A method and system for adaptive detection and multi-frame obstacle confirmation using dual-frequency forward-looking sonar
By combining low-frequency wide-area search with high-frequency local verification in a closed-loop detection mechanism, the problems of resource waste and false alarms in obstacle avoidance of underwater vehicles using dual-frequency forward-looking sonar have been solved, achieving more efficient and reliable obstacle confirmation.
Patent Information
- Application Number
- CN202611316513.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2046-08-28
AI Technical Summary
Existing dual-frequency forward-looking sonar lacks a risk-driven linkage mechanism between low-frequency wide-area search and high-frequency fine verification in obstacle avoidance of underwater vehicles. Single-frame acoustic images are prone to false alarms, dual-frequency acoustic images are difficult to fuse directly, launch control and detection results lack a closed loop, and safety launch control is insufficient, resulting in resource waste and potential safety risks.
Candidate obstacles are obtained by low-frequency wide-area search, local re-detection areas are generated based on risk parameters, local verification is performed by high frequency, and obstacles are confirmed by dual-frequency coordinate registration and multi-frame stability. A closed-loop detection mechanism is established, and detection parameters are adjusted to optimize resource utilization and reduce false alarms.
Reduce false alarm rate, improve detection resource utilization efficiency, enhance obstacle avoidance real-time performance and reliability, reduce invalid scans, and ensure system security.
Smart Images

Figure CN122860425B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of underwater acoustic detection, obstacle avoidance for underwater vehicles, forward-looking sonar imaging, active sonar adaptive detection control, and underwater intelligent sensing. Specifically, it relates to a method and system that utilizes a first operating frequency for wide-area search, generates local re-detection areas based on candidate obstacle risks, performs local verification using a second operating frequency, and combines dual-frequency coordinate registration and multi-frame stability for obstacle confirmation. It can be applied to forward-looking obstacle avoidance, close-range situational awareness, and safety alarms for underwater platforms such as UUVs, AUVs, ROVs, manned submersibles, underwater work platforms, and seabed mining equipment. It can also be used for obstacle detection in front of fixed or mobile underwater work equipment. Background Technology
[0002] When underwater vehicles navigate autonomously, berth, traverse complex terrain, approach seabed targets, and conduct seabed mining operations, they need to acquire real-time information on obstacles, seabed undulations, suspended objects, structures, and other platforms within a certain angle range ahead. Since underwater optical and electromagnetic methods are significantly affected by distance, turbidity, and attenuation, active forward-looking sonar is a crucial sensor for underwater vehicles to achieve obstacle avoidance and situational awareness at both near and long ranges.
[0003] Common forward-looking sonars transmit acoustic signals forward through a transducer array. The receiving array collects the echoes and performs matched filtering, beamforming, imaging, and target detection to form a range-azimuth acoustic map. The system then outputs the obstacle location and alarm results to the host computer or the underwater vehicle control system.
[0004] 2. Current operating methods of dual-frequency forward-looking sonar
[0005] To balance detection range and resolution, some forward-looking sonars employ dual-frequency or multi-frequency operation. Generally, lower operating frequencies result in less propagation loss, longer detection range, and wider coverage, making them suitable for wide-area searches; higher operating frequencies have shorter wavelengths and higher range and angular resolution, making them suitable for close-range detailed observation or target contour verification.
[0006] Common operating modes of existing dual-frequency sonars include fixed-frequency operation, alternating operation at fixed periods, manual frequency switching, or operation according to a fixed transmission sequence after parameters are set by the host computer. While these methods are simple to implement, they typically do not fully utilize the detection results of the previous frame for real-time feedback control of the transmission strategy for the next frame.
[0007] 3. Shortcomings of existing technologies
[0008] 1) There is a lack of risk-driven linkage mechanism between low-frequency wide-area search and high-frequency fine verification. Existing solutions usually involve fixed alternating transmission or manual frequency switching. High-frequency sonar may perform ineffective scans of areas without targets, resulting in a waste of transmission energy, processing power, and communication bandwidth.
[0009] 2) Single-frame sonar images have a high rate of false alarms. Underwater reverberation, speckle, seabed undulations, shadows, bubbles, suspended objects, and sidelobe interference can all create bright spots or edges resembling targets in the acoustic image, making single-frame threshold detection prone to false alarms.
[0010] 3) Dual-frequency acoustic maps are not easily fused directly. Different operating frequencies correspond to different beamwidths, distance sampling, angular resolutions, propagation losses, acoustic map scales, and effective fields of view. Simple superposition or manual comparison makes it difficult to reliably identify the same obstacle.
[0011] 4) Lack of a closed loop between transmission control and detection results. Existing methods often operate continuously with preset parameters and cannot dynamically reduce the detection sector, change the beam density, adjust the transmission pulse width, or insert verification Ping according to the obstacle risk level.
[0012] 5) Insufficient safety control during transmission. Active sonar has a high transmission power. If the system automatically continues to transmit using the old parameters after power-on or continues to transmit after a communication link is lost, it may lead to unnecessary safety risks and energy consumption.
