A radar image object detection method, device, equipment and storage medium
By acquiring depth information contribution values from radar images and introducing an attention mechanism, combined with Mask R-CNN model training, the problem of low object detection accuracy in radar images is solved, achieving higher detection accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG POWER GRID CO LTD
- Filing Date
- 2023-02-17
- Publication Date
- 2026-06-02
AI Technical Summary
Object detection accuracy in radar images is low, and conventional image detection methods cannot be effectively applied due to the lack of rotation semantic invariance.
By acquiring initial detection contribution values from different depth information in radar sample images, an attention mechanism is introduced, and the Mask R-CNN model is used for training. The model parameters are then adjusted to improve detection accuracy.
It improves the accuracy of object detection in radar images, especially after the introduction of the attention mechanism, which retains more original feature information and further enhances detection precision.
Smart Images
Figure CN116129243B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, device, and storage medium for object detection in radar images. Background Technology
[0002] In engineering applications, radar imagery provides reliable information for applications. Accurate detection and identification of objects in radar images allows for better engineering design and maintenance.
[0003] Because the underground medium is not homogeneous and contains debris, useful information in radar images is often hidden in the noisy background, resulting in low object detection accuracy. Furthermore, radar images differ from conventional images. For example, radar images do not possess rotational semantic invariance like conventional images. Therefore, directly applying detection methods used for conventional images to radar images will also lead to low object detection accuracy. Summary of the Invention
[0004] This invention provides a method, apparatus, device, and storage medium for object detection in radar images, in order to improve the accuracy of object detection in radar images.
[0005] According to one aspect of the present invention, a method for object detection in radar images is provided, the method comprising:
[0006] Acquire radar sample images and determine the initial detection contribution value of different depth information in the radar sample images to the object to be detected;
[0007] The radar sample image is input into the deep learning model to be trained, and an attention mechanism is introduced into the deep learning model to be trained according to the initial detection contribution value to perform object detection on the radar sample image.
[0008] Based on the loss function of the object detection of the deep learning model to be trained, the model parameters of the deep learning model to be trained and the initial detection contribution value are adjusted to obtain the target deep learning model and the target detection contribution value.
[0009] The target deep learning model and the target detection contribution value are used to perform object detection on the target radar image.
[0010] Optionally, determining the initial detection contribution value of different depth information in the radar sample image to the object to be detected includes:
[0011] The target detection depth range of the radar sample image is determined based on the depth range of the object to be detected in the radar sample image.
[0012] In the radar sample image, the initial detection contribution value of depth information outside the target detection depth range is determined to be zero.
[0013] Optionally, an attention mechanism is introduced into the deep learning model to be trained based on the initial detection contribution value, including:
[0014] In the deep learning model to be trained, the radar sample image undergoes its first convolution operation to obtain the first convolution result;
[0015] The first convolution result is weighted according to the initial detection contribution value, and the weighted result is used to replace the first convolution result for subsequent calculations, so as to introduce an attention mechanism into the deep learning model to be trained.
[0016] Optionally, adjusting the model parameters of the deep learning model to be trained and the initial detection contribution value according to the loss function of the object detection of the deep learning model to be trained includes:
[0017] Based on the loss function of the deep learning model to be trained, which includes classification loss, anchor box loss, and cross-entropy loss in object detection, the model parameters of the deep learning model to be trained and the initial detection contribution value are adjusted.
[0018] Optionally, the deep learning model to be trained includes: a deep learning model with the Mask R-CNN model as the backbone network module.
[0019] Optionally, object detection is performed on the target radar image using the target deep learning model and the target detection contribution value, including:
[0020] The target radar image is input into the target deep learning model, and after the first convolution operation of the target radar image, the second convolution result is obtained;
[0021] The second convolution result is weighted according to the target detection contribution value, and the weighted result is used to replace the second convolution result for subsequent calculations to obtain the object detection result of the target radar image.
[0022] According to another aspect of the present invention, an object detection apparatus for radar images is provided, the apparatus comprising:
[0023] The initial detection contribution value determination module is used to acquire radar sample images and determine the initial detection contribution values of different depth information in the radar sample images to the object to be detected.
