A small object collision avoidance system for electric vertical takeoff and landing aircraft

Abstract

This invention discloses a collision avoidance system for small flying objects used in electric vertical takeoff and landing (EVTOL) aircraft, belonging to the field of aircraft collision avoidance technology. The system includes an infrared detection module, a target identification and threat assessment module, a collision risk calculation module, an auxiliary ranging module, a fusion correction module, and an avoidance decision module. The infrared detection module outputs target infrared detection data; the target identification and threat assessment module identifies the target type and generates an initial threat level; the collision risk calculation module generates target collision risk data using a corresponding calculation strategy; the auxiliary ranging module acquires target distance-related data when preset trigger conditions are met; the fusion correction module corrects the target collision risk data; and the avoidance decision module generates warning information or avoidance control commands. This system can achieve graded detection, risk calculation, and avoidance control of small flying objects.

Description

Technical Field

[0001] This invention relates to the field of aircraft collision avoidance technology, and in particular to a small-scale object collision avoidance system for electric vertical take-off and landing aircraft. Background Technology

[0002] With the development of the low-altitude economy, electric vertical takeoff and landing (EVTOL) aircraft, as an important vehicle for urban air transportation, are gradually being applied to low-altitude commuting, logistics transportation, and emergency rescue scenarios. Urban low-altitude airspace is characterized by a variety of aircraft types, complex flight paths, and the potential presence of small flying objects such as consumer drones, birds, and balloons. These small flying objects are typically small in size, have unstable trajectories, and some lack cooperative beacons or transponders, which can easily impact the low-altitude flight safety of EVTOL aircraft.

[0003] Existing civil aviation collision avoidance systems typically rely on transponder communication, active radar, optoelectronic vision, or lidar. Among these, transponder-based collision avoidance systems struggle to detect small flying objects without cooperative beacons; radar systems generally suffer from large size, power consumption, and payload requirements, and are susceptible to multipath reflections and complex backgrounds in low-altitude environments; visible light vision systems are significantly affected by lighting conditions, weather, and visibility, exhibiting decreased detection stability at night, in fog, or under strong backlight conditions; and lidar has limited effectiveness in detecting some transparent, light-absorbing, or weakly reflective targets, while also being costly and power-consuming.

[0004] For electric vertical takeoff and landing (EVTOL) aircraft, limitations in onboard power supply, payload space, and installation location make it difficult to rely on high-power active detection equipment for continuous omnidirectional detection over extended periods. Furthermore, small flying objects are small in scale and have short appearance times; if avoidance is only initiated when the target is confirmed to be close, insufficient avoidance time may result. Conversely, triggering active detection or avoidance actions for all suspicious targets can easily lead to false triggers, increased energy consumption, or frequent flight path adjustments. Therefore, existing technologies still need to improve the detection, risk assessment, and avoidance control capabilities for small flying objects in low-altitude operation scenarios for EVTOL aircraft. Summary of the Invention

[0005] To address the shortcomings of existing electric vertical takeoff and landing (EVTOL) aircraft in detecting small flying objects without cooperative beacons, the high power consumption and payload of active detection equipment, and the susceptibility to false triggering or untimely avoidance during collision avoidance, this invention provides a small flying object collision avoidance system for EVTOL aircraft. This system can perform target identification, threat assessment, and collision risk calculation based on infrared detection data, and, when triggering conditions are met, introduce auxiliary ranging data to correct the collision risk, thereby generating corresponding warning information or avoidance control commands.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A small object collision avoidance system for electric vertical takeoff and landing aircraft includes: The infrared detection module is used to perform infrared detection on small flying objects within a preset airspace around the electric vertical take-off and landing aircraft, and output target infrared detection data. The target identification and threat assessment module is used to identify the target type based on the target infrared detection data and generate the initial threat level of the target; The collision risk calculation module is used to generate target collision risk data based on the initial threat level and the target infrared detection data, using a calculation strategy corresponding to the initial threat level. The auxiliary ranging module is used to acquire target distance-related data when the initial threat level or the target collision risk data meets the preset triggering conditions; The fusion correction module is used to correct the target collision risk data based on the target infrared detection data and the target distance-related data; The avoidance decision module is used to generate early warning information or avoidance control commands based on the target collision risk data before or after correction.

[0007] By adopting the above technical solution and setting up modules such as infrared detection, target identification and threat assessment, collision risk calculation, auxiliary ranging, fusion correction and avoidance decision-making, the system can first complete the passive detection and preliminary risk assessment of small flying objects based on infrared detection data, and then introduce target distance-related data to correct the collision risk when the triggering conditions are met, thereby taking into account low power consumption detection, reliability of risk assessment and timeliness of avoidance control.

[0008] Further configuration: The target infrared detection data includes at least one of the following: target azimuth angle, target elevation angle, thermal radiation intensity, infrared image sequence, and target imaging size; The infrared detection module includes multiple infrared thermal imaging sensors positioned at different locations on the electric vertical takeoff and landing aircraft. These multiple infrared thermal imaging sensors are used to form multi-directional infrared detection coverage of a preset airspace around the electric vertical takeoff and landing aircraft.

