Development and application method and system of high-sensing obstacle avoidance cone barrel

Abstract

This invention discloses a method and system for the research and application of highly perceptive obstacle avoidance cones, relating to the field of intelligent transportation infrastructure perception and recognition technology. The method includes acquiring the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones; constructing a perception enhancement model and a multimodal perception gain matrix through joint modeling; obtaining prior information of the perception model; using a data-physical dual-drive multi-parameter intelligent identification algorithm to perform joint inversion and multimodal fusion identification of multi-source perception data under physical constraints to obtain initial identification results, model parameters to be calibrated, and fusion identification parameters to be calibrated, thereby constructing an obstacle avoidance comprehensive test platform and acquiring an obstacle avoidance perception database; after joint calibration of the model parameters to be calibrated and the fusion identification parameters to be calibrated, constructing a perception-enhanced obstacle avoidance cone equipment for verification; determining the obstacle avoidance warning threshold based on the verification results; establishing an obstacle avoidance safety decision-making and engineering application system; and improving the recognition robustness, perception accuracy, and engineering application capabilities of the obstacle avoidance cones.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation infrastructure perception and recognition technology, specifically involving a method and system for the research and application of a high-perception obstacle avoidance cone. Background Technology

[0002] Obstacle avoidance cones, as crucial safety facilities in road construction zones, emergency response zones, and temporary traffic guidance areas, are no longer merely serving a traditional warning function in advanced intelligent driving and multimodal environmental perception systems. They are increasingly becoming important targets influencing the environmental understanding, path planning, and risk decision-making of autonomous vehicles. Existing obstacle avoidance cones are typically made of materials such as plastic or rubber, which suffer from issues such as small geometric dimensions, limited radar cross-section, unstable retroreflective characteristics, and significant decreases in perception confidence under complex conditions such as rain, fog, strong light, and high-speed dynamics. These issues can easily lead to missed detections, false detections, or inaccurate positioning by millimeter-wave radar, lidar, and visual sensors.

[0003] Existing obstacle avoidance target perception and recognition methods mostly focus on target detection and analysis in single modalities, single materials, or steady-state environments. They largely rely on passive reflection characteristics or simple data-driven algorithms, failing to fully consider the coupling relationship between the geometric scattering characteristics of obstacle avoidance cones, retroreflective material properties, and active near-infrared pulse-coded signals. They also lack systematic modeling of the multi-parameter intertwined influence mechanism of "geometry-material-supplementary lighting coding." Furthermore, existing intelligent recognition methods still suffer from insufficient generalization ability in complex traffic scenarios, sensitivity to environmental clutter, and difficulty in achieving high-precision, low-latency recognition under high-speed dynamic conditions. They cannot yet meet the application requirements of robust perception of small obstacles in advanced intelligent driving scenarios.

[0004] Therefore, how to provide a method for the research and application of highly perceptive obstacle avoidance cones to improve the recognition robustness, perception accuracy and engineering application capabilities of obstacle avoidance cones in complex environments, high-speed dynamic scenes and under the condition of multi-modal sensor collaborative perception has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a method and system for the development and application of highly perceptive obstacle avoidance cones.

[0006] The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for the development and application of a high-sensitivity obstacle avoidance cone, comprising: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cones, and construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; use the perception enhancement model and the multimodal perception gain matrix as prior information of the perception model; Based on the prior information of the perception model, a pre-constructed dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multimodal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints, so as to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated. An obstacle avoidance integrated test platform is constructed based on the model parameters to be calibrated and the fusion identification parameters to be calibrated. The obstacle avoidance integrated test platform is used to conduct multi-condition static and dynamic tests on the obstacle avoidance cone to obtain an obstacle avoidance perception database. Based on the obstacle avoidance perception database and the initial identification results, the model parameters to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters. A perception-enhanced obstacle avoidance cone device is constructed based on the aforementioned combination of perception enhancement parameters. The perception-enhanced obstacle avoidance cone equipment is verified using the calibrated model parameters and the calibrated fusion identification parameters. Based on the verification results, the obstacle avoidance warning threshold is determined, and an obstacle avoidance safety decision-making and engineering application system is established.

[0007] Secondly, this invention provides a high-sensitivity obstacle avoidance cone research and application system, including: The acquisition module is used to acquire the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones, and to construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; the perception enhancement model and the multimodal perception gain matrix are used as prior information of the perception model; The identification module is used to perform joint inversion and multimodal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints based on the prior information of the perception model and using a pre-built data-object dual-drive multi-parameter intelligent identification algorithm to obtain the initial identification result, the model parameters to be calibrated, and the fusion identification parameters to be calibrated. The calibration module is used to construct an obstacle avoidance integrated test platform based on the parameters of the model to be calibrated and the fusion identification parameters to be calibrated, and to conduct multi-condition static and dynamic tests on the obstacle avoidance cone using the obstacle avoidance integrated test platform to obtain an obstacle avoidance perception database; based on the obstacle avoidance perception database and the initial identification results, the parameters of the model to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters; The construction module is used to construct a perception-enhanced obstacle avoidance cone device based on the combination of the perception enhancement parameters. The verification module is used to verify the perception-enhanced obstacle avoidance cone equipment using the calibrated model parameters and the calibrated fusion identification parameters, and to determine the obstacle avoidance warning threshold based on the verification results, thereby establishing an obstacle avoidance safety decision-making and engineering application system.

[0008] Thirdly, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps described in any of the above-mentioned methods for the development and application of high-sensitivity obstacle avoidance cones.

[0009] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps described in any of the above-described methods for the development and application of high-sensitivity obstacle avoidance cones.

[0010] This invention provides a method for the development and application of high-sensitivity obstacle avoidance cones. By jointly modeling the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones, a perception enhancement model and a multimodal perception gain matrix are constructed. This enables the coupled analysis of multiple physical quantities and the quantitative evaluation of the perception enhancement effect, providing reliable physical prior information for subsequent identification.

[0011] Based on prior information from the perception model, a dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multimodal fusion identification of multi-source perception data under physical constraints. By embedding physical mechanism constraints into the identification process, the problem of insufficient generalization ability of pure data-driven methods in complex environments is overcome, and the identification accuracy and robustness of the position, category and motion state of obstacle avoidance cones are significantly improved.

[0012] By constructing an obstacle avoidance integrated test platform, conducting static and dynamic tests under multiple working conditions, obtaining an obstacle avoidance perception database, and jointly calibrating the parameters of the model to be calibrated and the fusion identification parameters, the optimal combination of perception enhancement parameters was obtained, providing data support and parameter basis for the development of perception-enhanced obstacle avoidance cone equipment.

