Detection method, device and equipment of ghost probe target and medium

Abstract

The invention discloses a ghost probe target detection method, device and equipment and a medium, and the method comprises the steps: collecting first driving data in a vehicle driving process and image data on a front road, and determining second driving data and driving orientation of a pedestrian, and the position information of a target vehicle and the pedestrian; determining a pulse transmission frequency according to the first driving data; performing feature extraction on the image data to obtain image feature data, and inputting the image feature data into a pulse neural network for prediction according to a pulse transmission frequency to obtain a first prediction result; performing fuzzy clustering according to the first driving data, the second driving data, the driving orientation and the position information, and performing minimization processing on a target function corresponding to the membership degree to obtain a second prediction result corresponding to the ghost probe target; and fusing the first prediction result and the second prediction result to obtain a comprehensive prediction result corresponding to the ghost probe target. The method can improve the detection precision of the ghost probe target, and can be widely applied to the technical field of vehicles.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a method, apparatus, equipment and medium for detecting ghost-protruding targets. Background Technology

[0002] In recent years, with the rapid development of the social economy, traffic flow has increased rapidly, road transportation has become increasingly busy, and the probability of traffic accidents has also risen accordingly. Among them, "ghost peek" is a common traffic accident, which generally refers to a situation where the driver's vision is obstructed by surrounding vehicles or objects, creating a blind spot, and a moving object suddenly appears from the blind spot.

[0003] In related technologies, vehicles are generally equipped with intelligent driving systems that can intelligently detect unexpected pedestrians appearing out of nowhere, thereby proactively avoiding them or alerting the driver to take evasive action. However, this approach typically relies on information detected by the vehicle itself. Due to factors such as the sensing range of onboard sensors and target occlusion, the ability to identify unexpected pedestrians is relatively weak, and the identification timing is often late. This makes it difficult to slow down or change lanes in advance, reducing driving safety or causing sudden deceleration, which negatively impacts the driving experience.

[0004] In summary, the problems with the relevant technologies urgently need to be addressed. Summary of the Invention

[0005] The purpose of this application is to at least partially solve one of the technical problems existing in the related art.

[0006] Therefore, one object of the embodiments of this application is to provide a method, apparatus, device and medium for detecting ghost-protruding targets.

[0007] To achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of this application include:

[0008] On the one hand, embodiments of this application provide a method for detecting ghost-protruding targets, the method comprising:

[0009] The system collects first driving data and image data of the road ahead during the vehicle's journey, and uses the vehicle-road-cloud system to determine the second driving data of pedestrians, their driving direction, and the location information of the target vehicle and pedestrians.

[0010] Based on the first driving data, determine the pulse transmission frequency;

[0011] Image feature data is obtained by extracting features from the image data, and the image feature data is input into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result;

[0012] Fuzzy clustering is performed based on the first driving data, the second driving data, the driving direction, and the location information, and the objective function corresponding to the membership degree is minimized to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position;

[0013] The first prediction result and the second prediction result are fused to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0014] In addition, the ghost-peeping target detection method according to the above embodiments of this application may also have the following additional technical features:

[0015] Furthermore, in one embodiment of this application, determining the pulse transmission frequency based on the first driving data includes:

[0016] Based on the first driving data, the first driving speed and acceleration of the target vehicle are determined;

[0017] The pulse transmission frequency is determined based on the first driving speed and the acceleration.

[0018] Furthermore, in one embodiment of this application, the step of extracting image feature data from the image data includes:

[0019] The image data is input into a convolutional neural network;

[0020] The image feature data is obtained by extracting features from the image data using the convolutional neural network.

[0021] Furthermore, in one embodiment of this application, the step of inputting the image feature data into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result includes:

[0022] The image feature data is input into a spiking neural network for prediction based on the pulse transmission frequency, and initial prediction results are obtained at multiple time steps.

[0023] The initial prediction results are summarized to obtain the summarized prediction results;

[0024] The aggregated prediction result is corrected based on the first driving data to obtain the first prediction result.

