edge detection is triggered repeatedly. The hot spot region corresponding to the newly detected edge is compared with the set threshold until the set threshold is not met, and the loop ends.

6. The sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar according to claim 5, characterized in that, The module M2 adopts: Module M2.1: Upsamples radar data through interpolation; Module M2.2: Performs range normalization processing on the upsampled radar data; Module M2.3: Performs mean filtering on the normalized radar data; Module M2.4: Gaussian filtering is used to remove Gaussian noise from the filtered radar data.

7. The sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar according to claim 5, characterized in that, The module M4 adopts: Module M4.1: Extracts the magnitude of the gradient corresponding to the hot spot region as a feature; Module M4.2: Extracts the side length of the heatmap region as a feature; Module M4.3: Extracts kurtosis from heatmap regions as a feature; Module M4.4: Extracts the peak half-width of a heatmap region as a feature; Module M4.5: Constructs a feature vector by splicing the modulus of the gradient corresponding to the extracted hot spot region, the side length of the heat map region, the kurtosis of the heat map region, and the peak half-width of the heat map region.

8. The sandy shallow-buried waste detection system based on an intelligent vehicle and IR-UWB radar according to claim 5, characterized in that, The module M5 employs the following method: inputting the constructed feature vectors into a support vector machine to classify garbage and noise.

Images