[0013] Therefore, it is necessary to propose a dual-frequency forward-looking sonar adaptive detection method based on risk area re-detection, so that low-frequency wide-area search, high-frequency local verification, dual-frequency fusion, multi-frame confirmation and next cycle parameter adjustment form a closed loop, thereby reducing false alarms, reducing invalid scans and improving the real-time performance and reliability of underwater obstacle avoidance. Summary of the Invention
[0014] The purpose of this invention is to provide a dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method and system for obstacle avoidance of underwater vehicles, so as to solve the problems of fixed transmission, indiscriminate scanning, many false alarms in a single frame, difficulty in dual-frequency fusion, and low efficiency of detection resource utilization of existing dual-frequency forward-looking sonar.
[0015] The core idea of this invention is as follows: a wide-area search is performed at a first operating frequency to obtain candidate obstacles; risk parameters are calculated based on the distance, azimuth, echo intensity, confidence level, continuous frame stability, and relative motion trend of the candidate obstacles; when the risk parameters meet the triggering conditions, one or more risk regions are generated, and the risk regions are locally re-detected at a second operating frequency; then the dual-frequency results are mapped to a unified range-azimuth coordinate system, and obstacles are confirmed by combining dual-frequency spatial consistency, multi-frame continuity, and relative motion trend; finally, the confirmation results are fed back to the task scheduling module to adjust the operating frequency, detection sector, beam density, pulse width, ping rate, and reception time window of subsequent pings.
[0016] The main highlights of this invention include:
[0017] 1) A closed-loop detection mechanism is proposed, consisting of "low-frequency wide-area search - high-frequency risk area re-detection - dual-frequency fusion - multi-frame confirmation - next Ping parameter feedback".
[0018] 2) High-frequency verification is not a fixed full-field scan, but automatically generates local re-detection areas based on low-frequency candidate obstacles, reducing invalid emission and invalid beamforming.
[0019] 3) Risk parameters are determined not only by the intensity of a single frame echo, but also by factors such as distance, azimuth, confidence level, number of consecutive frames, relative motion trend, and relationship with the safety envelope of the aircraft.
[0020] 4) Establish a dual-frequency distance-azimuth coordinate registration mechanism so that acoustic maps with different frequencies, beamwidths and sampling scales can be associated with targets under a unified coordinate system.
[0021] 5) Obstacles are identified through dual-frequency consistency and multi-frame stability, reducing false alarms caused by reverberation, speckle, seabed undulation and transient interference.
[0022] 6) The obstacle confirmation results are used in reverse to plan the next cycle of detection missions, so as to achieve adaptive adjustment of Ping rate, pulse width, sector range, beam density and receiving time window.
[0023] To achieve the above objectives, the present invention adopts the following technical solution:
[0024] A dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method includes the following steps:
[0025] Step S1: The system is powered on and parameters are verified. The sonar enters a safe standby state, waiting for and verifying the working parameters and transmission authorization.
[0026] Step S2: When in the allowed transmission state, perform a wide-area search at the first operating frequency F1 to obtain the F1 range-azimuth acoustic map of the forward field of view;
[0027] Step S3: Perform candidate obstacle detection on the F1 distance-azimuth acoustic map to obtain a set of candidate regions, and extract feature information of each candidate region;
[0028] Step S4: Calculate the risk parameters for each candidate region based on the feature information;
[0029] Step S5: When the risk parameter exceeds the first threshold, generate one or more local re-detection regions based on the corresponding candidate regions;
[0030] Step S6: Perform local re-detection on the local re-detection area at the second operating frequency F2 to obtain the F2 range-azimuth acoustic map;
[0031] Step S7: Map the F1 distance-azimuth acoustic map and the F2 distance-azimuth acoustic map to a unified distance-azimuth coordinate system, and perform dual-frequency target association and consistency judgment;
[0032] Step S8: Based on the consistency judgment result, perform multi-frame stability analysis by combining the multi-frame state records of candidate obstacles, confirm stable obstacles and output their information and risk level;
[0033] Step S9: Based on the confirmation result of the stable obstacle, adaptively adjust the detection parameters of subsequent detection cycles.
[0034] Furthermore, in step S4, the calculation of the risk parameter comprehensively considers one or more of the following factors: detection confidence, distance risk, orientation risk, relative motion trend, number of consecutive frames, target size, degree of overlap with the vehicle's safety envelope, and dual-frequency historical consistency.
[0035] Furthermore, in step S5, the local re-detection area is defined by an azimuth range and a distance range, which are jointly determined based on the beamwidth of the first operating frequency F1, the size of the candidate target, the positioning error, the platform motion error, the sound speed error, and the safety margin.
[0036] Furthermore, in step S6, the local redetection only covers the local redetection area, rather than scanning the entire forward field of view; and the beam density, pulse width, reception time window, and repetition count of the second operating frequency F2 are adjusted according to the magnitude of the risk parameter.
[0037] Furthermore, in step S7, the dual-frequency target association and consistency judgment includes: calculating the spatial overlap, center distance deviation, center orientation deviation, contour similarity, intensity ratio, shadow consistency, and confidence consistency of the F1 candidate target and the F2 verification target under a unified coordinate system; when the preset spatial consistency condition is met, the confirmation level of the candidate target is increased; otherwise, its confirmation level is reduced or it is marked as a target to be observed.