[0025] Furthermore, in one embodiment of this application, fusing the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost-peeping target includes:

[0026] Determine the first confidence level corresponding to the spiking neural network and the second confidence level corresponding to the fuzzy clustering;

[0027] Based on the first confidence level and the second confidence level, determine the first weight corresponding to the first prediction result and the second weight corresponding to the second prediction result;

[0028] Based on the first weight and the second weight, the first prediction result and the second prediction result are weighted and fused to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0029] Furthermore, in one embodiment of this application, the method further includes:

[0030] If the comprehensive prediction results indicate the presence of a "ghost vehicle" (a vehicle that suddenly appears out of nowhere), an alarm message is sent to the driver of the target vehicle.

[0031] On the other hand, embodiments of this application provide a detection device for ghost-protruding targets, the device comprising:

[0032] The data acquisition unit is used to collect the first driving data and image data of the road ahead during the vehicle's driving process, and to determine the second driving data, driving direction, and location information of the target vehicle and pedestrian through the vehicle-road-cloud system.

[0033] The processing unit is used to determine the pulse transmission frequency based on the first driving data;

[0034] The first prediction unit is used to extract features from the image data to obtain image feature data, and input the image feature data into the pulse neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result.

[0035] The second prediction unit is used to perform fuzzy clustering based on the first driving data, the second driving data, the driving direction, and the location information, and to minimize the objective function corresponding to the membership degree to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position;

[0036] The fusion unit is used to fuse the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0037] On the other hand, embodiments of this application provide an electronic device, including:

[0038] At least one processor;

[0039] At least one memory for storing at least one program;

[0040] When the at least one program is executed by the at least one processor, the at least one processor implements the above-described method for detecting ghost targets.

[0041] On the other hand, embodiments of this application also provide a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the above-described method for detecting ghost-peeping targets.

[0042] On the other hand, embodiments of this application also provide a computer program product, which includes a computer program stored in a computer-readable storage medium. The processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the aforementioned method for detecting ghost-peeping targets.

[0043] The advantages and beneficial effects of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application:

[0044] This application discloses a method, apparatus, device, and medium for detecting "ghost peek-out" targets. It collects first driving data and image data from the road ahead during vehicle travel, and determines second driving data, driving direction, and position information of the target vehicle and pedestrian using a vehicle-road cloud system. Based on the first driving data, a pulse transmission frequency is determined. Image feature data is extracted from the image data and input into a spiking neural network according to the pulse transmission frequency to predict the "ghost peek-out" target, resulting in a first prediction result. Fuzzy clustering is performed based on the first driving data, the second driving data, the driving direction, and the position information, and the objective function corresponding to the membership degree is minimized to obtain a second prediction result corresponding to the "ghost peek-out" target. The cluster centers of the fuzzy clustering are the initial distance between the target vehicle and the pedestrian, and the distance between the collision point and the center point of the target vehicle when it reaches the collision point. The first and second prediction results are fused to obtain a comprehensive prediction result corresponding to the "ghost peek-out" target. This application utilizes a dual-modal fusion of spiking neural network (SNN) temporal prediction and adaptive fuzzy clustering to significantly improve the detection accuracy of pedestrians appearing out of nowhere, thereby enhancing the user's driving experience. SNN dynamically processes image features using pulse frequency to adapt to changes in vehicle speed; fuzzy clustering integrates multi-dimensional data such as pedestrian orientation, speed, and distance, quantifying collision risk by optimizing the membership function. The complementary dual prediction results effectively reduce the false positive rate and are suitable for complex road scenarios. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of this application or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions in this application. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0046] Figure 1 This is a schematic diagram of the implementation environment for a method for detecting ghost-protruding targets provided in this application embodiment;

[0047] Figure 2 This is a flowchart illustrating a method for detecting a ghost-protruding target provided in an embodiment of this application;

[0048] Figure 3 This is a schematic diagram of the structure of a ghost-protruding target detection device provided in the embodiments of this application;

[0049] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0050] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0051] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0053] In recent years, with the rapid development of the social economy, traffic flow has increased rapidly, road transportation has become increasingly busy, and the probability of traffic accidents has also risen accordingly. Among them, "ghost peek" is a common traffic accident, which generally refers to a situation where the driver's vision is obstructed by surrounding vehicles or objects, creating a blind spot, and a moving object suddenly appears from the blind spot.