[0013] This invention can significantly improve the recognition confidence, perception robustness, and parameter inversion accuracy of small obstacle avoidance targets in complex environments and high-speed dynamic scenarios. It can also enhance the detection range, positioning accuracy, and engineering applicability of multimodal sensors for obstacle avoidance cones, providing important technical support for road construction safety, normal operation, and intelligent traffic supervision in high-level intelligent driving environments.

[0014] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the development and application method of a high-sensitivity obstacle avoidance cone provided in an embodiment of the present invention; Figure 2This is a flowchart illustrating another method for developing and applying a high-sensitivity obstacle avoidance cone provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a high-sensitivity obstacle avoidance cone research and application system provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0017] To address the issue that existing obstacle avoidance cones cannot meet the robust perception requirements of small obstacles in advanced intelligent driving scenarios, this invention provides a method for developing and applying high-perception obstacle avoidance cones. (See [link]). Figure 1 , Figure 1 This is a flowchart illustrating the development and application method of a high-sensitivity obstacle avoidance cone provided by an embodiment of the present invention, specifically including the following steps: Step S101: Obtain the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cone, and construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; use the perception enhancement model and the multimodal perception gain matrix as prior information of the perception model.

[0018] In this embodiment of the invention, the physical property parameters and environmental configuration parameters of the obstacle avoidance cone are obtained, and a perception enhancement model and a multimodal perception gain matrix are constructed through joint modeling, including: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cone; Based on joint modeling of physical property parameters and environmental configuration parameters, a geometric scattering characteristic model, an antireflection material response model, and an active near-infrared coded response model are constructed. Based on the retroreflective material response model, the active near-infrared coded response model, and environmental configuration parameters, a control equation for active-passive composite detection is constructed. The geometric scattering characteristic model, the retroreflective material response model, the active near-infrared coded response model, and the active-passive composite detection control equation are used as the sensing enhancement model; Based on the perception enhancement model, the perception gain parameters corresponding to each sensor mode are calculated using preset evaluation indicators, and a multimodal perception gain matrix is ​​constructed.

[0019] See Figure 2 , Figure 2This is a flowchart illustrating another method for developing and applying a high-sensitivity obstacle avoidance cone provided in an embodiment of the present invention. First, the physical property parameters and environmental configuration parameters of the obstacle avoidance cone are obtained. The physical property parameters include geometric structure parameters, surface retroreflective material parameters, active near-infrared pulse coding parameters, and near-infrared receiving channel parameters. The environmental configuration parameters include sensor observation parameters, environmental propagation parameters, environmental noise parameters, and synchronous sampling parameters.

[0020] In this embodiment of the invention, geometric parameters may include the overall dimensions of the obstacle avoidance cone, the cone tilt angle, the dimensions of the base structure, and the local curved surface shape parameters; surface retroreflective material parameters may include the type of retroreflective film, retroreflective coefficient, reflective film coverage area, surface aging degree, material attenuation parameters, and reflection response parameters under the working bands of lidar and near-infrared; active near-infrared pulse coding parameters may include emission intensity, coding frequency, pulse duty cycle, modulation method, and emission timing parameters; sensor observation parameters may include the installation position, observation distance, observation angle, and sampling frequency of the millimeter-wave radar, lidar, and camera; near-infrared receiving channel parameters may include receiver type, receiving sensitivity, exposure or integration time, and timestamp parameters; environmental propagation parameters may include environmental transmittance, ambient light intensity, rain and fog conditions, and road background reflection conditions; environmental noise parameters may include environmental clutter intensity and background noise level; and synchronous sampling parameters may include the synchronization clock parameters and sampling timing parameters between the active near-infrared pulse coding transmitter and the near-infrared receiving channel.

[0021] In this embodiment of the invention, the sensing enhancement model includes a geometric scattering feature model, an inverse reflection material response model, an active near-infrared coded response model, and an active-passive composite detection control equation.

[0022] Based on electromagnetic scattering theory, and combining geometric structural parameters, surface retroreflective material parameters, and sensor observation parameters, a geometric scattering characteristic model is established to calculate the equivalent radar cross-section and electromagnetic echo intensity of the obstacle avoidance cone. Based on photometry and retroreflection principles, and combining surface retroreflective material parameters and environmental propagation parameters, an retroreflective material response model is established to calculate the passive reflection response signal of the obstacle avoidance cone. Based on surface retroreflective material parameters, active near-infrared pulse coding parameters, near-infrared receiving channel parameters, and environmental propagation parameters, an active near-infrared coded response model is established to calculate the active near-infrared coded response signal.

[0023] Based on the passive reflection response signal, the active near-infrared coded response signal, environmental noise parameters, and synchronous sampling parameters, an active-passive composite detection control equation is established to calculate the active-passive composite detection response. Using at least one of the effective detection distance, recognition confidence, and encoding / decoding reliability of millimeter-wave radar, lidar, camera, and near-infrared receiving channel before and after perception enhancement as preset evaluation indicators, the perception gain parameters corresponding to each sensor mode are calculated, and a multimodal perception gain matrix is ​​constructed.

[0024] In this embodiment of the invention, the equivalent radar cross section, electromagnetic echo intensity, passive reflection response signal, active near-infrared coded response signal, active-passive composite detection response, and multimodal sensing gain matrix together constitute the prior information of the sensing model.

[0025] In this embodiment of the invention, the process of constructing the active-passive composite detection control equations is as follows: The passive reflection response signal of the obstacle avoidance cone, the active near-infrared coded response signal, environmental noise, and the synchronous sampling relationship of the near-infrared receiving channel are collectively modeled as a combined active and passive detection response. The passive reflection response signal is mainly determined by the retroreflective material's reflection coefficient, the reflective material's coverage area, the sensor's incident angle, the detection distance, and the environmental transmittance. The active near-infrared coded response signal is mainly determined by the emission intensity of the active near-infrared pulse-coded transmitter mounted on the obstacle avoidance cone, the pulse coding function, the retroreflective material response, the environmental transmittance, the sensitivity of the near-infrared receiver mounted on the intelligent vehicle or roadside sensing device, and the synchronous sampling relationship between the active near-infrared pulse-coded transmitter and the near-infrared receiver.