[0024] An attention mechanism introduction module is used to input the radar sample image into the deep learning model to be trained, and to introduce an attention mechanism into the deep learning model to be trained according to the initial detection contribution value to perform object detection on the radar sample image;
[0025] The parameter adjustment module is used to adjust the model parameters of the deep learning model to be trained and the initial detection contribution value according to the loss function of the object detection of the deep learning model to be trained, so as to obtain the target deep learning model and the target detection contribution value.
[0026] The object detection module is used to perform object detection on the target radar image using the target deep learning model and the target detection contribution value.
[0027] Optionally, the initial detection contribution value determination module includes:
[0028] The target detection depth range determination unit is used to determine the target detection depth range of the radar sample image based on the depth range of the object to be detected in the radar sample image.
[0029] The initial detection contribution value determination unit is used to determine, in the radar sample image, that the initial detection contribution value of depth information outside the target detection depth range is zero for the object to be detected.
[0030] Optional attention mechanism introduction modules include:
[0031] The first convolution result determination unit is used to obtain the first convolution result after the first convolution operation of the radar sample image in the deep learning model to be trained.
[0032] An attention mechanism introduction unit is used to weight the first convolution result according to the initial detection contribution value, and replace the first convolution result with the weighted result for subsequent calculations, so as to introduce an attention mechanism into the deep learning model to be trained.
[0033] Optional parameter adjustment module, including:
[0034] The parameter adjustment unit is used to adjust the model parameters of the deep learning model to be trained and the initial detection contribution value according to the loss function of the object detection of the deep learning model to be trained, which includes classification loss, anchor box loss and cross-entropy loss.
[0035] Optionally, the deep learning model to be trained includes: a deep learning model with the Mask R-CNN model as the backbone network module.
[0036] Optional, the object detection module includes:
[0037] The second convolution result determination unit is used to input the target radar image into the target deep learning model, and after the first convolution operation of the target radar image, obtain the second convolution result;
[0038] The object detection result determination unit is used to perform weighted processing on the second convolution result according to the target detection contribution value, and replace the second convolution result with the weighted processing result for subsequent calculations to obtain the object detection result of the target radar image.
[0039] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:
[0040] At least one processor; and
[0041] A memory communicatively connected to the at least one processor; wherein,
[0042] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the object detection method for radar images according to any embodiment of the present invention.
[0043] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the object detection method for radar images according to any embodiment of the present invention.
[0044] The technical solution of this invention involves acquiring radar sample images and determining the initial detection contribution values of different depth information in the radar sample images to the objects to be detected; inputting the radar sample images into a deep learning model to be trained, and introducing an attention mechanism into the deep learning model to be trained based on the initial detection contribution values to perform object detection on the radar sample images; adjusting the model parameters and initial detection contribution values of the deep learning model to be trained based on the loss function of object detection in the deep learning model to be trained, thereby obtaining the target deep learning model and the target detection contribution values; and using the target deep learning model and the target detection contribution values to perform object detection on the target radar image, thus solving the problem of object detection in radar images. By introducing an attention mechanism based on the detection contribution values for different depth information in the radar image, the accuracy of object detection in radar images can be improved.
[0045] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0047] Figure 1 This is a flowchart of a radar image object detection method according to Embodiment 1 of the present invention;
[0048] Figure 2 This is a flowchart of an object detection method for radar images provided according to Embodiment 2 of the present invention;
[0049] Figure 3 This is a schematic diagram of a deep learning model structure to be trained that incorporates an attention mechanism, according to Embodiment 2 of the present invention.