[0009] By adopting the above technical solution, and by including at least one of the following in the target infrared detection data: azimuth angle, elevation angle, thermal radiation intensity, infrared image sequence, and target imaging size, and by using multiple infrared thermal imaging sensors in different orientations to form multi-directional detection coverage, it is beneficial to obtain information on the target's spatial position, thermal radiation state, and imaging changes, thereby improving the completeness of detection of small flying objects around the aircraft and the basis for subsequent identification and tracking.

[0010] Further configuration: The target identification and threat assessment module is used to extract target identification features based on the target infrared detection data, and compare the target identification features with a pre-built infrared feature library of small flying objects to determine the target type; The target recognition features include geometric morphological features extracted based on the target contour, thermal radiation center distribution features extracted based on the thermal radiation distribution of the target region, and thermal radiation ripple frequency features extracted based on the thermal radiation intensity changes of the same target region in multiple consecutive infrared images.

[0011] By adopting the above technical solution, geometric morphological features, thermal radiation center distribution features, and thermal radiation ripple frequency features are extracted from the target infrared detection data and compared with the infrared feature library of small flying objects. This helps to distinguish small targets such as drones, birds, and balloons from multiple dimensions such as morphology, thermal distribution, and time changes, and reduces the risk of misidentification caused by a single infrared intensity judgment.

[0012] Further configuration: The infrared feature library for small flying objects stores infrared feature samples corresponding to different target types; The target recognition and threat assessment module is used to input the target recognition features into the target recognition model to obtain the target feature vector, and compare the target feature vector with the infrared feature sample to output the target type classification result.

[0013] By adopting the above technical solution, storing infrared feature samples corresponding to different target types in the infrared feature library of small flying objects, and inputting the target recognition features into the target recognition model to obtain the target feature vector and then performing similarity comparison, it is beneficial to improve the stability of the target type classification results and make the target recognition process have a clearer data comparison basis.

[0014] Further configuration: The target recognition and threat assessment module generates an initial threat index based on the target type danger coefficient, the normalized value of the pixel area occupied by the target in the infrared image, the normalized value of the target's angular velocity in the field of view, and the thermal radiation contrast intensity of the target relative to the background, and determines the initial threat level based on the comparison result of the initial threat index and the preset threat threshold.

[0015] By adopting the above technical solution, an initial threat index is generated by combining the target type hazard coefficient, pixel area normalization value, angular velocity normalization value, and thermal radiation contrast intensity. This allows threat assessment to consider target type, target size, movement trend, and thermal radiation salience simultaneously, avoiding the determination of threat level based solely on a single target feature and improving the rationality of the initial threat level classification.

[0016] Further settings: The initial threat level includes low threat, medium threat, and high threat; the target collision risk data includes at least one of relative collision time (TTC), target trajectory prediction results, target approach trend, and target risk level. When the initial threat level is low threat, the collision risk calculation module generates the relative collision time (TTC) based on the change in the target's imaging scale in the infrared image; When the initial threat level is medium or high threat, the collision risk calculation module performs trajectory prediction based on the target azimuth information, target pitch information and target imaging scale information, and generates the target collision risk data based on the trajectory prediction results.

[0017] By adopting the above technical solution, the initial threat level is divided into low threat, medium threat and high threat, and different collision risk calculation strategies are adopted according to different threat levels. Low threat targets can be quickly calculated by using infrared imaging scale changes, while medium and high threat targets use trajectory prediction to generate collision risk data, thereby achieving hierarchical processing under limited airborne computing resources.

[0018] Further configuration: The collision risk calculation module determines the target equivalent angular diameter based on the pixel angle of the target-enclosed region in the infrared image, and generates the relative collision time TTC based on the rate of change of the target equivalent angular diameter over time; When the rate of change of the target equivalent angular diameter over time is less than or equal to zero, the collision risk calculation module sets the relative collision time (TTC) to a safe value.

[0019] By adopting the above technical solution, the equivalent angular diameter of the target is determined by the pixel angle of the target's surrounding area, and the relative collision time (TTC) is generated based on the rate of change of the equivalent angular diameter of the target over time. This enables the system to judge the approach trend of the target based on the change in the infrared imaging scale without obtaining the absolute distance of the target. When the rate of change is less than or equal to zero, the TTC is set as a safe value, which helps to reduce false triggering of targets that are far away or moving parallel to the target.

[0020] Further configuration: The collision risk calculation module uses an extended Kalman filter for trajectory prediction; The observation inputs of the extended Kalman filter include the target azimuth angle, the target elevation angle, and the target equivalent angular diameter; When a sudden change in the thermal radiation intensity of the target is detected, the collision risk calculation module increases the observation noise of the corresponding infrared observation data to reduce the impact of the corresponding frame of infrared observation data on the trajectory prediction results.

[0021] By adopting the above technical solution, using extended Kalman filtering for trajectory prediction, and using the target azimuth, target elevation angle, and target equivalent angular diameter as observation inputs, it is beneficial to perform continuous trajectory estimation for medium and high threat targets. When the target thermal radiation intensity changes abruptly, increasing the corresponding observation noise can reduce the interference of abnormal infrared frames on the trajectory prediction results and improve the stability of trajectory prediction.