[0054] In related technologies, vehicles are generally equipped with intelligent driving systems that can intelligently detect unexpected pedestrians appearing out of nowhere, thereby proactively avoiding them or alerting the driver to take evasive action. However, this approach typically relies on information detected by the vehicle itself. Due to factors such as the sensing range of onboard sensors and target occlusion, the ability to identify unexpected pedestrians is relatively weak, and the identification timing is often late. This makes it difficult to slow down or change lanes in advance, reducing driving safety or causing sudden deceleration, which negatively impacts the driving experience.

[0055] It's easy to understand that dealing with "ghost peeks" (suddenly appearing out of nowhere) requires decision-making based on comprehensive perception information across the entire scene. Simply optimizing the performance of a single vehicle's sensors and decision-making algorithms is insufficient to effectively handle such situations. Among related technologies, vehicle-road-cloud collaborative integrated technology, through the introduction and fusion of roadside perception, enhances the perception capabilities of individual vehicles, compensating for their shortcomings. However, in practical applications, while cloud platforms can solve computing power issues, their overall architecture suffers from excessively long latency chains, making real-time performance difficult to guarantee.

[0056] In view of this, this application provides a method for detecting "ghost peek-out" targets. The method involves collecting first driving data and image data from the road ahead during vehicle travel, and determining second driving data, driving direction, and position information of the target vehicle and pedestrian using a vehicle-road cloud system. Based on the first driving data, a pulse transmission frequency is determined. Feature extraction is performed on the image data to obtain image feature data, which is then input into a spiking neural network based on the pulse transmission frequency to predict the "ghost peek-out" target, resulting in a first prediction result. Fuzzy clustering is performed based on the first driving data, the second driving data, the driving direction, and the position information, and the objective function corresponding to the membership degree is minimized to obtain a second prediction result corresponding to the "ghost peek-out" target. The cluster centers of the fuzzy clustering are the initial distance between the target vehicle and the pedestrian, and the distance between the collision point and the center point of the target vehicle when it reaches the collision point. The first prediction result and the second prediction result are fused to obtain a comprehensive prediction result corresponding to the "ghost peek-out" target. This application utilizes a dual-modal fusion of spiking neural network (SNN) temporal prediction and adaptive fuzzy clustering to significantly improve the detection accuracy of pedestrians appearing out of nowhere, thereby enhancing the user's driving experience. SNN dynamically processes image features using pulse frequency to adapt to changes in vehicle speed; fuzzy clustering integrates multi-dimensional data such as pedestrian orientation, speed, and distance, quantifying collision risk by optimizing the membership function. The complementary dual prediction results effectively reduce the false positive rate and are suitable for complex road scenarios.

[0057] Please refer to Figure 1 , Figure 1 This diagram illustrates the implementation environment of a ghost-peeping target detection method provided in this embodiment. In this implementation environment, the main hardware and software components involved include a terminal device 110 and a backend server 120. The terminal device 110 and the backend server 120 are connected by communication.

[0058] Specifically, the ghost-peeping target detection method provided in this application embodiment can be executed independently on the terminal device 110, independently on the backend server 120, or based on data interaction between the terminal device 110 and the backend server 120. The terminal device 110 can be an in-vehicle terminal; the backend server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0059] The terminal device 110 and the backend server 120 can establish a communication connection via a wireless network or a wired network. This wireless or wired network uses standard communication technologies and / or protocols. The network can be the Internet or any other network, including but not limited to any combination of Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), mobile, wired or wireless networks, private networks, or virtual private networks.

[0060] Of course, this is understandable. Figure 1 The implementation environment described in this application is only one of the optional application scenarios for the ghost-peeping target detection method provided in this embodiment. The actual application is not fixed. Figure 1 The software and hardware environment shown.

[0061] Below, in conjunction with the aforementioned description of the implementation environment, a method for detecting ghost-protruding targets provided in the embodiments of this application will be introduced and explained.