[0026] The control equation for the combined active and passive detection can be expressed as: ; ; ; in, This indicates the combined active and passive detection response signal received by the sensor; Indicates a passive reflection response signal; Indicates an active near-infrared coded response signal; Indicates environmental noise parameters; Indicates the passive reflection ratio; Indicates the reflection coefficient of retroreflective materials; Indicates the area covered by retroreflective material; Indicates the sensor's incident angle; This indicates the distance between the obstacle avoidance cone and the sensor; Indicates environmental transmittance; Indicates the active near-infrared coding response ratio; Indicates the intensity of the active near-infrared light source; Represents the near-infrared pulse coding function; Indicates the encoding frequency; Indicates the pulse width.

[0027] The aforementioned synchronous sampling parameters include the sampling frequency of the near-infrared receiving channel, the sampling timestamp, the sampling window length, the clock offset between the transmitting and receiving ends, and the coding period parameters of the active near-infrared pulse-coded signal. Among these, the coding frequency and pulse width are used to define the timing characteristics of the active near-infrared pulse-coded signal and are parameters used for code identification and time-domain alignment in the synchronous sampling relationship.

[0028] In this embodiment of the invention, based on the geometric scattering feature model, the retroreflective material response model, the active near-infrared coding response model, and the active-passive composite detection control equation, at least one of the effective detection distance, identification confidence, and coding-decoding reliability of millimeter-wave radar, lidar, camera, and near-infrared receiving channel before and after perception enhancement is used as a preset evaluation index to calculate the perception gain parameters corresponding to each sensor mode and construct a multimodal perception gain matrix.

[0029] To quantify the effects of different perception enhancement mechanisms on various intelligent driving sensor modes, this embodiment of the invention selects parallel evaluation indicators for each sensor mode: for millimeter-wave radar mode, the evaluation indicators are equivalent radar cross-section, electromagnetic echo intensity, maximum effective detection range, and recognition confidence; for lidar mode and camera mode, the evaluation indicators are passive optical response, maximum effective detection range, and recognition confidence; and for near-infrared receiving channel mode, the evaluation indicators are decoding accuracy, signal-to-noise ratio, maximum effective detection range, and recognition confidence of active near-infrared pulse coding response. At least one or more of these evaluation indicators can be selected to participate in the evaluation, depending on the specific mode, to construct a multi-modal perception gain matrix. : ; ; ; ; in, Indicates the first Effective detection range gain of sensor-like modes; In the enhanced perception state, the first The maximum effective detection range of the sensor-like modality; Indicates the first state under the reference condition The maximum effective detection range of the sensor-like modality; This represents a positive constant introduced to prevent the denominator from being zero; Indicates the first Confidence gain for sensor-like modal identification; In the enhanced perception state, the first Confidence level for sensor-like modal identification; Indicates the first state under the reference condition Confidence level for sensor-like modal identification; This indicates the decoding reliability gain of the near-infrared receiving channel for the active near-infrared pulse-coded response; This indicates the decoding reliability of the near-infrared receiving channel for the active near-infrared pulse-coded response under enhanced sensing conditions; This indicates the decoding reliability of the near-infrared receiving channel for the active near-infrared pulse-coded response under reference conditions; A category index representing sensor modes, including cameras. LiDAR Millimeter-wave radar and near-infrared receiving channel To prevent the introduction of positive numbers with a denominator of zero, for cameras, LiDAR, and millimeter-wave radar, since no encoding / decoding reliability indicators are involved, the third column of their corresponding matrices is set to 0.

[0030] The maximum effective detection distance, recognition confidence, and decoding reliability under the enhanced perception state were obtained through field tests on an obstacle avoidance integrated test platform or closed road after the enhanced perception obstacle avoidance cones or combinations of enhanced perception parameters were deployed. The corresponding parameters under the baseline state were obtained through tests under the same conditions without active near-infrared coded illumination, without scattering enhancement, or with ordinary obstacle avoidance cones. By comparing the enhanced state with the baseline state under the same conditions, the perception gain parameters of each sensor mode were obtained.

[0031] The rows of the multimodal sensing gain matrix correspond to sensor modes such as camera, lidar, millimeter-wave radar, and near-infrared receiving channel, while the columns correspond to effective detection distance gain, recognition confidence gain, and encoding / decoding reliability gain. For cameras, lidar, and millimeter-wave radar that do not involve active near-infrared pulse coding and decoding, their encoding / decoding reliability gain term is 0 or does not participate in the fusion weight calculation of the corresponding mode.

[0032] Step S102: Based on the prior information of the perception model, the pre-built data-object dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multi-modal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints, so as to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated.

[0033] First, the pre-constructed data-physical dual-drive multi-parameter intelligent identification algorithm provided in this embodiment of the invention will be described. The data-physical dual-drive multi-parameter intelligent identification algorithm includes an analytical model unit, a machine learning surrogate model unit, a multi-parameter joint inversion unit, a sensing response signal adaptive filtering unit, a physical constraint neural network unit, an encoded signal decoding feature extraction unit, and a multi-source data fusion identification unit.

[0034] The system comprises the following components: an analytical model unit to characterize the electromagnetic scattering and optical propagation mechanisms of obstacle avoidance cones; a machine learning surrogate model unit to establish a nonlinear mapping relationship between multi-source sensing responses and target parameters; a multi-parameter joint inversion unit to construct a multi-parameter joint inversion framework of "geometry, reflection, and light source"; a sensing response signal adaptive filtering unit to suppress low signal-to-noise ratio clutter and environmental clutter; a physical constraint neural network unit to embed the electromagnetic scattering constraint relationship determined by the geometric scattering feature model, the passive reflection constraint relationship determined by the retroreflective material response model, the active near-infrared propagation and coding consistency constraint relationship determined by the active near-infrared coding response model, and the active-passive composite detection constraint relationship determined by the active-passive composite detection control equation into the model training and identification process; a coding signal decoding feature extraction unit to decode and extract features from the active near-infrared pulse coding signal; and a multi-source data fusion identification unit to combine echo intensity, point cloud distribution, infrared pulse frequency, and visual response information to jointly identify the spatial location, category attributes, and motion state parameters of obstacle avoidance cones.

[0035] In this embodiment of the invention, the prior information of the perception model is used to provide mechanistic constraints and fusion basis for the pre-constructed data-object dual-drive multi-parameter intelligent identification algorithm. The multi-source perception data includes at least millimeter-wave radar echo intensity data, lidar point cloud distribution data, camera visual image data, active near-infrared pulse-coded response data collected by the near-infrared receiving channel, and corresponding time synchronization data.