[0022] Further configuration: The auxiliary ranging module includes a low-power millimeter-wave radar; The preset triggering conditions include: the initial threat level is medium or high threat, or the target collision risk data indicates that the target collision risk reaches a preset risk threshold; The auxiliary ranging module determines the ranging direction based on the target azimuth and elevation angles in the target infrared detection data, and controls the low-power millimeter-wave radar to perform directional ranging in the ranging direction to obtain the target's absolute distance and / or radial velocity.

[0023] By adopting the above technical solution, and by setting up a low-power millimeter-wave radar, ranging is triggered when the initial threat level is medium or high threat, or when the target collision risk reaches a preset risk threshold, the system does not need to continuously turn on the active ranging device; at the same time, directional ranging is performed based on the target azimuth and target elevation angles, which can obtain the target's absolute distance and / or radial velocity, improving the accuracy of collision risk correction in medium and high risk scenarios.

[0024] Further configuration: The avoidance decision module divides the target risk level into a warning zone, an avoidance zone, and an emergency zone based on the target collision risk data and the initial threat level, and generates warning information, active avoidance control instructions, or emergency avoidance control instructions respectively.

[0025] By adopting the above technical solution, the avoidance decision module divides the target collision risk data and initial threat level into warning zones, avoidance zones, and emergency zones, and generates warning information, active avoidance control commands, or emergency avoidance control commands respectively. This enables the system to take graded responses according to the risk level, reducing excessive avoidance caused by low-risk targets, while improving the timeliness of avoidance under high-risk targets. In summary, the present invention has the following beneficial effects: This invention uses an infrared detection module to passively detect small flying objects within a preset airspace around an electric vertical takeoff and landing (EVTOL) aircraft. Combined with a target identification and threat assessment module, it determines the target type and initial threat level. This enables the system to acquire infrared detection data of small flying objects without relying on target cooperative beacons, and to identify and preliminarily differentiate targets such as drones, birds, and balloons. It is suitable for detecting small flying objects operating at low altitudes without transponders or cooperative signals.

[0026] This invention employs different collision risk calculation strategies based on the initial threat level: for low-threat targets, the relative collision time (TTC) is generated based on changes in the target's imaging scale in infrared images; for medium- or high-threat targets, trajectory prediction is performed and target collision risk data is generated based on target azimuth, elevation, and imaging scale information. Thus, the system can perform low-computational screening for low-risk targets and more refined trajectory prediction for medium- and high-risk targets, reducing the ineffective use of airborne computing resources.

[0027] This invention uses an auxiliary ranging module to acquire target distance-related data when the initial threat level or target collision risk data meets preset triggering conditions. The fusion correction module then corrects the target collision risk data based on the target infrared detection data and the target distance-related data. This eliminates the need for the system to continuously operate the active ranging device. At the same time, it can introduce target absolute distance and radial velocity information in medium- to high-risk scenarios, improving the reliability of target approach trend, relative collision time, or trajectory prediction results.

[0028] This invention generates early warning information or avoidance control commands based on target collision risk data before or after correction through an avoidance decision module. It also divides the target risk level into early warning zone, avoidance zone, and emergency zone according to the target collision risk data and initial threat level. This enables the system to take graded responses to small flying objects with different risk levels, which helps to reduce frequent false triggers caused by low-risk targets and improves the timeliness of active avoidance or emergency avoidance under high-risk targets. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall system architecture of the embodiment; Figure 2 This is a schematic diagram illustrating the decision-making process avoidance in an implementation example. Detailed Implementation

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the following embodiments are used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Where there is no conflict, the technical features in the following embodiments can be combined with each other.

[0031] like Figure 1 and Figure 2As shown, this embodiment provides a collision avoidance system for small flying objects used in electric vertical takeoff and landing (EVTOL) aircraft. This system is applied to EVTOL aircraft during low-altitude urban cruising, takeoff and landing, hovering, or route changes to detect, identify, calculate collision risks, and control the avoidance of small flying objects such as uncooperative drones, birds, balloons, and floating objects without cooperative beacons. The system includes an infrared detection module, a target identification and threat assessment module, a collision risk calculation module, an auxiliary ranging module, a fusion correction module, and an avoidance decision module.

[0032] The infrared detection module is used to detect small flying objects within a predetermined airspace around the electric vertical takeoff and landing (EVTOL) aircraft and output target infrared detection data. The infrared detection module consists of multiple infrared thermal imaging sensors distributed at different locations on the aircraft fuselage. It passively receives infrared energy radiated by the small flying object itself or its propulsion system, which may include a motor, battery, or other heat-generating components. Because the infrared detection module uses a passive receiving method, it does not rely on target transponder signals and does not require actively emitting high-power detection beams, making it suitable for low-power detection of small flying objects without cooperative beacons.