[0062] Please refer to Figure 2 , Figure 2 This is a schematic diagram of a method for detecting a "ghost peek" target provided in an embodiment of this application. The method for detecting a "ghost peek" target includes, but is not limited to:

[0063] Step 210: Collect the first driving data and image data of the road ahead during the vehicle's driving process, and determine the second driving data, driving direction, and location information of the target vehicle and pedestrian through the vehicle-road cloud system;

[0064] Step 220: Determine the pulse transmission frequency based on the first driving data;

[0065] Step 230: Extract features from the image data to obtain image feature data, and input the image feature data into a spiking neural network according to the pulse transmission frequency to predict the ghost peeking target and obtain a first prediction result;

[0066] Step 240: Perform fuzzy clustering based on the first driving data, the second driving data, the driving direction, and the location information, and minimize the objective function corresponding to the membership degree to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position;

[0067] Step 250: Fuse the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0068] This application provides a method for detecting "ghost pedestrian" targets. This method significantly improves the detection accuracy of ghost pedestrian targets by fusing temporal prediction using a spiking neural network (SNN) and adaptive fuzzy clustering, thereby improving the user's driving experience. SNN dynamically processes image features using pulse frequency to adapt to changes in vehicle speed; fuzzy clustering integrates multi-dimensional data such as pedestrian orientation, speed, and distance, and quantifies collision risk by optimizing the membership function. The complementary dual prediction results effectively reduce the false positive rate and are suitable for complex road scenarios.

[0069] Specifically, in this embodiment, the system can collect first driving data (such as vehicle speed, acceleration, heading angle, etc.) and image data of the road ahead in real time. Simultaneously, it acquires second driving data (walking speed, gait), driving direction (face / torso direction), and real-time position information of the vehicle and pedestrian through a vehicle-road-cloud cooperative system. This data constitutes the basic input for dynamic perception and can subsequently be used by spiking neural networks and fuzzy clustering.

[0070] In this embodiment, the vehicle's driving state (such as acceleration or deceleration) can be determined based on the vehicle's first driving data, thereby dynamically adjusting the pulse transmission frequency. Specifically, in this embodiment, the first driving speed and acceleration of the target vehicle can be determined based on the first driving data, and then the pulse transmission frequency can be determined. For example, when the vehicle speed is high or the acceleration is large, the pulse transmission frequency can be increased to enhance the temporal parsing capability of the spiking neural network for image features; conversely, the frequency can be decreased to reduce computational energy consumption. This allows for adaptation to real-time changes in traffic scenarios, improving processing accuracy and saving computational resources.

[0071] The characteristics and complexity of "ghost peek" incidents pose significant challenges to existing target identification technologies. Spiking Neural Networks (SNNs), with their unique pulse coding mechanism and event-driven characteristics, can process spatiotemporal information more efficiently and exhibit stronger adaptability and real-time response capabilities to dynamically changing environments. In this embodiment, during image processing, convolutional networks can extract image feature data such as pedestrian contours and motion trajectories, and encode them into pulse sequences based on pulse transmission frequency, which are then input into the spiking neural network. The spiking neural network utilizes its event-driven characteristics to capture the temporal characteristics of pedestrian movement (such as sudden lateral movement) through a membrane potential accumulation-trigger mechanism, outputting a vision-based first prediction result. In this embodiment, the first prediction result can be the predicted probability of a "ghost peek" target appearing.

[0072] Specifically, in this embodiment, the image data collected by the roadside camera is first processed by a convolutional neural network (CNN) to extract key features from the image data, and then the extracted image feature data is encoded into a pulse sequence. This converts the position, speed, and orientation of vehicles and pedestrians in the image data into the pulse firing pattern of neurons, which is then input into the input layer of the spiking neural network.

[0073] After receiving a pulse sequence, the spiking neural network begins forward propagation computation. Input layer neurons transmit the pulse signals to hidden layer neurons, which process and extract features from the input signals based on predefined connection weights and a neuron model (such as the LIF model). Through continuous learning and weight adjustment, the neurons in the hidden layers gradually learn to recognize traffic scene feature patterns when a pedestrian suddenly appears. When dealing with a scenario where a pedestrian might suddenly appear, the hidden layer neurons can capture and integrate features such as changes in vehicle movement and environmental anomalies when a pedestrian suddenly appears. After processing through multiple hidden layers, the output layer neurons finally output the first prediction result based on the received signals: whether a pedestrian suddenly appears (this can be specific probability data).