[0038] Furthermore, in step S9, the detection parameters include one or more of the following: operating frequency, detection sector, beam density, pulse width, ping rate, and reception time window. If there is no target ahead or the risk is below a preset threshold, a low-density wide-area search is performed using the first operating frequency F1. If a low-risk target exists at a distance, the search refresh rate of the first operating frequency F1 is increased or the local beam spacing is reduced. If a medium- to high-risk target exists, a local verification ping is inserted using the second operating frequency F2, and the beam density of the local re-detection area is increased. If the target enters the safe distance threshold, the local ping rate is increased, the range is shortened, and the reception time window is adjusted.
[0039] Furthermore, in step S1, launch authorization and security state machine management are also performed; the security state machine includes at least the following states: security standby, parameter valid, launch allowed, launch executed, launch prohibited, link break protection, and fault protection; launch is only allowed when parameter verification is passed, launch authorization is valid, and the system status is normal.
[0040] Furthermore, when a communication link is lost, parameter timeout occurs, parameter verification error occurs, power amplifier malfunction occurs, power supply voltage malfunction occurs, transducer protection status occurs, temperature exceeds the limit, or the host computer revokes authorization, the safety state machine revokes the transmission permission signal, stops transmission, and records the log; the system will not automatically resume transmission until it obtains valid parameters and transmission authorization again.
[0041] This invention also discloses a dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation system for implementing the method, comprising:
[0042] A dual-frequency forward-looking transducer array is used to transmit acoustic signals at a first operating frequency F1 and a second operating frequency F2 and to receive echo signals.
[0043] Dual-frequency transmission drive unit, used to generate drive signals according to the transmission mission;
[0044] A multi-channel receiving and acquisition unit is used to amplify, filter, control the gain, and convert the echo signal to digital.
[0045] The FPGA preprocessing unit is used for real-time preprocessing of multi-channel digital data;
[0046] The GPU beamforming and fusion confirmation unit is used for beamforming, candidate target detection, risk parameter calculation, dual-frequency coordinate registration, dual-frequency target association, multi-frame stability analysis and obstacle confirmation.
[0047] The ARM main control and task scheduling unit is used to communicate with the host computer, parse parameters, schedule detection tasks, and adaptively adjust subsequent detection parameters based on obstacle confirmation results.
[0048] Launch authorization and security state machine, used to constrain launch behavior for security purposes;
[0049] The dual-frequency coordinate registration module is used to map acoustic maps of different frequencies to a unified distance-azimuth coordinate system.
[0050] The risk area generation and re-detection planning module is used to generate local re-detection areas and determine local verification parameters based on the risk parameters of candidate obstacles.
[0051] Furthermore, the first operating frequency F1 is 450kHz, used for wide-area search; the second operating frequency F2 is 750kHz, used for high-resolution verification of the local re-detection area; the system is integrated into an underwater vehicle, underwater work platform or seabed mining equipment, and connected to a host computer or vehicle control system via Ethernet, serial port, fiber optic or CAN interface.
[0052] Compared with the prior art, the present invention has the following beneficial effects:
[0053] 1) Reduce false alarm rate: By combining F1 candidate detection, F2 local verification, dual-frequency spatial consistency and multi-frame stability judgment, false alarms caused by reverberation, speckle, seabed undulation and transient interference are reduced.
[0054] 2) Improve the efficiency of detection resource utilization: The F2 does not perform indiscriminate scanning of the entire field of view, but only performs local re-detection of risk areas, which can reduce invalid transmission, invalid reception time windows, invalid beamforming and data transmission volume.
[0055] 3) Improve close-range confirmation capability: Using higher frequencies and higher beam densities in risk areas can obtain more detailed local acoustic images, which is beneficial for confirming obstacle boundaries, locations and stability.
[0056] 4) Balancing long-range search and short-range obstacle avoidance: F1 maintains wide-area search capability, while F2 provides local fine-grained verification capability. The two work together dynamically through a closed-loop task scheduling mechanism.
[0057] 5) Improve real-time performance: The system automatically adjusts the range, receiving time window and beam density according to the current risk, and can concentrate computing resources on key areas in high-risk close-range scenarios.
[0058] 6) Improve system security: Launch authorization, parameter verification, link break protection, and abnormal launch stop mechanisms can prevent sonar from launching automatically in unauthorized or abnormal states.
[0059] 7) Adaptable to different hardware implementations: This solution can be implemented in FPGA+ARM+GPU architecture, as well as on DSP, CPU, SoC or other heterogeneous computing platforms, and has good engineering adaptability. Attached Figure Description
[0060] Figure 1 This is a schematic diagram of the components of a dual-frequency forward-looking sonar adaptive detection system;
[0061] Figure 2 The flowchart shows the overall methodology for risk area re-detection.
[0062] Figure 3This diagram illustrates the wide-area search sector and local re-detection in the risk area; the rectangle at the bottom represents the sonar / UUV; the entire sector represents the F1 wide-area search area; the diagonal striped area represents the F2 risk area, i.e., the local re-detection area; the hollow circles in the diagonal stripes represent candidate obstacles.