[0050] Figure 4 This is a schematic diagram of a radar sample image provided according to Embodiment 2 of the present invention;
[0051] Figure 5 This is a schematic diagram of the structure of a radar image object detection device according to Embodiment 3 of the present invention;
[0052] Figure 6 This is a schematic diagram of the structure of an electronic device that implements the object detection method for radar images according to embodiments of the present invention. Detailed Implementation
[0053] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0054] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0055] Example 1
[0056] Figure 1 This is a flowchart of an object detection method for radar images according to Embodiment 1 of the present invention. This embodiment is applicable to situations where object detection is performed in radar images. The method can be executed by a radar image object detection device, which can be implemented in hardware and / or software and can be configured in an electronic device such as a computer. Figure 1 As shown, the method includes:
[0057] Step 110: Acquire radar sample images and determine the initial detection contribution value of different depth information in the radar sample images to the object to be detected.
[0058] In this embodiment, the radar sample image can be a sagittal and transverse section imaging (B-scan) radar image. The abscissa of a B-scan radar image is generally the horizontal distance, and the ordinate is the propagation time of the radar signal in the medium. The ordinate can reflect the depth of the detected object. The physical meanings of the abscissa and ordinate of a B-scan radar image are different. B-scan radar images differ from conventional images. B-scan radar images do not possess semantic invariance after rotation. The importance of the signal varies at different ordinates (i.e., depths) in a B-scan radar image.
[0059] For example, an n-dimensional vector can be given to represent the initial detection contribution values of different depth information in the radar sample image to the object to be detected. Here, n can be the number of rows in the radar sample image. The sum of the n initial detection contribution values is 1. Each initial detection contribution value can be any value, and their sum is 1; or, they can be the same value, i.e., 1 / n.
[0060] Step 120: Input the radar sample images into the deep learning model to be trained, and introduce an attention mechanism into the deep learning model to be trained based on the initial detection contribution value to perform object detection on the radar sample images.
[0061] Introducing an attention mechanism into the deep learning model to be trained based on the initial detection contribution value can enhance the attention to different depth information in the radar sample image by using the initial detection contribution value. Specifically, introducing an attention mechanism into the deep learning model to be trained can be achieved by using an n-dimensional vector to weight the feature information at different depths in the radar sample image. Using the deep learning model to be trained to perform object detection on the radar sample image can be achieved by iteratively training the deep learning model to obtain the target deep learning model.
[0062] Step 130: Based on the loss function of the object detection of the deep learning model to be trained, adjust the model parameters and initial detection contribution value of the deep learning model to be trained to obtain the target deep learning model and the object detection contribution value.
[0063] The loss function can include, but is not limited to, classification loss functions. The deep learning model to be trained is trained using the loss function, which can be achieved through iterative training to minimize the loss function. When the loss function is minimized, the corresponding model parameters and detection contribution values can be used as the model parameters and object detection contribution values of the target deep learning model, respectively.
[0064] Step 140: Use a target deep learning model and target detection contribution value to perform object detection on the target radar image.
[0065] Object detection in target radar images can be achieved by inputting the target radar image into a target deep learning model and weighting the target radar image according to the target detection contribution value, i.e., introducing an attention mechanism. The final result is the object detection result of the target radar image obtained by the target deep learning model with the attention mechanism.
[0066] The technical solution of this embodiment acquires radar sample images and determines the initial detection contribution values of different depth information in the radar sample images to be detected; inputs the radar sample images into a deep learning model to be trained, and introduces an attention mechanism into the deep learning model to be trained based on the initial detection contribution values to perform object detection on the radar sample images; adjusts the model parameters and initial detection contribution values of the deep learning model to be trained based on the loss function of object detection in the deep learning model to be trained, thereby obtaining the target deep learning model and the target detection contribution value; and uses the target deep learning model and the target detection contribution value to perform object detection on the target radar image, thus solving the problem of object detection in radar images. This method improves the accuracy of object detection in radar images by introducing an attention mechanism based on the detection contribution values of different depth information in the radar image.
[0067] Example 2
[0068] Figure 2 This is a flowchart of an object detection method for radar images according to Embodiment 2 of the present invention. This embodiment is a further refinement of the above technical solution, and the technical solution in this embodiment can be combined with various optional solutions in one or more of the above embodiments. Figure 2 As shown, the method includes:
[0069] Step 210: Acquire radar sample images and determine the initial detection contribution value of different depth information in the radar sample images to the object to be detected.