[0036] In this embodiment of the invention, based on prior information from the perception model, a pre-constructed dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multi-modal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints, obtaining initial identification results, model parameters to be calibrated, and fusion identification parameters to be calibrated, including: Mechanism modeling and nonlinear feature mapping are performed on multi-source sensing data to obtain mechanism feature information and data-driven feature information. By combining prior information from the perception model, joint inversion processing is performed on mechanistic feature information and data-driven feature information to obtain initial inversion parameters; Based on the initial inversion parameters, clutter suppression and signal enhancement processing are performed on the multi-source sensing data to obtain the filtered sensing response characteristics. The physical constraint relationship in the perception enhancement model is embedded into the identification process, and the filtered perception response features and initial inversion parameters are optimized by physical constraints to obtain identification features that satisfy the physical mechanism constraints. Decode and extract features from active near-infrared pulse-coded response data in multi-source sensing data to obtain coded near-infrared signal features; The fusion weights of each sensor mode are determined based on the multimodal sensing gain matrix. The identification features that meet physical consistency, the encoded near-infrared signal features, and the multi-source sensing data are fused and identified to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated.

[0037] Specifically, the analytical model unit is used to perform mechanism modeling on multi-source sensing data by utilizing the prior information of the sensing model, including the equivalent radar cross-section, electromagnetic echo intensity, passive reflection response signal, active near-infrared coded response signal, and active-passive composite detection response, to obtain mechanism feature information including the electromagnetic scattering feature information and optical propagation feature information of the obstacle avoidance cone. The machine learning agent model unit is used to perform nonlinear feature mapping processing on multi-source perception data to obtain data-driven identification feature information of obstacle avoidance cones. The multi-parameter joint inversion unit is used to combine prior information from the perception model to perform joint inversion processing on electromagnetic scattering feature information, optical propagation feature information, and data-driven identification feature information to obtain the initial inversion parameters of the obstacle avoidance cone. The adaptive filtering unit for sensing response signal is used to perform clutter suppression and signal enhancement processing on multi-source sensing data according to the initial inversion parameters, and to obtain the filtered sensing response characteristics. The physical constraint neural network unit is used to embed the electromagnetic scattering constraint relationship, passive reflection constraint relationship, active near-infrared propagation and coding consistency constraint relationship and active-passive composite detection constraint relationship defined by the geometric scattering feature model, retroreflection material response model, active near-infrared coding response model and active-passive composite detection control equation into the identification process. It performs physical constraint optimization on the filtered sensing response features and initial inversion parameters to obtain identification features that meet the physical mechanism constraints. The encoded signal decoding feature extraction unit is used to decode and extract features from the measured active near-infrared pulse encoded response data collected by the near-infrared receiving channel, using the active near-infrared encoded response signal calculated by the active near-infrared encoded response model as a reference, to obtain the encoded near-infrared signal features. Among them, the near-infrared receiving channel is a receiving channel set on intelligent driving vehicles, roadside sensing devices or obstacle avoidance integrated test platforms, used to receive coded near-infrared signals emitted by the active near-infrared pulse coding emission module at the obstacle avoidance cone end; the obstacle avoidance cone end mainly undertakes the active near-infrared pulse coding emission and retroreflection enhancement functions, while the receiving channel side undertakes the functions of actual response acquisition, decoding and feature extraction.

[0038] The multi-source data fusion identification unit is used to determine the fusion weight of each sensor mode based on the multi-modal sensing gain matrix, and to use the fusion weight to fuse and identify the identification features, coded near-infrared signal features and multi-source sensing data to obtain the initial identification results including initial position parameters, initial category parameters and initial motion state parameters, as well as the model parameters to be calibrated and the fusion identification parameters to be calibrated. The model parameters to be calibrated include geometric scattering characteristic model parameters, retroreflective material response model parameters, active near-infrared coded response model parameters, and active-passive composite detection control equation parameters.

[0039] In this embodiment of the invention, the physical constraint relationship in the perception enhancement model is embedded into the identification process, and the filtered perception response features and initial inversion parameters are optimized for physical constraints to obtain identification features that satisfy the physical mechanism constraints, including: Electromagnetic scattering constraint relationship is constructed based on the geometric scattering characteristic model, passive reflection constraint relationship is constructed based on the retroreflective material response model, active near-infrared propagation and coding consistency constraint relationship is constructed based on the active near-infrared coding response model, and active-passive composite detection constraint relationship is constructed based on the active-passive composite detection control equation. A joint loss function is constructed based on the constraints of electromagnetic scattering, passive reflection, active near-infrared propagation and coding consistency, and active-passive composite detection. Based on the joint loss function, the identified features are obtained by iteratively solving using the filtered sensing response features and the initial inversion parameters as input.

[0040] Specifically, the electromagnetic scattering constraint relationship, passive reflection constraint relationship, active near-infrared propagation and coding consistency constraint relationship, and active-passive composite detection constraint relationship are embedded into the joint loss function of the physical constraint neural network unit; Based on the joint loss function, the physical constraint neural network unit is iteratively trained; the filtered sensing response features and initial inversion parameters are used as inputs, and the intermediate identification features are output through forward propagation of the physical constraint neural network unit. Error correction and feature reconstruction are performed on the intermediate identification features to obtain identification features that satisfy the physical mechanism constraints.

[0041] In this embodiment of the invention, to satisfy both electromagnetic scattering and optical propagation laws while fitting data, a joint loss function is constructed, consisting of data fitting loss, electromagnetic scattering constraint loss, optical propagation constraint loss, and regularization loss. Specifically, the electromagnetic scattering constraint loss is used to ensure that the equivalent radar cross-section predicted by the model is consistent with the calculation result of the geometric scattering characteristic model, while the optical propagation constraint loss is used to ensure that the sensor response signal predicted by the model is consistent with the calculation result of the active-passive composite detection control equation.

[0042] Specifically, based on the filtered sensing response features and the initial inversion parameters, a data fitting loss term is constructed to characterize the consistency between the identification results and the measured sensing response; An electromagnetic scattering constraint loss term is constructed based on the geometric scattering feature model. The electromagnetic scattering constraint loss term is used to characterize the deviation between the identification result and the electromagnetic scattering constraint relationship. An optical propagation constraint loss term is constructed based on the retroreflective material response model, the active near-infrared coded response model, and the active-passive composite detection control equation. The optical propagation constraint loss term is used to characterize the deviation between the sensor response signal predicted by the model and the passive reflection response signal, the active near-infrared coded response signal, and the active-passive composite detection response. A regularized loss term is constructed based on the parameter size and parameter variation amplitude of the physical constraint neural network. The regularized loss term is used to suppress model overfitting and improve the generalization ability under complex working conditions. By weighting and combining the data fitting loss term, electromagnetic scattering constraint loss term, optical propagation constraint loss term, and regularization loss term, a joint loss function for the physical constraint neural network unit is obtained, and an optimization model for the joint constraint of data fitting and physical mechanism is established. Finally, based on the joint loss function, the optimized model is iteratively trained and its parameters are updated to obtain the identification features that satisfy the physical mechanism constraints.