[0033] In a preferred embodiment, the infrared thermal imaging sensor employs an uncooled focal plane infrared detector. Multiple uncooled focal plane infrared detectors are respectively arranged on the nose, tail, left wing, right wing, belly, and top of the electric vertical takeoff and landing (EVTOL) aircraft. Each detector's field of view can be set to 100° horizontally and 80° vertically, with at least 10° of overlapping field of view between adjacent sensors. This achieves multi-directional coverage of a preset airspace around the aircraft and facilitates target handover between adjacent fields of view. Each detector can achieve frame rate synchronization via an onboard hardware synchronization trigger board, with a refresh rate of no less than 60Hz, enabling infrared images acquired from multiple directions to participate in target identification, trajectory prediction, and fusion correction under a unified time reference.

[0034] The target infrared detection data includes at least one of the following: target azimuth angle, target elevation angle, thermal radiation intensity, infrared image sequence, and target imaging size. Specifically, the target azimuth angle characterizes the target's horizontal direction of approach relative to the aircraft, and the target elevation angle characterizes the target's vertical direction of approach relative to the aircraft; thermal radiation intensity characterizes the difference in infrared radiation between the target and the background; the infrared image sequence records the target's position, outline, grayscale, and thermal radiation changes in multiple consecutive infrared images; and the target imaging size characterizes the target's pixel area, enclosed region size, pixel angle, or equivalent angular diameter in the infrared image. Of the above data, the azimuth and elevation angles are primarily used for target spatial positioning, trajectory prediction, and auxiliary ranging direction determination, while thermal radiation intensity and infrared image sequences are primarily used for target identification, threat assessment, and collision risk calculation.

[0035] The target identification and threat assessment module is used to identify target types based on infrared detection data and generate an initial threat level for the target. Specifically, the module first performs target detection and segmentation on the infrared image sequence to obtain the target region; then, it extracts target identification features based on the target region. These target identification features include geometric morphological features extracted based on the target contour, thermal radiation center distribution features extracted based on the thermal radiation distribution of the target region, and thermal radiation ripple frequency features extracted based on the changes in thermal radiation intensity of the same target region in multiple consecutive infrared images.

[0036] Geometric features can include the aspect ratio of the target bounding box, the ratio of the major and minor axes of the target outline, the area of ​​the target region, the complexity of the outline, or edge morphology features. Thermal radiation center distribution features can include the location of the thermal radiation centroid within the target region, the offset between the thermal radiation centroid and the geometric center, the distribution of high thermal radiation areas, or the gray-level statistical features of the target region. Thermal radiation ripple frequency features can be obtained by performing time-series analysis of the thermal radiation intensity of the same target region after target tracking and matching in multiple consecutive frames of infrared images; for example, the time-series signal formed by the change of the mean gray level, peak gray level, or local heat source intensity within the target region over time can be extracted, and frequency domain analysis can be performed on this time-series signal to obtain the thermal radiation ripple frequency features. Here, "same target region" refers to the target region belonging to the same small flying object after cross-frame target association, rather than the same pixel region under fixed image coordinates.

[0037] The target recognition and threat assessment module compares target recognition features with a pre-built infrared feature library of small flying objects to determine the target type. This library stores infrared feature samples corresponding to different target types, including infrared thermal images of real-flying UAVs, infrared thermal images of birds, and target infrared data obtained through software thermodynamic simulation. The module employs a lightweight convolutional neural network, inputting the extracted target recognition features into a target recognition model to obtain a target feature vector. This target feature vector is then compared with the infrared feature samples using cosine similarity to output a target type classification result. The target types can include consumer-grade UAVs, birds, balloons, plastic floats, or other low-altitude small targets.

[0038] After target type identification, the target identification and threat assessment module performs a quantitative calculation of the threat level. Specifically, the module generates an initial threat index TH based on the target type hazard coefficient, the normalized pixel area occupied by the target in the infrared image, the normalized angular velocity of the target in the field of view, and the thermal radiation contrast intensity of the target relative to the background. The initial threat index TH can be calculated using the following formula:

[0039] Wherein, C is the target type hazard coefficient, which is retrieved based on the target identification results to determine the hazard coefficient of the corresponding target type. The C value for each target type can be determined by expert scoring. S is the normalized value of the pixel area occupied by the target in the infrared image, used to characterize the target imaging scale. W is the normalized value of the angular velocity of the target in the field of view, used to characterize the degree of motion change of the target in the infrared field of view. L is the thermal radiation contrast intensity of the target relative to the background, used to characterize the difference in thermal radiation between the target and the background. w1, w2, w3, and w4 are weight coefficients, whose values ​​can be obtained by machine learning regression from historical flight test data, or by constructing a judgment matrix containing multi-dimensional threat assessment factors and solving it using the analytic hierarchy process (AHP) for consistency verification.

[0040] The target identification and threat assessment module compares the calculated initial threat index TH with preset threat thresholds TH1 and TH2 to determine the initial threat level. Specifically, when TH is less than TH1, it is classified as low threat; when TH is greater than or equal to TH1 and less than TH2, it is classified as medium threat; and when TH is greater than or equal to TH2, it is classified as high threat. If the threshold expression "TH1≤TH≤TH2 is medium threat, TH≥TH2 is high threat" is used in actual implementation, then in the boundary case where TH equals TH2, it can be classified as high threat according to the conservative principle to avoid boundary overlap leading to unclear execution results.