[0070] In an optional embodiment of the present invention, determining the initial detection contribution value of different depth information of the object to be detected in the radar sample image includes: determining the target detection depth range of the radar sample image based on the depth range of the object to be detected in the radar sample image; and determining that the initial detection contribution value of depth information outside the target detection depth range of the radar sample image is zero.
[0071] The uppermost part of a B-scan radar image represents the propagation of radar waves from the air to the ground surface; this part generally does not contain objects to be detected. Therefore, objects to be detected only exist within a certain depth range in the B-scan radar image. To improve the accuracy of object detection, objects can be detected within the target detection depth range of the radar sample image. Specifically, the initial detection contribution value corresponding to the non-target detection depth range of the radar sample image can be set to zero, thereby improving the accuracy of object detection.
[0072] Furthermore, in practical applications, before determining the initial detection contribution value of different depth information in the radar sample image to the object to be detected, the non-target detection depth range in the radar sample image where the object to be detected does not exist can be eliminated to reduce the workload of image detection.
[0073] Step 220: Input the radar sample image into the deep learning model to be trained, and obtain the first convolution result after the radar sample image is convolved for the first time in the deep learning model to be trained.
[0074] The deep learning models to be trained include: a deep learning model with Mask R-CNN as the backbone network module. Convolution operations can obtain image features of radar sample images, and attention mechanisms can be better introduced into deep learning models based on image features.
[0075] In this embodiment, an attention mechanism is introduced after the first convolution operation, which can retain more of the original feature information of the radar image. Introducing the attention mechanism on the basis of the original feature information can improve the accuracy of object detection.
[0076] Step 230: Weight the first convolution result according to the initial detection contribution value, and replace the first convolution result with the weighted result for subsequent calculations, so as to introduce an attention mechanism into the deep learning model to be trained and perform object detection on the radar sample image.
[0077] The first convolution result can be an n×m matrix A, and the initial detection contribution values can form an n-dimensional column vector w. The weighted result of the first convolution result, weighted according to the initial detection contribution values, can be B = w·A. By replacing A with B for subsequent calculations, different depth information in the radar image can be introduced into the deep learning model with different levels of importance, thereby improving the performance of target detection in radar images.
[0078] Figure 3 This is a schematic diagram of a deep learning model structure to be trained that incorporates an attention mechanism, according to Embodiment 2 of the present invention. Figure 3 As shown, in this embodiment, an attention mechanism can be introduced into the Mask R-CNN model. Specifically, the attention mechanism can be introduced after the first convolution operation in the Mask R-CNN model.
[0079] Step 240: Based on the loss functions of the deep learning model to be trained, which include classification loss, anchor box loss, and cross-entropy loss in object detection, adjust the model parameters and initial detection contribution value of the deep learning model to be trained to obtain the target deep learning model and the object detection contribution value.
[0080] In this embodiment, the loss function L can include three parts: classification loss, anchor box loss, and cross-entropy loss. Specifically, Figure 4 This is a schematic diagram of a radar sample image provided according to Embodiment 2 of the present invention. In this embodiment, the loss function L = Lcls + Lbox + Lmask. Wherein, Lcls represents the classification loss, Lbox represents the bounding-box loss, and Lmask represents the cross-entropy loss of the template (mask). The bounding-box can be... Figure 4 The smallest rectangle that can enclose a convex shape. The template can be... Figure 4 Convex line.
[0081] Step 250: Input the target radar image into the target deep learning model, and after the first convolution operation of the target radar image, obtain the second convolution result.
[0082] Step 260: Weight the second convolution result according to the target detection contribution value, and use the weighted result to replace the second convolution result for subsequent calculations to obtain the object detection result of the target radar image.
[0083] In this process, the deep learning model to be trained is obtained by training the target deep learning model and the target detection contribution value through radar sample images. This can improve the accuracy of object detection when performing object detection on the target radar image.