[0063] Figure 4 Diagram showing the transition between the exploration mission and safe launch status;
[0064] Figure 5 This is a schematic diagram of dual-frequency coordinate registration and fusion processing;
[0065] Figure 6 This is a schematic diagram illustrating the confirmation of stable obstacles across multiple frames.
[0066] Figure 7 This is a flowchart illustrating the data and control flow between FPGA, ARM, and GPU.
[0067] Figure 8 This is a schematic diagram of the Ping sequence for search, verification, fusion, and feedback. Detailed Implementation
[0068] The specific embodiments of the present invention will be described in detail below. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.
[0069] like Figure 1 and Figure 7 As shown, the present invention provides a dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation system, comprising a dual-frequency forward-looking transducer array 101, a dual-frequency transmission drive unit 102, a multi-channel receiving and acquisition unit 103, an FPGA preprocessing unit 104, an ARM main control and task scheduling unit 105, a GPU beamforming and fusion confirmation unit 106, a host computer or UUV control interface 107, a transmission authorization and safety state machine 108, a dual-frequency coordinate registration module 109, and a risk area generation and re-detection planning module 110. These modules can be integrated into a single sonar device, or distributed among the sonar electronics compartment, FPGA / ARM control board, GPU computing unit, and host computer according to the hardware architecture of the underwater platform.
[0070] The dual-frequency forward-looking transducer array 101 is used to transmit acoustic signals at a first operating frequency F1 and a second operating frequency F2 to a predetermined sector in front of the sonar, and to receive echo signals from obstacles, the seabed, and water scatterers. The first operating frequency F1 can be selected as a lower frequency for wide-area searching over a larger area; the second operating frequency F2 can be selected as a frequency higher than F1 for local high-resolution re-detection of risk areas.
[0071] In one embodiment, F1 is a frequency around 450 kHz, and F2 is a frequency around 750 kHz; F1 is used for searching a forward-looking sector of approximately 130°, and F2 is used for verification of local sectors generated based on the risk area. However, the above frequencies, bandwidths, and sector ranges are only examples, and in actual applications, they can be adjusted according to the transducer, platform speed, range, and target type.
[0072] The dual-frequency transmit drive unit 102 generates drive signals for the corresponding operating frequency, pulse width, bandwidth, amplitude, array element delay, and transmit sector based on the transmit task issued by the ARM main control and task scheduling unit 105. This unit may include a power drive circuit, a transmit switch, a transmit / receive isolation or protection circuit, a power amplifier power supply, and a status acquisition circuit.
[0073] The dual-frequency transmission drive unit 102 receives the transmission permission signal output by the transmission authorization and security state machine 108. It will only perform transmission when the parameter verification is successful, the communication handshake is effective, the power amplifier is in normal status, the transducer is in normal status, and the system is in a transmission permission state.
[0074] The multi-channel receiving and acquisition unit 103 is used to amplify, filter, control the gain, and perform analog-to-digital conversion on the echo received by the array. This unit can receive TVG parameters, receiving time window parameters, and sampling control parameters issued by the ARM master control and task scheduling unit 105. The acquired multi-channel digital data is sent to the FPGA preprocessing unit 104.
[0075] The FPGA preprocessing unit 104 is used to perform real-time preprocessing on multi-channel received data, including but not limited to digital downconversion, decimation, channel amplitude and phase correction, matched filtering, pulse compression, data truncation, buffering, timestamp marking, channel number marking, and data packing. The FPGA preprocessing results are sent to the GPU beamforming and fusion confirmation unit 106 via a high-speed data interface.
[0076] The FPGA preprocessing unit 104 can also add metadata to the data packet according to the frequency type, transmission sequence number, receiving window, range and re-detection area flag of the current Ping, so that the GPU beamforming and fusion confirmation unit 106 can distinguish between wide-area search Ping and risk area verification Ping.
[0077] The ARM main control and task scheduling unit 105 is responsible for communicating with the host computer or UUV control interface 107 to complete parameter reception, parameter parsing, CRC verification, version number management, handshake, status reporting, log recording, and detection task scheduling. Based on the candidate target, risk parameters, and obstacle confirmation results output by the GPU beamforming and fusion confirmation unit 106, this unit generates the detection parameters for the next Ping.
[0078] The ARM main control and task scheduling unit 105 is connected to the transmission authorization and safety state machine 108. After the sonar is powered on, if no new valid transmission parameters and transmission permission commands are received, it remains in a safe standby state and does not automatically resume the previous transmission task. If the communication link is lost, parameters time out, status is abnormal, or the host computer revokes the transmission authorization, the transmission stops and enters a protection state.
[0079] The GPU beamforming and fusion confirmation unit 106 is used to perform beamforming on the multi-channel data preprocessed by the FPGA preprocessing unit 104 to generate distance-azimuth acoustic maps corresponding to the first and second operating frequencies; and to perform target candidate detection, risk parameter calculation, dual-frequency coordinate registration, dual-frequency target association, multi-frame stability analysis, and obstacle confirmation on the acoustic maps.
[0080] The GPU beamforming and fusion verification unit 106 may include a wide-area search processing module, a candidate target detection module, a risk scoring module, a local verification processing module, a dual-frequency fusion module, a multi-frame verification module, and an alarm output module. These modules may be implemented using rule-based algorithms, statistical detection algorithms, machine learning algorithms, or combinations thereof.