[0041] The collision risk calculation module generates target collision risk data based on the initial threat level and target infrared detection data, using a calculation strategy corresponding to the initial threat level. The target collision risk data includes at least one of the following: relative collision time (TTC), target trajectory prediction results, target approach trend, and target risk level. Here, the target risk level can be the preliminary risk level formed during the collision risk calculation phase, and the avoidance decision module can further combine the initial threat level and the corrected target collision risk data to form the final avoidance level.

[0042] When the initial threat level is low, the collision risk calculation module employs a low-computation infrared estimation strategy. Specifically, the collision risk calculation module generates the relative collision time (TTC) based on the target's imaging scale changes in the infrared image. The module determines the target's equivalent angular diameter θ based on the pixel angle of the target's enclosed region in the infrared image, and calculates the rate of change of the target's equivalent angular diameter θ over time. The relative collision time (TTC) is generated. The target equivalent angular diameter θ can be the angle subtended by the pixels corresponding to the diagonal of the target bounding box, or it can be calculated based on the pixel size of the target bounding region and the field of view parameters of the infrared thermal imaging sensor.

[0043] In one specific calculation method, let the relative velocity be V. rel Let the actual width of the target be W and the target distance be D. Under small angle conditions, we can approximate the following: If the target approaches the aircraft along the line of sight, the target distance D decreases, and the target's equivalent angular diameter θ increases. Greater than 0. At this point, we can approximate the result as follows:

[0044] in, Let be the rate of change of the target equivalent angular diameter over time. When the value is less than or equal to 0, it indicates that the target is moving away from the aircraft, moving parallel to the aircraft, or showing no tendency to approach. In this case, the collision risk calculation module sets the relative collision time TTC to infinity or another safe value. When the value is greater than 0, the system continues to monitor TTC. In this way, collision risk pre-screening can be performed with low power consumption and low computing power without directly obtaining the target's absolute distance and true relative velocity during the low-threat phase.

[0045] When the initial threat level is medium or high, the collision risk calculation module employs a trajectory prediction strategy. Specifically, the collision risk calculation module performs trajectory prediction based on target azimuth information, target elevation information, and target imaging scale information, and generates target collision risk data based on the trajectory prediction results. Target azimuth information can be derived from the target azimuth angle, target elevation information can be derived from the target elevation angle, and target imaging scale information can include the target imaging size, the pixel angle of the target's enclosed region, or the target's equivalent angular diameter.

[0046] In a preferred embodiment, the collision risk calculation module employs an extended Kalman filter for trajectory prediction. The observation inputs to the extended Kalman filter include the target azimuth angle, target elevation angle, and target equivalent angular diameter. Its state equation is established using a modified polar coordinate system to mitigate the trajectory estimation instability caused by the inability to directly observe the target distance in pure azimuth tracking. The collision risk calculation module recursively estimates the target's motion state based on continuous infrared observation data, smoothly predicting the target's motion trend over the next N seconds, and calculates the relative collision time (TTC), target approach trend, or preliminary target risk level based on the filtered trajectory data.

[0047] During infrared observation, the infrared radiation intensity of the target may fluctuate drastically due to background flicker, occlusion, changes in target attitude, or changes in heat sources. To reduce the impact of abnormal infrared observations on trajectory prediction results, when the collision risk calculation module detects a sudden change in the target's thermal radiation intensity, it adaptively increases the observation noise matrix of the corresponding infrared observation data and decreases the Kalman gain corresponding to that frame of observation data. This reduces the impact of that frame of infrared observation data on the trajectory prediction results, thereby preventing trajectory drift.

[0048] The auxiliary ranging module is used to acquire target distance-related data when the initial threat level or target collision risk data meets preset trigger conditions. In one embodiment, the auxiliary ranging module includes a low-power millimeter-wave radar. To reduce power consumption and electromagnetic exposure, the low-power millimeter-wave radar is in standby / dormant state by default. When the initial threat level of the target reaches medium or high threat, or when the target collision risk data indicates that the target collision risk reaches a preset risk threshold, the auxiliary ranging module wakes up the low-power millimeter-wave radar. The preset risk threshold may include a relative collision time TTC less than a first preset threshold T1.

[0049] The auxiliary ranging module determines the ranging direction based on the target's azimuth and elevation angles from the target's infrared detection data, and controls the low-power millimeter-wave radar to perform directional ranging in that direction to obtain the target's absolute range and / or radial velocity. The absolute range is used to correct errors when retrieving range from pure infrared angles and imaging scales, while the radial velocity characterizes the target's approach or departure speed along the aircraft's line-of-sight. Therefore, the system can activate active ranging only when medium-to-high threats or risks reach a threshold, thus balancing low-power detection with range accuracy in high-risk scenarios.