[0084] The technical solution of this embodiment involves acquiring radar sample images and determining the initial detection contribution values of different depth information in the radar sample images to the object to be detected; inputting the radar sample images into the deep learning model to be trained, and obtaining the first convolution result after the first convolution operation of the radar sample images in the deep learning model to be trained; weighting the first convolution result according to the initial detection contribution values, and replacing the first convolution result with the weighted result for subsequent calculations, thereby introducing an attention mechanism into the deep learning model to be trained to detect objects in the radar sample images; and adjusting the model parameters of the deep learning model to be trained based on the loss functions of classification loss, anchor box loss, and cross-entropy loss included in the object detection of the deep learning model to be trained. The initial detection contribution value is used to obtain the target deep learning model and the target detection contribution value. The target radar image is input into the target deep learning model, and after the first convolution operation of the target radar image, the second convolution result is obtained. The second convolution result is weighted according to the target detection contribution value, and the weighted result replaces the second convolution result for subsequent calculations to obtain the object detection result of the target radar image. This method solves the object detection problem in radar images. By introducing an attention mechanism through the detection contribution value of different depth information in the radar image, the accuracy of object detection in radar images can be improved. Furthermore, introducing an attention mechanism after the first convolution operation can retain more original features of the image, further improving the accuracy of object detection.
[0085] The acquisition, storage, and application of radar images involved in the technical solutions of this invention comply with relevant laws and regulations and do not violate public order and good morals.
[0086] Example 3
[0087] Figure 5 This is a schematic diagram of the structure of a radar image object detection device according to Embodiment 3 of the present invention. Figure 5 As shown, the device includes: an initial detection contribution value determination module 310, an attention mechanism introduction module 320, a parameter adjustment module 330, and an object detection module 340. Wherein:
[0088] The initial detection contribution value determination module 310 is used to acquire radar sample images and determine the initial detection contribution values of different depth information of the object to be detected in the radar sample images.
[0089] Attention mechanism introduction module 320 is used to input radar sample images into the deep learning model to be trained, and introduce an attention mechanism into the deep learning model to be trained based on the initial detection contribution value to perform object detection on the radar sample images.
[0090] The parameter adjustment module 330 is used to adjust the model parameters and initial detection contribution value of the deep learning model to be trained according to the loss function of object detection of the deep learning model to be trained, so as to obtain the target deep learning model and the target detection contribution value.
[0091] The object detection module 340 is used to perform object detection on the target radar image using a target deep learning model and target detection contribution value.
[0092] Optionally, the initial detection contribution value determination module 310 includes:
[0093] The target detection depth range determination unit is used to determine the target detection depth range of the radar sample image based on the depth range of the object to be detected in the radar sample image.
[0094] The initial detection contribution value determination unit is used to determine that the initial detection contribution value of depth information within the non-target detection depth range of the radar sample image is zero.
[0095] Optionally, the attention mechanism introduction module 320 includes:
[0096] The first convolution result determination unit is used to obtain the first convolution result after the first convolution operation of the radar sample image in the deep learning model to be trained.
[0097] An attention mechanism is introduced into the unit, which is used to weight the first convolution result based on the initial detection contribution value, and then replace the first convolution result with the weighted result for subsequent calculations, so as to introduce the attention mechanism into the deep learning model to be trained.
[0098] Optionally, the parameter adjustment module 330 includes:
[0099] The parameter tuning unit is used to adjust the model parameters and initial detection contribution value of the deep learning model to be trained, based on the loss function of object detection, which includes classification loss, anchor box loss, and cross-entropy loss.
[0100] Optionally, the deep learning model to be trained includes: a deep learning model with the Mask R-CNN model as the backbone network module.
[0101] Optionally, the object detection module 340 includes:
[0102] The second convolution result determination unit is used to input the target radar image into the target deep learning model and obtain the second convolution result after the first convolution operation of the target radar image;
[0103] The object detection result determination unit is used to weight the second convolution result according to the target detection contribution value, and then use the weighted result to replace the second convolution result for subsequent calculations to obtain the object detection result of the target radar image.
[0104] The object detection device for radar images provided in this embodiment of the invention can execute the object detection method for radar images provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the method.