[0081] The host computer or UUV control interface 107 is used to send operating modes, ranges, frequency strategies, detection sectors, safety distances, risk thresholds, transmission authorizations, TVG parameters, and other control parameters to the sonar, and to receive obstacle locations, risk levels, acoustic image data, system status, and log information output by the sonar. This interface can be Ethernet, serial port, fiber optic, CAN, synchronous trigger line, or other commonly used interfaces on underwater platforms.
[0082] The launch authorization and safety state machine 108 is used to impose safety constraints on the active sonar launch behavior. This state machine includes at least the following states: safe standby, parameter valid, launch allowed, launch executed, launch prohibited, link loss protection, and fault protection.
[0083] In a preferred embodiment, the following conditions must be met before transmission: the CRC check of the parameters issued by the host computer or UUV control interface 107 is passed; the parameter version number or timestamp is valid; the transmission authorization token is valid; the power amplifier, power supply and transducer are in normal status; the receiving or transmitting channel is not in a protected state; and the current Ping interval meets the constraints of sound propagation round-trip time and processing time.
[0084] The dual-frequency coordinate registration module 109 is used to map the acoustic maps of the first and second operating frequencies to a unified distance-azimuth coordinate system. This module can establish frequency-related mapping relationships based on the sound velocity settings, sampling rate, distance gate, array geometry parameters, channel amplitude and phase correction parameters, beam pointing, beamwidth, and installation calibration parameters corresponding to each frequency.
[0085] Under a unified coordinate system, the system can calculate the spatial overlap, center position deviation, distance deviation, azimuth deviation, intensity ratio, shadow relationship and confidence consistency of dual-frequency targets, thereby determining whether the low-frequency candidate target has been verified and confirmed by the high-frequency target.
[0086] The risk area generation and re-detection planning module 110 generates one or more re-detection areas based on the risk parameters of candidate obstacles and determines the local verification parameters for the second operating frequency. The local verification parameters include, but are not limited to, the center azimuth, azimuth width, distance range, beam density, receive time window, transmit pulse width, bandwidth, Ping insertion timing, and repetition count of the verification sector.
[0087] The functions and connections of each module can be summarized as shown in Table 1.
[0088] Table 1: Functions and Connections of Each Module
[0089] label Module Name Main functions Connection 101 Dual-frequency forward-looking transducer array Transmit F1 / F2 acoustic signals and receive echoes Connected to dual-frequency transmission drive unit 102 and multi-channel receiving and acquisition unit 103 102 Dual-frequency transmission drive unit Execution frequency, pulse width, array element delay, and sector corresponding launch mission Controlled by the ARM master control and task scheduling unit 105 and the transmit authorization and security state machine 108, the dual-frequency forward-looking transducer array 101 is driven. 103 Multi-channel receiving and acquisition unit Echo amplification, filtering, TVG, ADC acquisition The input comes from the dual-frequency forward-looking transducer array 101, and the output goes to the FPGA preprocessing unit 104. 104 FPGA preprocessing unit DDC, decimation, amplitude and phase correction, matched filtering, packetization Receive the acquired data and output it to the GPU beamforming and fusion confirmation unit 106 105 ARM main control and task scheduling unit Parameter parsing, handshake, logging, task scheduling 107 Connects to host computer or UUV control interface; 108 Transmission authorization and security state machine; 102 Dual-frequency transmission drive unit; 104 FPGA preprocessing unit; 106 GPU beamforming and fusion confirmation unit. 106 GPU beamforming and fusion confirmation unit 106 Beamforming, candidate detection, risk scoring, dual-frequency fusion, multi-frame confirmation Receives data from FPGA preprocessing unit 104, and outputs candidate targets and confirmation results to ARM / host computer. 107 Host computer or UUV control interface 107 Send parameters, receive audio and video signals and alarms. Connecting the ARM main controller and the task scheduling unit 105 108 Launch authorization and security state machine 108 Determine whether to allow launch; prohibit launch in case of link failure or malfunction. Receives status feedback from the ARM main controller and task scheduling unit 105, and controls the dual-frequency transmission drive unit 102. 109 Dual-frequency coordinate registration module 109 Map the F1 / F2 sound maps to a unified coordinate system Receive dual-frequency acoustic images and output registration target data. 110 Risk Area Generation and Re-detection Planning Module 110 Generate ROI and next Ping verification parameters The risk results are received and fed back to the ARM main control and task scheduling unit 105.
[0090] like Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 8 As shown, the present invention provides a dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method, comprising the following steps:
[0091] Step S1: System power-on and parameter verification
[0092] After power-on, the sonar enters a safe standby state. The ARM main control and task scheduling unit 105 waits for the host computer or UUV control interface 107 to send operating parameters and transmission authorization. The system verifies the frequency, bandwidth, pulse width, range, detection sector, TVG, transmission voltage, ping rate, reception time window, risk threshold, CRC checksum, parameter version number, and authorization token. If the parameters fail verification or transmission authorization is not obtained, the transmission authorization and safety state machine 108 maintains a transmission-prohibited state.