[0074] In this embodiment, fuzzy clustering analysis is also initiated to predict the ghost-protruding target. Specifically, the concept of the fuzzy clustering algorithm is as follows:

[0075] In classical fuzzy set theory, membership is defined by the degree of membership of an element to each cluster center, and the membership function value of each element is restricted to the range [0, 1]. Mathematically, a fuzzy set F is defined as:

[0076] F={(x,μ F (x))|x∈E}(1)

[0077] Where, μ F (x) represents the membership value of element x to the universal set E. For each element x, μ F (x) are all between 0 and 1. Because fuzzy sets define the fuzziness of elements with respect to cluster centers, thus achieving soft classification of data, Bezdek et al. proposed the fuzzy clustering (FCM) algorithm based on fuzzy sets.

[0078] The traditional FCM algorithm is a method for constrained nonlinear programming problems. Its basic idea is as follows: For the current set E, firstly, c cluster centers are randomly determined; then, the membership degree of each element to the current cluster center is calculated based on the Euclidean distance between each element and the cluster center, and a membership matrix is ​​generated to update the cluster centers; when the iteration condition is met (generally, the iteration condition is set to: the change in cluster centers is less than a set value or the maximum number of iterations is reached), the iteration ends, maximizing inter-cluster differences and minimizing intra-cluster differences. Traditional FCM divides the entire set E into c fuzzy clusters. In each iteration, the cluster centers and membership values ​​are updated to minimize the objective function value. The objective function is used to classify the feature vectors by minimizing the objective function, which is expressed as:

[0079]

[0080] Formula (2) calculates J by updating the membership matrix U and the cluster center V. f The minimum value of x, where x k The elements k and c in the universal set are... i Representing cluster center i, by evaluating element x k With cluster center c i The membership relationship between them is used to calculate the member function value u. ik .in addition, The following formula gives u ik With c i The update method, where dis(x) k ,c i ) is element x k With cluster center c i Euclidean distance between them:

[0081]

[0082] FCM is widely used in image clustering and segmentation, but its application in intelligent decision-making is limited. When applied to ghost-protruding target decision-making, the original algorithm's properties and applicable working conditions are not compatible because the traditional FCM algorithm only considers distance information, which has a large error in vehicle-road-cloud systems and has very low practicality.

[0083] Traditional fuzzy clustering algorithms calculate membership degrees based solely on distance. Obviously, in the process of identifying pedestrian targets, traditional membership degree determination is ineffective. Therefore, this application's embodiments design an adaptive fusion method to calculate membership degrees by comprehensively considering the pedestrian target's facial orientation, walking speed, and distance to the estimated collision point.

[0084] The specific calculation formula is as follows:

[0085] v veh =Vcar sinφ,v person =V per sinθ(5)

[0086] P v =v veh t+Position veh ,P p =v person t+Position person (6)

[0087] Dis vp (t,c i )=||P v -P p || (7)

[0088] Among them, v veh It is the projection of the car's speed in the longitudinal direction of travel, v person It is the projection of the pedestrian's walking speed onto the lateral direction of the car's travel. Position represents the current position information of the vehicle / person in the vehicle-road cloud system. P represents the position information of the vehicle / person in the vehicle-road cloud system after a unit of time t. Dis vp This represents the distance between pedestrians and vehicles after a unit of time. α is the direction the pedestrian is facing. If the target is walking out of the road, the value is 10000 (to ensure minimal membership degree), otherwise it is 1.

[0089] In this embodiment of the application, the objective function of fuzzy clustering is:

[0090]

[0091] Where u ik is the membership degree between the current frame element k and the cluster center i, and N is the number of frames in the vehicle's motion after the pedestrian target is detected. The objective function of the algorithm in this application can be minimized using the Lagrange multiplier method. According to formula (4),

[0092]

[0093] In this application, two cluster centers are defined: c1 is the initial distance from the pedestrian target identified by the vehicle-road cloud perception, and c2 is the distance from the collision point to the vehicle's center point, representing the distance from the vehicle's current position to the collision point. By continuously updating the distance membership, the likelihood of a target suddenly appearing in the current scene can be assessed.