[0043] In this embodiment of the invention, the joint loss function includes: ; ; ; ; ; in, Indicates the total loss; This represents the data fitting loss; This represents the first adjustable weighting coefficient; This represents the electromagnetic scattering confinement loss; This represents the second adjustable weighting coefficient; This represents the optical propagation constraint loss; This represents the third adjustable weighting coefficient; Indicates the regularization loss; This represents the obstacle avoidance cone recognition result predicted by the optimized model; This indicates the measured calibration results; This represents the equivalent radar cross section predicted by the optimized model; This represents the equivalent radar cross section calculated from the geometric scattering characteristic model; This represents the sensor response signal predicted by the optimization model; Indicates a passive reflection response signal; Indicates an active near-infrared coded response signal; The parameters represent the physical constraints of the neural network unit.

[0044] In this embodiment of the invention, the obstacle avoidance cone recognition result predicted by the optimized model is a structured identification result output by the physical constraint neural network unit during iterative training and forward propagation, including at least the position parameters, category parameters, and motion state parameters of the obstacle avoidance cone. This structured identification result is fused by the multi-source data fusion identification unit to form an initial identification result, which is further corrected during subsequent experimental platform calibration. The equivalent radar cross section predicted by the optimized model and the sensor response signal predicted by the optimized model are respectively predicted by the physical constraint neural network unit based on the filtered sensing response characteristics and the initial inversion parameters, and are used to constrain consistency with the calculation results of the geometric scattering feature model and the active-passive composite detection control equation.

[0045] In this embodiment of the invention, the active near-infrared pulse-coded response data in multi-source sensing data is decoded and its features are extracted to obtain coded near-infrared signal features, including: The active near-infrared pulse coding response data and the corresponding time synchronization data collected by the near-infrared receiving channel are input to the coding signal decoding feature extraction unit. Based on the active near-infrared coded response signal, active near-infrared pulse coding parameters and time synchronization data, the active near-infrared pulse coding response data is encoded, time-domain decoded and frequency-domain analyzed to obtain the coded sequence information. The encoded sequence information is denoised, segmented, and normalized to obtain a standardized encoded signal; Based on the standardized coded signal, the active near-infrared illumination intensity, pulse interval features, and temporal variation features are extracted to obtain the coded near-infrared signal features.

[0046] In this embodiment of the invention, the fusion weights of each sensor mode are determined based on the multimodal sensing gain matrix. The identified features, coded near-infrared signal features, and multi-source sensing data are then fused and identified to obtain initial identification results, model parameters to be calibrated, and fusion identification parameters to be calibrated, including: Based on the temporal markers and spatial observation information in multi-source sensing data, the identified features and coded near-infrared signal features are temporally aligned and spatially correlated to obtain multi-source correlated features. Based on the multimodal sensing gain matrix, the fusion weights corresponding to the millimeter-wave radar mode, lidar mode, camera mode, and near-infrared receiving channel mode are determined, and each fusion weight and its update parameters are used as the initial fusion identification parameters. Based on the initial fusion identification parameters, the multi-source correlation features, coded near-infrared signal features and multi-source sensing data are fused to obtain the fusion identification features. Based on the fusion identification features, the spatial position, category attributes and temporal changes of the obstacle avoidance cones are jointly solved to obtain the initial identification results and the model parameters to be calibrated, including the initial position parameters, initial category parameters and initial motion state parameters. The initial fusion identification parameters are then used as the fusion identification parameters to be calibrated.

[0047] Step S103: Construct an obstacle avoidance integrated test platform based on the model parameters to be calibrated and the fusion identification parameters to be calibrated, and conduct multi-condition static and dynamic tests on the obstacle avoidance cone using the obstacle avoidance integrated test platform to obtain an obstacle avoidance perception database; perform joint calibration on the model parameters to be calibrated and the fusion identification parameters to be calibrated based on the obstacle avoidance perception database and the initial identification results to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters.

[0048] In this embodiment of the invention, based on the parameters of the model to be calibrated and the fusion identification parameters to be calibrated, the geometric structure parameters, surface retroreflective material parameters, active near-infrared illumination intensity, pulse coding parameters, near-infrared receiving channel synchronous sampling parameters, millimeter-wave radar observation distance, lidar observation distance, camera observation distance, observation angle, relative motion speed, and environmental conditions of the obstacle avoidance cone are set to construct a comprehensive obstacle avoidance test platform adapted to multi-modal sensors such as millimeter-wave radar, lidar, and cameras. The comprehensive obstacle avoidance test platform includes a dynamic reflection coefficient control device, an active near-infrared signal coding intensity fine control device, an environmental simulation device, and a synchronous acquisition and monitoring system.

[0049] The environmental simulation device can simulate typical complex working conditions such as rain, fog, and strong light; the dynamic adjustment device for reflectivity coefficient is used to change the attenuation degree of reflective film or the surface reflection state; the active near-infrared signal encoding intensity fine adjustment device is used to change the intensity, frequency and encoding strategy of near-infrared supplementary light; the synchronous acquisition and monitoring system is used to record millimeter-wave radar, lidar, visual images and near-infrared response data in real time.

[0050] Then, by conducting a series of static and dynamic tests, an obstacle avoidance perception database of the obstacle avoidance cones under different working conditions, distances, angles, and driving speeds was obtained, and the model sensitivity, perception boundary, and error correction parameters were corrected based on the test results.

[0051] In this embodiment of the invention, joint calibration is performed on the model parameters to be calibrated and the fusion identification parameters to be calibrated based on the obstacle avoidance perception database to obtain a combination of calibrated model parameters, calibrated fusion identification parameters, and perception enhancement parameters, including: Based on the obstacle avoidance perception database and the pre-calibrated spatial location, category attributes and motion state parameters of the obstacle avoidance cones, the perception effectiveness of the initial identification results is evaluated, and the evaluation results of the obstacle avoidance cones' position identification error, category identification accuracy, motion state identification error and perception gain of each sensor mode are obtained. Based on the perception gain evaluation results, position identification error, category identification accuracy, and motion state identification error, the parameters of the model to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters.