[0050] The fusion correction module is used to correct target collision risk data based on target infrared detection data and target range-related data. When the low-power millimeter-wave radar is activated, the fusion correction module can input the absolute range and radial velocity fed back from the millimeter-wave radar, along with the target azimuth, elevation, and equivalent angular diameter output from the infrared detection module, into the measurement vector of the extended Kalman filter for measurement-level data fusion. This corrects the TTC error and trajectory prediction error obtained from pure infrared inversion. The corrected target collision risk data may include the corrected relative collision time (TTC), the corrected target trajectory prediction result, the corrected target approach trend, or the corrected preliminary target risk level.

[0051] In one implementation, the fusion correction module also executes auxiliary verification logic. When the infrared detection module misjudges a target as a high threat due to environmental background flicker, local heat source interference, or other reasons, and the low-power millimeter-wave radar confirms that the target's absolute distance is much greater than the safety boundary, the fusion correction module can use the radar distance determination result as the basis for correction, reducing the risk level of the target collision risk data, and enabling the avoidance decision module to suppress unnecessary avoidance control commands, thereby preventing the aircraft from making unnecessary avoidance actions. Conversely, when the target is made of radar-absorbing material or has a special configuration that results in a small radar cross-section and radar miss, but the infrared detection module continuously outputs high-risk thermal radiation characteristics, and the initial threat index TH is greater than or equal to the second threat threshold TH2, the fusion correction module or the avoidance decision module can trigger a missed alarm defense mechanism and execute avoidance according to the infrared high-risk characteristic judgment result.

[0052] After the current collision risk is eliminated, the auxiliary ranging module can control the low-power millimeter-wave radar to switch back to standby / sleep mode. Eliminating the collision risk can include: the target's initial threat level decreasing to low threat, and the relative collision time TTC exceeding a first preset threshold T1 for more than 2 seconds; or the target leaving the infrared detection module's field of view for more than 1 second. In this way, the system acquires distance and velocity information during the high-risk phase and resumes low-power operation after the risk is eliminated.

[0053] The avoidance decision module is used to generate early warning information or avoidance control instructions based on the target collision risk data before or after correction. Specifically, the avoidance decision module divides the target risk level into early warning zone, avoidance zone, and emergency zone according to the target collision risk data and the initial threat level, and generates early warning information, active avoidance control instructions, or emergency avoidance control instructions respectively.

[0054] When the relative collision time TTC is greater than the first preset threshold T1 and the initial threat level is low threat, the avoidance decision module determines that the target is in the warning zone, generates a warning message, prompts the pilot or autopilot system to pay attention to the target, and does not execute active avoidance actions. When the relative collision time TTC is less than or equal to the first preset threshold T1 and greater than the second preset threshold T2, or the initial threat level is medium threat, the avoidance decision module determines that the target is in the avoidance zone, triggers active avoidance, and selects a horizontal avoidance, vertical avoidance, or hovering waiting strategy based on the relative position of the target and the aircraft; horizontal avoidance may include yaw or lateral maneuvers, and vertical avoidance may include climb or descent. When the relative collision time TTC is less than or equal to the second preset threshold T2, or the initial threat level is high threat, the avoidance decision module determines that the target is in the emergency zone, immediately executes maximum performance avoidance maneuvers, and may trigger bounce avoidance or rotor thrust vectoring assist actions.

[0055] The first preset threshold T1 can be the predicted time required for the aircraft to complete a smooth avoidance maneuver in standard cruise mode. The setting of T1 should ensure that the aircraft has sufficient time after entering the avoidance zone to evade the collision path in a manner that meets the requirements of load comfort and flight control stability, while avoiding premature triggering that would lead to frequent course adjustments. The second preset threshold T2 can be the sum of the total system response delay time and safety margin required for the aircraft to perform extreme maneuvers; the total system response delay time can include sensor acquisition delay, control command processing delay, and the physical action response time of the flight control actuators. When TTC is less than T2, it indicates that even if the system responds immediately at full power, it is difficult to completely eliminate the collision risk, therefore an emergency avoidance maneuver needs to be triggered. When the risk assessment result corresponding to the initial threat level is inconsistent with the risk assessment result corresponding to TTC, the avoidance decision module adopts a conservative principle, that is, if any indicator points to a higher risk level, the corresponding warning information or avoidance control command is generated according to the higher risk level.

[0056] In one specific operational process, the system first uses an infrared detection module to continuously collect infrared data within a preset airspace around the aircraft, and outputs target infrared detection data such as target azimuth, target pitch angle, thermal radiation intensity, infrared image sequence, and target imaging size. The target identification and threat assessment module extracts target identification features from the target infrared detection data and compares them with a small aircraft infrared feature database to determine the target type. Subsequently, the target identification and threat assessment module calculates an initial threat index TH based on the target type's hazard coefficient, pixel area normalization value, angular velocity normalization value, and thermal radiation contrast intensity, and determines whether the threat level is low, medium, or high based on the comparison results of TH with preset threat thresholds TH1 and TH2.