[0094] In this embodiment, the initial distance (c1) between the vehicle and the pedestrian and the estimated collision point distance (c2) are used as cluster centers. The adaptive membership degree is calculated by fusing the pedestrian's orientation, the velocity vector projections of both sides (v_x / v_y decomposition), and the dynamic position (Δd). Specifically, if the pedestrian is facing outward from the road, a very low membership degree is directly assigned to eliminate interference; otherwise, the membership degree distribution is optimized by minimizing the objective function (including Lagrange constraints) to generate a second prediction result based on the motion situation.

[0095] Finally, the system fuses the visual predictions of the spiking neural network with the kinematic predictions of fuzzy clustering to obtain a comprehensive prediction result for the ghost-peeping target.

[0096] It is understood that the method for detecting a "ghost peek" target provided in this application embodiment collects first driving data and image data of the road ahead during vehicle travel, and determines second driving data, driving direction, and position information of the target vehicle and pedestrian through a vehicle-road cloud system; determines a pulse transmission frequency based on the first driving data; extracts features from the image data to obtain image feature data, and inputs the image feature data into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target, obtaining a first prediction result; performs fuzzy clustering based on the first driving data, the second driving data, the driving direction, and the position information, and minimizes the objective function corresponding to the membership degree to obtain a second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point; and fuses the first prediction result and the second prediction result to obtain a comprehensive prediction result corresponding to the ghost peek target. This method significantly improves the detection accuracy of "ghost pedestrian" targets by fusing temporal prediction using a spiking neural network (SNN) and adaptive fuzzy clustering, thereby enhancing the user's driving experience. SNN dynamically processes image features using pulse frequency to adapt to changes in vehicle speed; fuzzy clustering integrates multi-dimensional data such as pedestrian orientation, speed, and distance, and quantifies collision risk by optimizing the membership function. The complementary dual prediction results effectively reduce the false positive rate and are suitable for complex road scenarios.

[0097] Specifically, in some embodiments, the step of inputting the image feature data into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result includes:

[0098] The image feature data is input into a spiking neural network for prediction based on the pulse transmission frequency, and initial prediction results are obtained at multiple time steps.

[0099] The initial prediction results are summarized to obtain the summarized prediction results;

[0100] The first prediction result is obtained by correcting the aggregated prediction result based on the first driving data.

[0101] In this embodiment, to improve the accuracy and reliability of the judgment, the output of the spiking neural network can be post-processed. Specifically, image feature data can be input into the spiking neural network according to the pulse transmission frequency to predict the ghosting target, obtaining initial prediction results at multiple time steps. Then, a voting mechanism can be used to statistically analyze the multiple judgment results. If the model determines the existence of the ghosting target multiple times in consecutive time steps, the existence of the ghosting target is more confidently confirmed. Therefore, a summary prediction result can be determined based on multiple initial prediction results. In this embodiment, other auxiliary information in the first driving data, such as vehicle braking signals and driver operation behavior, can also be combined to correct the judgment result, thereby obtaining the first prediction result.

[0102] Specifically, in some embodiments, fusing the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost peek target includes:

[0103] Determine the first confidence level corresponding to the spiking neural network and the second confidence level corresponding to the fuzzy clustering;

[0104] Based on the first confidence level and the second confidence level, determine the first weight corresponding to the first prediction result and the second weight corresponding to the second prediction result;

[0105] Based on the first weight and the second weight, the first prediction result and the second prediction result are weighted and fused to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0106] In this embodiment, the core logic of ensemble learning is to calculate the confidence scores of the judgment results of the spiking neural network and fuzzy clustering respectively through data training, where fuzzy clustering can be used as a rule-based safety net algorithm. This yields a first confidence score corresponding to the spiking neural network and a second confidence score corresponding to the fuzzy clustering. Furthermore, a first weight corresponding to the first prediction result and a second weight corresponding to the second prediction result can be determined. In the actual judgment process, the spiking neural network and the rule-based safety net algorithm run in parallel, simultaneously outputting the judgment results. The system performs a weighted fusion of the two results based on the pre-trained confidence score weights to determine the comprehensive prediction result corresponding to the "ghost peek" target. In some embodiments, if the comprehensive prediction result indicates the presence of a "ghost peek" target, a warning message can be sent to the driver of the target vehicle.