[0105] Example 4
[0106] Figure 6 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0107] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0108] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0109] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as object detection methods for radar images.
[0110] In some embodiments, the object detection method for radar images may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the object detection method for radar images described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the object detection method for radar images by any other suitable means (e.g., by means of firmware).
[0111] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0112] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0113] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0114] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0115] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0116] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0117] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0118] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method for object detection in radar images, characterized in that, The method includes: Acquire radar sample images and determine the initial detection contribution values of different depth information in the radar sample images to the object to be detected; the initial detection contribution values are n-dimensional vectors, where n is the number of rows in the radar sample images; The radar sample image is input into the deep learning model to be trained, and an attention mechanism is introduced into the deep learning model to be trained according to the initial detection contribution value to perform object detection on the radar sample image. Based on the loss function of the object detection of the deep learning model to be trained, the model parameters of the deep learning model to be trained and the initial detection contribution value are adjusted to obtain the target deep learning model and the target detection contribution value. The target deep learning model and the target detection contribution value are used to perform object detection on the target radar image.
2. The method according to claim 1, characterized in that, Determining the initial detection contribution value of different depth information in the radar sample image to the object to be detected includes: The target detection depth range of the radar sample image is determined based on the depth range of the object to be detected in the radar sample image. In the radar sample image, the initial detection contribution value of depth information outside the target detection depth range is determined to be zero.
3. The method according to claim 1, characterized in that, An attention mechanism is introduced into the deep learning model to be trained based on the initial detection contribution value, including: In the deep learning model to be trained, the radar sample image undergoes its first convolution operation to obtain the first convolution result; The first convolution result is weighted according to the initial detection contribution value, and the weighted result is used to replace the first convolution result for subsequent calculations, so as to introduce an attention mechanism into the deep learning model to be trained.
4. The method according to claim 1, characterized in that, Based on the loss function of the object detection in the deep learning model to be trained, the model parameters of the deep learning model to be trained and the initial detection contribution value are adjusted, including: Based on the loss function of the deep learning model to be trained, which includes classification loss, anchor box loss, and cross-entropy loss in object detection, the model parameters of the deep learning model to be trained and the initial detection contribution value are adjusted.
5. The method according to claim 1, characterized in that, The deep learning model to be trained includes: a deep learning model with Mask R-CNN model as the backbone network module.
6. The method according to claim 3, characterized in that, Object detection is performed on the target radar image using the aforementioned deep learning model for the target and the aforementioned target detection contribution value, including: The target radar image is input into the target deep learning model, and after the first convolution operation of the target radar image, the second convolution result is obtained; The second convolution result is weighted according to the target detection contribution value, and the weighted result is used to replace the second convolution result for subsequent calculations to obtain the object detection result of the target radar image.
7. An object detection device for radar images, characterized in that, The device includes: The initial detection contribution value determination module is used to acquire radar sample images and determine the initial detection contribution values of different depth information of the object to be detected in the radar sample images; the initial detection contribution value is an n-dimensional vector, where n is the number of rows in the radar sample image; An attention mechanism introduction module is used to input the radar sample image into the deep learning model to be trained, and to introduce an attention mechanism into the deep learning model to be trained according to the initial detection contribution value to perform object detection on the radar sample image; The parameter adjustment module is used to adjust the model parameters of the deep learning model to be trained and the initial detection contribution value according to the loss function of the object detection of the deep learning model to be trained, so as to obtain the target deep learning model and the target detection contribution value. The object detection module is used to perform object detection on the target radar image using the target deep learning model and the target detection contribution value.
8. The apparatus according to claim 7, characterized in that, The initial detection contribution value determination module includes: The target detection depth range determination unit is used to determine the target detection depth range of the radar sample image based on the depth range of the object to be detected in the radar sample image. The initial detection contribution value determination unit is used to determine, in the radar sample image, that the initial detection contribution value of depth information outside the target detection depth range is zero for the object to be detected.
9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the object detection method for radar images according to any one of claims 1-6.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the object detection method for radar images according to any one of claims 1-6.