[0057] When the target poses a low threat, the collision risk calculation module estimates the relative collision time (TTC) using changes in the target's imaging scale in the infrared image. If the target's equivalent angular diameter does not increase, the TTC is set to a safe value; if the target's equivalent angular diameter continues to increase, the TTC is used to determine whether an early warning or further auxiliary ranging should be triggered. When the target poses a medium or high threat, the collision risk calculation module uses an extended Kalman filter for trajectory prediction and generates target collision risk data based on the filtered trajectory data. When the initial threat level or target collision risk data meets preset triggering conditions, the auxiliary ranging module activates the low-power millimeter-wave radar and performs directional ranging according to the azimuth and elevation directions indicated by the infrared detection data to obtain absolute distance and / or radial velocity. The fusion correction module fuses and corrects the infrared observation data with target distance-related data, and the avoidance decision module generates early warning information or avoidance control commands based on the target collision risk data before or after correction.

[0058] In a low-threat target warning scenario, when an electric vertical takeoff and landing (EVA) aircraft is cruising at low speed over an urban area, a hovering consumer-grade balloon or a distant bird appears at a distance. The forward-facing infrared detection module receives the target's thermal radiation signal and outputs the target's azimuth, pitch angle, and infrared image sequence. The target identification and threat assessment module extracts the target's geometric and thermal radiation features, outputs a target feature vector via a lightweight convolutional neural network, and performs a cosine similarity comparison with a small flying object infrared feature library to determine if the target is a low-risk bird or other low-risk target. The target identification and threat assessment module calculates an initial threat index (TH) less than TH1, classifying the initial threat level as low threat. Under the low-threat strategy, the collision risk calculation module does not initiate complex trajectory prediction but instead extracts the target's equivalent angular diameter and calculates the time-to-collision (TTC) based on its rate of change. Upon receiving the result of low threat and a TTC greater than T1, the avoidance decision module determines that the target is in the warning zone. The system highlights the target information on the cockpit display or sends a warning signal to the autopilot system, but does not perform active avoidance.

[0059] In a scenario involving active avoidance of a medium-threat target, an electric vertical takeoff and landing (EVT) aircraft is cruising at low altitude at night when a small, unmanned drone (UAV) crossing its flight path without a transponder appears. The forward-facing infrared detection module receives thermal radiation signals from the UAV's motors or batteries, outputting the target's azimuth, pitch angle, and infrared image sequence. The target identification and threat assessment module extracts the target's geometric features, thermal radiation center distribution characteristics, and thermal radiation ripple frequency characteristics. After applying a lightweight convolutional neural network and cosine similarity comparison, the target is identified as a small UAV. The target identification and threat assessment module calculates a threat level (TH) greater than or equal to TH1 and less than TH2, classifying the initial threat level as medium threat. The auxiliary ranging module activates a dormant low-power millimeter-wave radar. The radar performs a directional scan based on the azimuth and pitch directions indicated by the infrared detection data, acquiring the target's absolute range and radial velocity. The collision risk calculation module employs an extended Kalman filter, combining the target azimuth, pitch, and equivalent angular diameter obtained from infrared detection, along with the absolute range and radial velocity fed back from radar, for target trajectory prediction. It then calculates the Time-to-Cost (TTC) based on the filtered trajectory data. The avoidance decision module, based on the medium threat level and the judgment result of T1 ≥ TTC > T2, determines that the target is within the avoidance zone. The flight control system then generates horizontal yaw, vertical climb, or descent commands based on the fused relative azimuth and velocity, enabling the aircraft to smoothly exit the collision path.

[0060] In high-threat target emergency avoidance and conflict handling scenarios, when a small UAV, flock of birds, or other small targets with high thermal radiation rapidly approach at close range during the takeoff, landing, or low-altitude cruise of an electric vertical takeoff and landing (EVT) aircraft, the infrared detection module outputs the target's azimuth, pitch angle, thermal radiation intensity, and infrared image sequence. The target identification and threat assessment module calculates a TH greater than or equal to TH2 based on the target type's hazard factor, target imaging area, field-of-view angular velocity, and thermal radiation contrast, classifying the initial threat level as high threat. The collision risk calculation module uses an extended Kalman filter for trajectory prediction and increases the corresponding observation noise when the target's thermal radiation intensity changes abruptly, reducing the impact of abnormal frames on the prediction results. After being triggered, the auxiliary ranging module acquires the target's absolute distance and radial velocity, and the fusion correction module corrects the target trajectory prediction results and TTC accordingly. When the TTC is less than or equal to T2, or the initial threat level remains high threat, the avoidance decision module determines that the target is in an emergency zone, generates an emergency avoidance control command, and executes maximum performance avoidance maneuvers, bounce avoidance, or rotor thrust vectoring assist actions.

[0061] In Example 3, if the infrared detection module misjudges a target as a high threat due to background heat flicker, but the low-power millimeter-wave radar confirms that the target's absolute distance is greater than the safety boundary, the fusion correction module reduces the risk level of the target collision risk data, and the avoidance decision module suppresses unnecessary avoidance control commands. If the target is made of a radar-absorbing material or has a special configuration that causes the radar to fail to detect the target, but the infrared detection module continuously outputs high-risk thermal radiation characteristics for a preset number of consecutive frames or a preset continuous duration, and TH is greater than or equal to TH2, the system triggers the missed alarm defense mechanism and generates avoidance control commands according to the infrared high-risk characteristic judgment result.