[0107] In this embodiment, when the first confidence level of the spiking neural network is higher than the second confidence level corresponding to the fuzzy clustering and reaches a certain threshold, the judgment result of the spiking neural network can be used first. If the confidence levels of the two are close or the confidence level of the fuzzy clustering is higher, the judgment logic of both is combined, and factors such as traffic scene and data credibility are comprehensively considered to arrive at the final judgment conclusion of the ghost peeping target. This ensemble learning mechanism can leverage the learning and adaptation capabilities of the spiking neural network to complex scenes, and can also use the fuzzy clustering algorithm to ensure the reliability of the judgment in known scenes, effectively improving the accuracy and stability of ghost peeping target judgment.

[0108] In this embodiment, the spiking neural network uses the aforementioned spiking neuron network structure to extract features and recognize patterns from the input data, outputting a preliminary result for judging ghost peeks. The rule-based baseline processing layer performs conditional matching on the input data based on preset traffic rules and fuzzy clustering results, outputting a rule-based judgment result. The confidence calculation layer calculates the confidence level of the outputs from both the spiking neural network and the rule-based baseline processing layer according to the aforementioned confidence evaluation method. The result fusion layer fuses the two judgment results using an ensemble strategy such as weighted averaging or voting based on the confidence calculation results. During the training phase, the ensemble learning architecture is trained using a large amount of historical data to optimize the spiking neural network parameters and the confidence weights of the two architectures, ensuring that the fused judgment result achieves optimal performance in various traffic scenarios.

[0109] The technical solution of this application has at least the following advantages:

[0110] 1. First, this application can serve as a target filtering module for "ghost pedestrians" (suddenly appearing from behind obstacles) in a cloud platform, providing accurate target filtering results for cloud-based decision-making and vehicle-side control. Traditional intelligent driving functions struggle to handle ghost pedestrian situations; this application's vehicle-road-cloud collaborative integrated system allows for reaction time in such situations. This application improves the real-time performance and accuracy of ghost pedestrian target determination through adaptive modifications to target acquisition using fuzzy clustering and the introduction of SNN.

[0111] 2. Secondly, to improve the flexibility of the current screening algorithm, this application integrates neural networks and machine learning, and fuses the results of two different fundamental models through confidence training. Because traditional DNN algorithms consume significant chip resources, their use in the ghost-peeping target judgment module can easily occupy the running space of other algorithms, affecting the overall performance of the cloud platform program. To address this issue, this application introduces a third-generation neural network (SNN) into the screening module. Its sparse computation and event-triggered characteristics require less computing power while still meeting the requirements of the ghost-peeping target judgment module.

[0112] Reference Figure 3This application also provides a device for detecting ghost-protruding targets, comprising:

[0113] The acquisition unit 310 is used to acquire first driving data and image data on the road ahead during the vehicle's driving process, and to determine the second driving data, driving direction, and location information of the target vehicle and pedestrian through the vehicle-road cloud system.

[0114] The processing unit 320 is used to determine the pulse transmission frequency based on the first driving data;

[0115] The first prediction unit 330 is used to extract features from the image data to obtain image feature data, and input the image feature data into the pulse neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result.

[0116] The second prediction unit 340 is used to perform fuzzy clustering based on the first driving data, the second driving data, the driving direction, and the location information, and to minimize the objective function corresponding to the membership degree to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position;

[0117] The fusion unit 350 is used to fuse the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost peek target.

[0118] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0119] Reference Figure 4 This application provides an electronic device, including:

[0120] At least one processor 410;

[0121] At least one memory 420 is used to store at least one program;

[0122] When at least one program is executed by at least one processor 410, the at least one processor 410 implements the above-described method for detecting ghost targets.

[0123] Similarly, the content of the above method embodiments is applicable to the embodiments of this electronic device. The specific functions implemented by the embodiments of this electronic device are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0124] This application embodiment also provides a computer-readable storage medium storing a program executable by a processor 410, which, when executed by the processor 410, is used to perform the above-described method for detecting ghost-peeping targets.