[0052] The perception gain evaluation results are obtained by comparing the perception modes of each sensor before and after perception enhancement based on the obstacle avoidance perception database. These include effective detection distance gain, recognition confidence gain, encoding / decoding reliability gain, and robustness evaluation results under different environmental conditions. Joint calibration aims to reduce position identification error, improve category identification accuracy, reduce motion state identification error, and enhance the perception gain of each mode. Through parameter sensitivity analysis and multi-objective weighted iterative optimization, the parameters of the geometric scattering feature model, retroreflective material response model, active near-infrared coded response model, active-passive composite detection control equations, and the fusion weights of each mode are updated synchronously.

[0053] The sensing enhancement parameter combination includes the geometry, deployment angle, and installation location of the millimeter-wave radar scattering enhancement structure; the material type, reflectivity, and coverage area of ​​the retroreflective coating; the light source intensity, coding frequency, pulse width, and transmission period of the active near-infrared pulse-coded emission module; and the triggering strategy, sampling window, and synchronization control parameters of the low-power control module. This sensing enhancement parameter combination is the optimal combination that meets the preset detection range, identification confidence level, decoding reliability, and power consumption constraints.

[0054] Step S104: Construct a perception-enhanced obstacle avoidance cone device based on the combination of perception enhancement parameters.

[0055] In this embodiment of the invention, based on the combination of perception enhancement parameters, a series of perception-enhanced obstacle avoidance cone devices are developed, integrating millimeter-wave radar scattering enhancement structure, retroreflective film, active near-infrared pulse code transmission module and low-power control module.

[0056] Specifically, based on the combination of perception enhancement parameters, the cone specifications, reflective film coverage scheme, millimeter-wave radar scattering enhancement structure layout scheme, coding strategy of active near-infrared pulse coding transmission module, and working mode of low-power control module are determined for different application scenarios, thereby forming a perception enhancement obstacle avoidance cone equipment suitable for high-speed construction areas, closed road test areas, and complex lighting environments.

[0057] Step S105: The perception-enhanced obstacle avoidance cone equipment is verified using the calibrated model parameters and the calibrated fusion identification parameters. Based on the verification results, the obstacle avoidance warning threshold is determined, and an obstacle avoidance safety decision-making and engineering application system is established.

[0058] In this embodiment of the invention, based on the calibrated model parameters and the calibrated fusion identification parameters, the perception-enhanced obstacle avoidance cone equipment is verified using an advanced intelligent driving system in a closed road scenario. Millimeter-wave radar echo data, lidar point cloud data, camera visual image data, and active near-infrared pulse-coded response data are collected under different driving speeds, detection distances, observation angles, and environmental conditions such as rain, fog, strong light, and weak light. The on-site measured perception data is compared with the model prediction results to obtain verification results such as effective detection distance, recognition confidence, position identification error, category identification accuracy, motion state identification error, and encoding / decoding reliability.

[0059] Based on the verification results, obstacle avoidance warning thresholds are determined. These thresholds include the minimum effective detection distance threshold, the minimum identification confidence threshold, the maximum position identification error threshold, the minimum encoding / decoding reliability threshold, and the minimum safe response time threshold. When the on-site measured perception data shows that any indicator is lower than the corresponding threshold, or when the safe response time calculated based on the relative distance and relative speed is lower than the preset value, an obstacle avoidance safety decision is generated, thereby establishing an obstacle avoidance safety decision and engineering application system.

[0060] Among these, obstacle avoidance safety decisions can be implemented by outputting deceleration prompts, braking warnings, lane change avoidance prompts, manual intervention prompts, or construction zone risk warnings to the advanced intelligent driving system. For example, when the perception-enhanced obstacle avoidance cone is stably identified at a preset distance and the identification confidence level is higher than a threshold, the system outputs a decision to proceed normally or maintain the lane; when the identification confidence level is lower than the threshold, the position identification error exceeds the threshold, or the safety reaction time is insufficient, the system outputs obstacle avoidance decisions such as deceleration, braking, or manual intervention.

[0061] In this embodiment of the invention, by jointly modeling the physical property parameters and environmental configuration parameters of the obstacle avoidance cone, a perception enhancement model and a multimodal perception gain matrix are constructed, realizing the quantitative evaluation of the coupling analysis of multiple physical quantities and the perception enhancement effect, and providing reliable physical prior information for subsequent identification.

[0062] Based on prior information from the perception model, a dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multimodal fusion identification of multi-source perception data under physical constraints. By embedding physical mechanism constraints into the identification process, the problem of insufficient generalization ability of pure data-driven methods in complex environments is overcome, and the identification accuracy and robustness of the position, category and motion state of obstacle avoidance cones are significantly improved.

[0063] By constructing an obstacle avoidance integrated test platform, conducting static and dynamic tests under multiple working conditions, obtaining an obstacle avoidance perception database, and jointly calibrating the parameters of the model to be calibrated and the fusion identification parameters, the optimal combination of perception enhancement parameters was obtained, providing data support and parameter basis for the development of perception-enhanced obstacle avoidance cone equipment.

[0064] This invention can significantly improve the recognition confidence, perception robustness, and parameter inversion accuracy of small obstacle avoidance targets in complex environments and high-speed dynamic scenarios. It can also enhance the detection range, positioning accuracy, and engineering applicability of multimodal sensors for obstacle avoidance cones, providing important technical support for road construction safety, normal operation, and intelligent traffic supervision in high-level intelligent driving environments.

[0065] Based on the same inventive concept, this invention also provides a high-sensitivity obstacle avoidance cone research and application system, see [link to relevant documentation]. Figure 3 , Figure 3 This is a schematic diagram of the structure of a high-sensitivity obstacle avoidance cone research and application system provided in an embodiment of the present invention. The high-sensitivity obstacle avoidance cone research and application system includes: The acquisition module 301 is used to acquire the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones, and to construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; the perception enhancement model and the multimodal perception gain matrix are used as prior information of the perception model. The identification module 302 is used to perform joint inversion and multimodal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints based on the prior information of the perception model and using a pre-built data-object dual-drive multi-parameter intelligent identification algorithm to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated. The calibration module 303 is used to construct an obstacle avoidance integrated test platform based on the parameters of the model to be calibrated and the fusion identification parameters to be calibrated, and to conduct multi-condition static and dynamic tests on the obstacle avoidance cones using the obstacle avoidance integrated test platform to obtain an obstacle avoidance perception database; based on the obstacle avoidance perception database and the initial identification results, the parameters of the model to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters; Module 304 is used to construct a perception-enhanced obstacle avoidance cone device based on a combination of perception enhancement parameters. The verification module 305 is used to verify the perception-enhanced obstacle avoidance cone equipment using the calibrated model parameters and the calibrated fusion identification parameters, and to determine the obstacle avoidance warning threshold based on the verification results, thereby establishing an obstacle avoidance safety decision-making and engineering application system.