[0093] Step S2: Wide-area search at the first operating frequency F1
[0094] When the system is in a transmit-allowed state, the dual-frequency transmit drive unit 102 performs a wide-area search ping at the first operating frequency F1. F1 is typically selected as a frequency with low propagation loss, suitable for a large range and a large field of view search. The multi-channel receive acquisition unit 103 acquires multi-channel echoes within a preset receive time window, the FPGA preprocessing unit 104 completes preprocessing, and the GPU beamforming and fusion confirmation unit 106 performs beamforming to obtain the F1 range-azimuth acoustic map.
[0095] Step S3: Candidate obstacle detection and feature extraction
[0096] The GPU beamforming and fusion confirmation unit 106 performs candidate target detection on the F1 distance-azimuth acoustic map, obtaining a candidate region set Ω={O1,O2,...,On}. Each candidate region Oi includes at least the center distance ri, center azimuth θi, echo intensity Ii, region area Ai, signal-to-noise ratio SNRi, detection confidence Ci, target boundary or contour, shadow features, and correlation results with the target in the previous frame.
[0097] Step S4: Calculation of risk parameters
[0098] For each candidate region, a risk parameter Ri is calculated. This risk parameter can be obtained by comprehensively considering factors such as detection confidence, distance risk, orientation risk, relative motion trend, number of consecutive occurrences, target size, safety envelope overlap, and dual-frequency historical consistency. For example, Ri = w1·Ci + w2·Dr(ri) + w3·Dθ(θi) + w4·Mv(i) + w5·Nh(i) + w6·Ah(i), where w1 to w6 are weights, Dr represents the function indicating higher risk with closer distance, Dθ represents the function indicating higher risk with closer proximity to the vehicle's direction of travel, Mv represents the relative proximity trend, Nh represents the number of consecutive occurrences, and Ah represents the target size or safety envelope overlap. The above formula is only an example; lookup tables, rule trees, fuzzy inference, or neural network scoring methods can also be used.
[0099] Step S5: Risk Area Generation
[0100] When the risk parameter Ri of a candidate region Oi exceeds the first threshold T1, or although it is lower than T1 but appears stably in multiple consecutive frames, the risk region generation and re-detection planning module 110 generates a local re-detection region ROIi based on the candidate region. ROIi can be composed of an azimuth range [θi-Δθi, θi+Δθi] and a range range [ri-Δri, ri+Δri], where Δθi and Δri are jointly determined by the F1 beamwidth, target size, positioning error, platform motion error, sound velocity error, and safety margin.
[0101] Step S6: Local re-detection at the second operating frequency F2
[0102] For the generated ROIi, the system inserts a local verification task at a second operating frequency F2 in one or more subsequent Pings. F2 is typically higher than F1 to obtain a higher resolution local acoustic map. The F2 verification task covers only the ROIi or several risk areas, rather than a fixed scan of the entire forward-looking field of view. The system can adjust the beam density, pulse width, reception time window, and number of repetitions of the F2 verification according to the Ri size.
[0103] Step S7: Map dual-frequency data to a unified distance-azimuth coordinate system
[0104] The dual-frequency coordinate registration module 109 maps the acoustic maps of F1 and F2 to a unified distance-azimuth coordinate grid based on the array calibration parameters, sampling rate, range gate, beam pointing, beamwidth, amplitude and phase correction parameters, and installation parameters corresponding to F1 and F2. For any frequency Fk, the mapping Mk:(rk,θk,Ik)→(r,θ,Ik') can be used, where (r,θ) are the unified coordinates, and Ik' is the intensity or characteristic after propagation loss, TVG, and frequency response correction.
[0105] Step S8: Dual-frequency target association and consistency judgment
[0106] Under a unified coordinate system, the system associates candidate targets F1 with verification targets F2, calculating spatial overlap, center distance deviation, center orientation deviation, contour similarity, intensity ratio, shadow consistency, and confidence consistency. When F2 detects a target within its corresponding ROI and satisfies the spatial consistency condition with candidate targets F1, the confirmation level of that candidate target is increased; when F2 does not detect a target or the results are inconsistent, the confirmation level is decreased or the target is retained as a target to be observed.
[0107] Step S9: Multi-frame stability analysis
[0108] The system establishes multi-frame state records for candidate obstacles, including target ID, historical distance, historical orientation, confidence sequence, number of dual-frequency confirmations, number of consecutive appearances, number of disappearances, and relative motion trend. The system can employ trajectory association, Kalman filtering, sliding window statistics, nearest neighbor association, Hungarian matching, particle filtering, or neural network tracking methods to analyze target continuity and stability.
[0109] Step S10: Stabilizing Obstacle Confirmation and Risk Level Output
[0110] When a candidate target meets the stable occurrence condition within N consecutive frames, or meets spatial consistency and shows a relative proximity trend in both F1 and F2, the system identifies it as a stable obstacle and outputs the obstacle's distance, orientation, risk level, relative motion trend, and alarm information. When a candidate target appears only in a single frame, has inconsistent dual frequencies, or exhibits unreasonable motion patterns, the system classifies it as transient interference, a low-confidence target, or a target awaiting verification.