[0125] Similarly, the content of the above method embodiments is applicable to the present computer-readable storage medium embodiments. The specific functions implemented by the present computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0126] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the aforementioned method for detecting ghost-peeping targets.

[0127] Similarly, the content of the above method embodiments is applicable to the embodiments of this computer program product. The specific functions implemented by the embodiments of this computer program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0128] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0129] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0130] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0131] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0132] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0133] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0134] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0135] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0136] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for detecting a ghost-protruding target, characterized in that, The method includes: The system collects first driving data and image data of the road ahead during the vehicle's journey, and uses the vehicle-road-cloud system to determine the second driving data of pedestrians, their driving direction, and the location information of the target vehicle and pedestrians. Based on the first driving data, determine the pulse transmission frequency; Image feature data is obtained by extracting features from the image data, and the image feature data is input into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result; Fuzzy clustering is performed based on the first driving data, the second driving data, the driving direction, and the location information, and the objective function corresponding to the membership degree is minimized to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position; The first prediction result and the second prediction result are fused to obtain the comprehensive prediction result corresponding to the ghost peek target.

2. The method for detecting a ghost-protruding target according to claim 1, characterized in that, The step of determining the pulse transmission frequency based on the first driving data includes: Based on the first driving data, the first driving speed and acceleration of the target vehicle are determined; The pulse transmission frequency is determined based on the first driving speed and the acceleration.

3. The method for detecting a ghost-protruding target according to claim 1, characterized in that, The step of extracting features from the image data to obtain image feature data includes: The image data is input into a convolutional neural network; The image feature data is obtained by extracting features from the image data using the convolutional neural network.

4. The method for detecting a ghost-protruding target according to claim 1, characterized in that, The step of inputting the image feature data into a spiking neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result includes: The image feature data is input into a spiking neural network for prediction based on the pulse transmission frequency, and initial prediction results are obtained at multiple time steps. The initial prediction results are summarized to obtain the summarized prediction results; The first prediction result is obtained by correcting the aggregated prediction result based on the first driving data.

5. A method for detecting a ghost-protruding target according to any one of claims 1-4, characterized in that, The process of fusing the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost-peeking target includes: Determine the first confidence level corresponding to the spiking neural network and the second confidence level corresponding to the fuzzy clustering; Based on the first confidence level and the second confidence level, determine the first weight corresponding to the first prediction result and the second weight corresponding to the second prediction result; Based on the first weight and the second weight, the first prediction result and the second prediction result are weighted and fused to obtain the comprehensive prediction result corresponding to the ghost peek target.

6. The method for detecting a ghost-protruding target according to claim 5, characterized in that, The method further includes: If the comprehensive prediction results indicate the presence of a "ghost vehicle" (a vehicle that suddenly appears out of nowhere), an alarm message is sent to the driver of the target vehicle.

7. A device for detecting ghost-protruding targets, characterized in that, The device includes: The data acquisition unit is used to collect the first driving data and image data of the road ahead during the vehicle's driving process, and to determine the second driving data, driving direction, and location information of the target vehicle and pedestrian through the vehicle-road-cloud system. The processing unit is used to determine the pulse transmission frequency based on the first driving data; The first prediction unit is used to extract features from the image data to obtain image feature data, and input the image feature data into the pulse neural network according to the pulse transmission frequency to predict the ghost peek target and obtain a first prediction result. The second prediction unit is used to perform fuzzy clustering based on the first driving data, the second driving data, the driving direction, and the location information, and to minimize the objective function corresponding to the membership degree to obtain the second prediction result corresponding to the ghost peek target; wherein, the cluster center of the fuzzy clustering is the initial distance between the target vehicle and the pedestrian and the distance between the collision point and the center point of the target vehicle when the target vehicle travels to the collision point position; The fusion unit is used to fuse the first prediction result and the second prediction result to obtain the comprehensive prediction result corresponding to the ghost peek target.

8. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a method for detecting ghost targets as described in any one of claims 1-6.

9. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to implement a method for detecting ghost-protruding targets as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a method for detecting ghost-protruding targets according to any one of claims 1-6.

Images