[0062] Through the above embodiments, the present invention can perform low-power collision risk estimation using infrared detection data in low-threat situations, improve the reliability of risk assessment through trajectory prediction and on-demand auxiliary ranging in medium and high-threat situations, and generate graded early warning or avoidance control commands based on target collision risk data and initial threat level, thereby realizing passive infrared detection, graded risk calculation, on-demand ranging correction and graded avoidance control for small flying objects.

[0063] The embodiments described above do not constitute a limitation on the scope of protection of this technical solution. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the above embodiments should be included within the scope of protection of this technical solution.

Claims

1. A small object collision avoidance system for electric vertical takeoff and landing aircraft, characterized in that, include: The infrared detection module is used to perform infrared detection on small flying objects within a preset airspace around the electric vertical take-off and landing aircraft, and output target infrared detection data. The target identification and threat assessment module is used to identify the target type based on the target infrared detection data and generate the initial threat level of the target; The collision risk calculation module is used to generate target collision risk data based on the initial threat level and the target infrared detection data, using a calculation strategy corresponding to the initial threat level. The auxiliary ranging module is used to acquire target distance-related data when the initial threat level or the target collision risk data meets the preset triggering conditions; The fusion correction module is used to correct the target collision risk data based on the target infrared detection data and the target distance-related data; The avoidance decision module is used to generate early warning information or avoidance control commands based on the target collision risk data before or after correction.

2. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 1, characterized in that, The target infrared detection data includes at least one of the following: target azimuth angle, target elevation angle, thermal radiation intensity, infrared image sequence, and target imaging size; The infrared detection module includes multiple infrared thermal imaging sensors positioned at different locations on the electric vertical takeoff and landing aircraft. These multiple infrared thermal imaging sensors are used to form multi-directional infrared detection coverage of a preset airspace around the electric vertical takeoff and landing aircraft.

3. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 1, characterized in that, The target identification and threat assessment module is used to extract target identification features based on the target infrared detection data, and compare the target identification features with a pre-built infrared feature library of small flying objects to determine the target type; The target recognition features include geometric morphological features extracted based on the target contour, thermal radiation center distribution features extracted based on the thermal radiation distribution of the target region, and thermal radiation ripple frequency features extracted based on the thermal radiation intensity changes of the same target region in multiple consecutive infrared images.

4. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 3, characterized in that, The infrared feature library for small flying objects stores infrared feature samples corresponding to different target types; The target recognition and threat assessment module is used to input the target recognition features into the target recognition model to obtain the target feature vector, and compare the target feature vector with the infrared feature sample to output the target type classification result.

5. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 1, characterized in that, The target identification and threat assessment module generates an initial threat index based on the target type hazard coefficient, the normalized value of the pixel area occupied by the target in the infrared image, the normalized value of the target's angular velocity in the field of view, and the thermal radiation contrast intensity of the target relative to the background. The module then determines the initial threat level based on the comparison between the initial threat index and a preset threat threshold.

6. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 2, characterized in that, The initial threat level includes low threat, medium threat and high threat, and the target collision risk data includes at least one of relative collision time (TTC), target trajectory prediction results, target approach trend and target risk level; When the initial threat level is low threat, the collision risk calculation module generates the relative collision time (TTC) based on the change in the target's imaging scale in the infrared image; When the initial threat level is medium or high threat, the collision risk calculation module performs trajectory prediction based on the target azimuth information, target pitch information and target imaging scale information, and generates the target collision risk data based on the trajectory prediction results.

7. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 6, characterized in that, The collision risk calculation module determines the target equivalent angular diameter based on the pixel angle of the target-enclosed region in the infrared image, and generates the relative collision time TTC based on the rate of change of the target equivalent angular diameter over time. When the rate of change of the target equivalent angular diameter over time is less than or equal to zero, the collision risk calculation module sets the relative collision time (TTC) to a safe value.

8. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 7, characterized in that, The collision risk calculation module uses an extended Kalman filter to predict the trajectory. The observation inputs of the extended Kalman filter include the target azimuth angle, the target elevation angle, and the target equivalent angular diameter; When a sudden change in the thermal radiation intensity of the target is detected, the collision risk calculation module increases the observation noise of the corresponding infrared observation data to reduce the impact of the corresponding frame of infrared observation data on the trajectory prediction results.

9. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 6, characterized in that, The auxiliary ranging module includes a low-power millimeter-wave radar; The preset triggering conditions include: the initial threat level is medium or high threat, or the target collision risk data indicates that the target collision risk reaches a preset risk threshold; The auxiliary ranging module determines the ranging direction based on the target azimuth and elevation angles in the target infrared detection data, and controls the low-power millimeter-wave radar to perform directional ranging in the ranging direction to obtain the target's absolute distance and / or radial velocity.

10. The small object collision avoidance system for electric vertical takeoff and landing aircraft according to claim 1, characterized in that, The avoidance decision module divides the target risk level into a warning zone, an avoidance zone, and an emergency zone based on the target collision risk data and the initial threat level, and generates warning information, active avoidance control instructions, or emergency avoidance control instructions respectively.

Images