[0111] Step S11: Adaptive adjustment of the next Ping probe parameters
[0112] Based on the obstacle confirmation results, the ARM main control and task scheduling unit 105 generates the next Ping or several subsequent Ping detection tasks. If there is no target ahead or the risk is low, the system can use F1 low-density wide-area search; if there is a distant low-risk target, the system can increase the F1 search refresh rate or reduce the local beam spacing; if there is a medium-to-high-risk target, the system inserts F2 local verification Ping and increases the ROI beam density; if the target enters the safe distance threshold, the system can increase the local Ping rate, shorten the range and the receiving time window, and output an alarm to the UUV.
[0113] Step S12: Ping rate and pulse width constraints
[0114] When adjusting the ping rate and pulse width, the system needs to meet constraints on sound propagation round-trip time, reception time window, processing delay, and safety margin. The minimum ping interval can be determined using the form Tping ≥ 2Rmax / c + Tproc + Tmargin, where Rmax is the current detection range, c is the speed of sound, Tproc is the processing delay, and Tmargin is the safety margin. The pulse width can be adjusted according to the range, resolution, transmission energy, and target risk level; in high ping rate or close-range verification modes, shorter pulses or smaller reception windows can be used.
[0115] Step S13: Abnormal Status and Launch Protection
[0116] When a communication link is lost, parameter timeout occurs, CRC error occurs, power amplifier malfunctions, power supply voltage malfunctions, transducer protection status occurs, temperature exceeds limits, task scheduling conflict occurs, or the host computer revokes authorization, the transmit authorization and safety state machine 108 revokes the transmit permission signal, the system stops transmitting, and logs the event. The system can continue to receive and upload status updates or wait for new valid parameters, but it must not automatically resume transmission without re-authorization.
[0117] The following examples are used to illustrate specific implementations of the present invention and are not intended to limit the scope of protection of the present invention.
[0118] Example 1: 450 kHz wide-area search and 750 kHz local verification
[0119] In one embodiment of forward-looking obstacle avoidance sonar, the first operating frequency F1 uses a frequency around 450 kHz, mainly used for searching a larger forward-looking sector; the second operating frequency F2 uses a frequency around 750 kHz, mainly used for local high-resolution verification of candidate obstacles detected by F1. The system first searches a sector of approximately 130° ahead using F1 to form a range-azimuth acoustic map. If a candidate obstacle is detected near azimuth θ=20° and distance r=60 m, and appears in two consecutive frames, the risk area generation and re-detection planning module 110 uses this candidate area as the center and generates an F2 verification area, for example, with an azimuth range of 15°, 25°, 15°, 25° and a distance range of [55 m, 65 m], according to the F1 beamwidth, target size, and positioning error.
[0120] The system then inserts an F2 local verification Ping, performing high-density beamforming only on that local area. If a target is detected in the F2 acoustic image near the same distance-azimuth coordinates, and the center deviation and overlap meet the thresholds, the target's confirmation level is increased; if the target continues to appear in several consecutive frames and the distance shows a decreasing trend, a stable obstacle confirmation result and a risk alarm are output.
[0121] Example 2: Priority scheduling of multiple risk areas
[0122] When F1 wide-area search simultaneously discovers multiple candidate regions, the system calculates risk parameters for each and sorts them from highest to lowest risk. If the GPU's processing power, launch time, or ping interval does not allow for reviewing all regions within one cycle, priority is given to reviewing targets that are close, located in the vehicle's direction of travel, have high confidence, and appear consecutively. Low-risk regions can be reviewed later or only tracked by F1.
[0123] Example 3: Handling of Unconfirmed Targets
[0124] If F2 fails to detect a target matching the candidate target of F1 during a re-check, the system does not immediately delete the target, but instead lowers its confirmation level based on historical frame records. If F1 continues to detect the target subsequently, F2 can be triggered again for re-checking; if the target disappears for several consecutive frames, it is determined to be transient interference or background reverberation. This mechanism can avoid missed confirmations due to the influence of attitude, obstruction, or signal-to-noise ratio on a single high-frequency echo.
[0125] Example 4: Integration with the launch safety state machine
[0126] After the sonar is powered off and restarted, the host computer interface can display the parameters used last time, but the sonar does not automatically transmit. Only after the host computer or UUV control interface 107 reissues valid parameters, passes CRC verification, and issues a transmission authorization will the sonar enter the transmission-allowed state. If the communication interruption exceeds the set time, the transmission authorization and safety state machine 108 revokes the transmission permission and stops subsequent Ping tasks. This mechanism is suitable for applications where the active sonar has high power and requires avoiding mis-transmissions.
[0127] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.