[0066] Optionally, module 301 acquires the physical property parameters and environmental configuration parameters of the obstacle avoidance cone, and constructs a perception enhancement model and a multimodal perception gain matrix through joint modeling, including: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cone; Based on joint modeling of physical property parameters and environmental configuration parameters, a geometric scattering characteristic model, an antireflection material response model, and an active near-infrared coded response model are constructed. Based on the retroreflective material response model, the active near-infrared coded response model, and environmental configuration parameters, a control equation for active-passive composite detection is constructed. The geometric scattering characteristic model, the retroreflective material response model, the active near-infrared coded response model, and the active-passive composite detection control equation are used as the sensing enhancement model; Based on the perception enhancement model, the perception gain parameters corresponding to each sensor mode are calculated using preset evaluation indicators, and a multimodal perception gain matrix is ​​constructed.

[0067] In this embodiment of the invention, by jointly modeling the physical property parameters and environmental configuration parameters of the obstacle avoidance cone, a perception enhancement model and a multimodal perception gain matrix are constructed, realizing the quantitative evaluation of the coupling analysis of multiple physical quantities and the perception enhancement effect, and providing reliable physical prior information for subsequent identification.

[0068] Based on prior information from the perception model, a dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multimodal fusion identification of multi-source perception data under physical constraints. By embedding physical mechanism constraints into the identification process, the problem of insufficient generalization ability of pure data-driven methods in complex environments is overcome, and the identification accuracy and robustness of the position, category and motion state of obstacle avoidance cones are significantly improved.

[0069] By constructing an obstacle avoidance integrated test platform, conducting static and dynamic tests under multiple working conditions, obtaining an obstacle avoidance perception database, and jointly calibrating the parameters of the model to be calibrated and the fusion identification parameters, the optimal combination of perception enhancement parameters was obtained, providing data support and parameter basis for the development of perception-enhanced obstacle avoidance cone equipment.

[0070] This invention can significantly improve the recognition confidence, perception robustness, and parameter inversion accuracy of small obstacle avoidance targets in complex environments and high-speed dynamic scenarios. It can also enhance the detection range, positioning accuracy, and engineering applicability of multimodal sensors for obstacle avoidance cones, providing important technical support for road construction safety, normal operation, and intelligent traffic supervision in high-level intelligent driving environments.

[0071] This invention also provides an electronic device, such as... Figure 4 As shown, it includes a processor 401, a communication interface 402, a memory 403, and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404. Memory 403 is used to store computer programs; When the processor 401 executes the program stored in the memory 403, it implements the method steps of any of the above-mentioned high-sensitivity obstacle avoidance cone research and application methods.

[0072] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0073] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0074] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0075] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0076] The present invention also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program implements the method steps of any of the above-described high-sensitivity obstacle avoidance cone development and application methods.

[0077] Optionally, the computer-readable storage medium may be non-volatile memory (NVM), such as at least one disk storage device.

[0078] Optionally, the aforementioned computer-readable storage medium may also be at least one storage device located remotely from the aforementioned processor.

[0079] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the steps described in any of the above-described methods for the development and application of high-sensitivity obstacle avoidance cones.

[0080] It should be noted that the terms "first," "second," etc., 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. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of systems and methods consistent with some aspects of the invention.

[0081] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0082] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0083] The method provided in this invention can be applied to electronic devices. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc. No limitation is made herein; any electronic device that can implement this invention falls within the protection scope of this invention.

[0084] For system / electronic device / storage medium embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments.

[0085] It should be noted that the system, electronic device and storage medium in the embodiments of the present invention are respectively the system, electronic device and storage medium of the above-mentioned high-sensitivity obstacle avoidance cone research and application method. Therefore, all embodiments of the above-mentioned high-sensitivity obstacle avoidance cone research and application method are applicable to the system, electronic device and storage medium, and can achieve the same or similar beneficial effects.

[0086] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for developing and applying a high-sensitivity obstacle avoidance cone, characterized in that, include: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cones, and construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; use the perception enhancement model and the multimodal perception gain matrix as prior information of the perception model; Based on the prior information of the perception model, a pre-constructed dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multimodal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints, so as to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated. An obstacle avoidance integrated test platform is constructed based on the parameters of the model to be calibrated and the fusion identification parameters to be calibrated. The obstacle avoidance integrated test platform is used to conduct multi-condition static and dynamic tests on the obstacle avoidance cone to obtain an obstacle avoidance perception database. Based on the obstacle avoidance perception database and the initial identification results, the model parameters to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters and the combination of perception enhancement parameters. A perception-enhanced obstacle avoidance cone device is constructed based on the aforementioned combination of perception enhancement parameters. The perception-enhanced obstacle avoidance cone equipment is verified using the calibrated model parameters and the calibrated fusion identification parameters. Based on the verification results, the obstacle avoidance warning threshold is determined, and an obstacle avoidance safety decision-making and engineering application system is established.

2. The method for developing and applying high-sensitivity obstacle avoidance cones according to claim 1, characterized in that, Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cones, and construct a perception enhancement model and a multimodal perception gain matrix through joint modeling, including: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cone; Based on the physical property parameters and the environmental configuration parameters, a geometric scattering characteristic model, an antireflective material response model, and an active near-infrared coded response model are constructed through joint modeling. Based on the retroreflective material response model, the active near-infrared coded response model, and the environmental configuration parameters, a control equation for active-passive composite detection is constructed. The geometric scattering feature model, the retroreflective material response model, the active near-infrared coded response model, and the active-passive composite detection control equation are used as the sensing enhancement model; Based on the aforementioned perception enhancement model, the perception gain parameters corresponding to each sensor mode are calculated using preset evaluation indicators to construct a multimodal perception gain matrix.