Claims
1. A method for adaptive detection and multi-frame obstacle confirmation using dual-frequency forward-looking sonar, characterized in that, Includes the following steps: Step S1: The system is powered on and parameters are verified. The sonar enters a safe standby state, waiting for and verifying the working parameters and transmission authorization. Step S2: When in the allowed transmission state, perform a wide-area search at the first operating frequency F1 to obtain the F1 range-azimuth acoustic map of the forward field of view; Step S3: Perform candidate obstacle detection on the F1 distance-azimuth acoustic map to obtain a set of candidate regions, and extract feature information of each candidate region; Step S4: Calculate the risk parameters for each candidate region based on the feature information; Step S5: When the risk parameter exceeds the first threshold, generate one or more local re-detection regions based on the corresponding candidate regions; Step S6: Perform local re-detection on the local re-detection area at the second operating frequency F2 to obtain the F2 range-azimuth acoustic map; Step S7: Map the F1 distance-azimuth acoustic map and the F2 distance-azimuth acoustic map to a unified distance-azimuth coordinate system, and perform dual-frequency target association and consistency judgment; Step S8: Based on the consistency judgment result, perform multi-frame stability analysis by combining the multi-frame state records of candidate obstacles, confirm stable obstacles and output their information and risk level; Step S9: Based on the confirmation result of the stable obstacle, adaptively adjust the detection parameters of subsequent detection cycles.
2. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S4, the calculation of the risk parameter comprehensively considers one or more of the following factors: detection confidence, distance risk, orientation risk, relative motion trend, number of consecutive frames, target size, degree of overlap with the vehicle's safety envelope, and dual-frequency historical consistency.
3. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S5, the local re-detection area is defined by an azimuth range and a distance range, which are determined jointly based on the beamwidth of the first operating frequency F1, the size of the candidate target, the positioning error, the platform motion error, the sound speed error, and the safety margin.
4. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S6, the local redetection only covers the local redetection area, rather than scanning the entire forward field of view; and the beam density, pulse width, reception time window, and repetition number of the second operating frequency F2 are adjusted according to the magnitude of the risk parameter.
5. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S7, the dual-frequency target association and consistency judgment includes: calculating the spatial overlap, center distance deviation, center orientation deviation, contour similarity, intensity ratio, shadow consistency and confidence consistency of the F1 candidate target and the F2 verification target under a unified coordinate system; when the preset spatial consistency condition is met, the confirmation level of the candidate target is increased; otherwise, its confirmation level is reduced or it is marked as a target to be observed.
6. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S9, the detection parameters include one or more of the following: operating frequency, detection sector, beam density, pulse width, ping rate, and reception time window. If there is no target ahead or the risk is below a preset threshold, a low-density wide-area search is performed using the first operating frequency F1. If a low-risk target exists at a distance, the search refresh rate of the first operating frequency F1 is increased or the local beam spacing is reduced. If a medium- to high-risk target exists, a local verification ping is inserted using the second operating frequency F2, and the beam density of the local re-detection area is increased. If the target enters the safe distance threshold, the local ping rate is increased, the range is shortened, and the reception time window is adjusted.
7. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 1, characterized in that, In step S1, launch authorization and security state machine management are also performed; the security state machine includes at least the following states: security standby, parameter valid, launch allowed, launch executed, launch prohibited, link break protection, and fault protection; launch is only allowed when parameter verification is passed, launch authorization is valid, and the system status is normal.
8. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation method according to claim 7, characterized in that, When a communication link is lost, parameter timeout occurs, parameter verification error occurs, power amplifier malfunction occurs, power supply voltage malfunction occurs, transducer protection status occurs, temperature exceeds the limit, or the host computer revokes the authorization, the safety state machine revokes the transmission permission signal, stops transmission, and records the log; the system will not automatically resume transmission until it obtains valid parameters and transmission authorization again.
9. A dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation system, used to implement the method as described in any one of claims 1 to 8, characterized in that, include: A dual-frequency forward-looking transducer array (101) is used to transmit acoustic signals at a first operating frequency F1 and a second operating frequency F2 and to receive echo signals. A dual-frequency transmission drive unit (102) is used to generate drive signals according to the transmission mission; A multi-channel receiving and acquisition unit (103) is used to amplify, filter, control the gain, and convert the echo signal to an analog-to-digital converter. The FPGA preprocessing unit (104) is used to perform real-time preprocessing of multi-channel digital data; The GPU beamforming and fusion confirmation unit (106) is used to perform beamforming, candidate target detection, risk parameter calculation, dual-frequency coordinate registration, dual-frequency target association, multi-frame stability analysis and obstacle confirmation. The ARM main control and task scheduling unit (105) is used to communicate with the host computer, parse parameters, schedule detection tasks, and adaptively adjust subsequent detection parameters according to the obstacle confirmation results. The launch authorization and security state machine (108) is used to impose security constraints on launch behavior; A dual-frequency coordinate registration module (109) is used to map acoustic maps of different frequencies to a unified distance-azimuth coordinate system; The risk area generation and re-detection planning module (110) is used to generate local re-detection areas and determine local verification parameters based on the risk parameters of candidate obstacles.
10. The dual-frequency forward-looking sonar adaptive detection and multi-frame obstacle confirmation system according to claim 9, characterized in that, The first operating frequency F1 is 450kHz, used for wide-area search; the second operating frequency F2 is 750kHz, used for high-resolution verification of the local re-detection area; the system is integrated into an underwater vehicle, underwater work platform or seabed mining equipment, and connected to a host computer or vehicle control system via Ethernet, serial port, fiber optic or CAN interface.
Citation Information
Patent Citations
Room characterization and correction for multi-channel audio
CN103621110A
Method for detecting obstacle in front of AUV (Autonomous Underwater Vehicle) for near-bottom navigation
CN122410539A