3. The method for developing and applying high-sensitivity obstacle avoidance cones according to claim 2, characterized in that, Based on the prior information of the perception model, a pre-constructed dual-drive multi-parameter intelligent identification algorithm is used to perform joint inversion and multi-modal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints, obtaining initial identification results, model parameters to be calibrated, and fusion identification parameters to be calibrated, including: Mechanism modeling and nonlinear feature mapping are performed on multi-source sensing data to obtain mechanism feature information and data-driven feature information. By combining the prior information of the perception model, the mechanism feature information and the data-driven feature information are jointly inverted to obtain the initial inversion parameters; Based on the initial inversion parameters, clutter suppression and signal enhancement processing are performed on the multi-source sensing data to obtain filtered sensing response characteristics. The physical constraint relationship in the perception enhancement model is embedded into the identification process, and the filtered perception response features and the initial inversion parameters are optimized by physical constraints to obtain identification features that satisfy the physical mechanism constraints. Decode and extract features from the active near-infrared pulse-coded response data in the multi-source sensing data to obtain coded near-infrared signal features; The fusion weights of each sensor mode are determined based on the multimodal sensing gain matrix. The identification features, the coded near-infrared signal features, and the multi-source sensing data are fused and identified to obtain the initial identification results, the model parameters to be calibrated, and the fusion identification parameters to be calibrated.

4. The method for developing and applying high-sensitivity obstacle avoidance cones according to claim 3, characterized in that, The physical constraint relationships in the perception enhancement model are embedded into the identification process. Physical constraint optimization is performed on the filtered perception response features and the initial inversion parameters to obtain identification features that satisfy the physical mechanism constraints, including: Electromagnetic scattering constraint relationship is constructed based on the geometric scattering characteristic model, passive reflection constraint relationship is constructed based on the retroreflective material response model, active near-infrared propagation and coding consistency constraint relationship is constructed based on the active near-infrared coding response model, and active-passive composite detection constraint relationship is constructed based on the active-passive composite detection control equation. A joint loss function is constructed based on the electromagnetic scattering constraint relationship, the passive reflection constraint relationship, the active near-infrared propagation and coding consistency constraint relationship, and the active-passive composite detection constraint relationship. Based on the joint loss function, the identified features are obtained by iteratively solving using the filtered perceived response features and the initial inversion parameters as input.

5. The method for developing and applying high-sensitivity obstacle avoidance cones according to claim 3, characterized in that, Based on the multimodal sensing gain matrix, the fusion weights of each sensor mode are determined. The identified features, the coded near-infrared signal features, and the multi-source sensing data are then fused and identified to obtain initial identification results, model parameters to be calibrated, and fusion identification parameters to be calibrated, including: Based on the time stamp and spatial observation information in the multi-source sensing data, the identification features and the coded near-infrared signal features are time-aligned and spatially correlated to obtain multi-source correlated features. Based on the multimodal sensing gain matrix, the fusion weights corresponding to the millimeter-wave radar mode, lidar mode, camera mode, and near-infrared receiving channel mode are determined, and each fusion weight and its update parameters are used as the initial fusion identification parameters. Based on the initial fusion identification parameters, the identification features, the coded near-infrared signal features, and the multi-source sensing data are fused to obtain the fusion identification features. Based on the fusion identification features, the spatial position, category attributes and temporal changes of the obstacle avoidance cone are jointly calculated to obtain initial identification results and model parameters to be calibrated, including initial position parameters, initial category parameters and initial motion state parameters. The initial fusion identification parameters are then used as fusion identification parameters to be calibrated.

6. The method for developing and applying high-sensitivity obstacle avoidance cones according to claim 1, characterized in that, Based on the obstacle avoidance perception database and the initial identification results, the model parameters to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain a combination of calibrated model parameters, calibrated fusion identification parameters, and perception enhancement parameters, including: Based on the obstacle avoidance perception database and the pre-calibrated spatial position, category attributes and motion state parameters of the obstacle avoidance cones, the perception performance of the initial identification results is evaluated to obtain the position identification error, category identification accuracy, motion state identification error and perception gain evaluation results of each sensor mode of the obstacle avoidance cones. Based on the perception gain evaluation results, the position identification error, the category identification accuracy, and the motion state identification error, the model parameters to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters, and the combination of perception enhancement parameters.

7. A high-sensitivity obstacle avoidance cone research and application system, characterized in that, include: The acquisition module is used to acquire the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones, and to construct a perception enhancement model and a multimodal perception gain matrix through joint modeling; the perception enhancement model and the multimodal perception gain matrix are used as prior information of the perception model; The identification module is used to perform joint inversion and multimodal fusion identification of the multi-source perception data of the obstacle avoidance cone under physical constraints based on the prior information of the perception model and using a pre-built data-object dual-drive multi-parameter intelligent identification algorithm to obtain the initial identification result, the model parameters to be calibrated, and the fusion identification parameters to be calibrated. The calibration module is used to construct an obstacle avoidance integrated test platform based on the parameters of the model to be calibrated and the fusion identification parameters to be calibrated, and to use the obstacle avoidance integrated test platform to conduct multi-condition static and dynamic tests on the obstacle avoidance cone to obtain an obstacle avoidance perception database. Based on the obstacle avoidance perception database and the initial identification results, the model parameters to be calibrated and the fusion identification parameters to be calibrated are jointly calibrated to obtain the calibrated model parameters, the calibrated fusion identification parameters and the combination of perception enhancement parameters. The construction module is used to construct a perception-enhanced obstacle avoidance cone device based on the combination of the perception enhancement parameters. The verification module is used to verify the perception-enhanced obstacle avoidance cone equipment using the calibrated model parameters and the calibrated fusion identification parameters, and to determine the obstacle avoidance warning threshold based on the verification results, thereby establishing an obstacle avoidance safety decision-making and engineering application system.

8. The high-sensitivity obstacle avoidance cone research and application system according to claim 7, characterized in that, The acquisition module acquires the physical attribute parameters and environmental configuration parameters of the obstacle avoidance cones, and constructs a perception enhancement model and a multimodal perception gain matrix through joint modeling, including: Obtain the physical property parameters and environmental configuration parameters of the obstacle avoidance cone; Based on the physical property parameters and the environmental configuration parameters, a geometric scattering characteristic model, an antireflective material response model, and an active near-infrared coded response model are constructed through joint modeling. Based on the retroreflective material response model, the active near-infrared coded response model, and the environmental configuration parameters, a control equation for active-passive composite detection is constructed. The geometric scattering feature model, the retroreflective material response model, the active near-infrared coded response model, and the active-passive composite detection control equation are used as the sensing enhancement model; Based on the aforementioned perception enhancement model, the perception gain parameters corresponding to each sensor mode are calculated using preset evaluation indicators to construct a multimodal perception gain matrix.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a computer program stored in a memory, it implements the high-sensitivity obstacle avoidance cone research and application method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the high-sensitivity obstacle avoidance cone research and application method according to any one of claims 1-6.

Images