Search

CN-121970740-A - Bird-repellent self-adaptive control method fusing bird feature AI identification

CN121970740ACN 121970740 ACN121970740 ACN 121970740ACN-121970740-A

Abstract

The invention discloses a bird-repellent self-adaptive control method fusing bird feature AI identification, and relates to the technical field of airspace safety control. The method first extracts the space-time trajectory features of the target birds and calculates the intent confidence coefficient, and activates the system only when an intrusion intent is identified. Subsequently, a historical tolerance index of the safety mask and the target is obtained, and safety interlock based on the spatial position is performed. In the decision stage, the actual latency of the target from stimulated to displaced is monitored, and based on the latency deviation and tolerance index, a model comprising an exponential function and a truncation logic is used to calculate a control gain. In the execution stage, the control gain is mapped to the peak height of the virtual repulsive potential energy field, a negative gradient vector is calculated, and the driving-off device is controlled to execute a non-aiming guiding action along the negative gradient direction. The bird repellent device solves the problems of long-term use failure of bird repellent equipment and secondary disasters caused by unordered frightening of birds, and realizes safe and efficient self-adaptive driving.

Inventors

  • ZHANG HAIFENG

Assignees

  • 深圳市天翊瑞霖智能科技有限公司

Dates

Publication Date
20260505
Application Date
20260203

Claims (8)

  1. 1. A bird-repellent self-adaptive control method integrating bird feature AI recognition is characterized by comprising the following steps: s1, extracting space-time track characteristics of target birds, and calculating an intention confidence coefficient, wherein when the intention confidence coefficient is larger than a first threshold value, an activation instruction is generated; s2, responding to the activation instruction, acquiring historical tolerance indexes of a safety mask and a target, and generating a driving-off enabling signal, wherein the generating logic of the driving-off enabling signal comprises Boolean operation based on the safety mask; Step S3, acquiring the actual latency of the target from stimulus to displacement, calculating a relative deviation ratio based on the difference between the actual latency and a safe escape time threshold, and calculating a control gain based on the relative deviation ratio and the historical tolerance index; And S4, mapping the control gain into the peak height of the virtual repulsive potential energy field, calculating a negative gradient vector of the virtual repulsive potential energy field, and controlling the driving-off equipment to execute actions along the direction of the negative gradient vector.
  2. 2. The method according to claim 1, wherein in said step S1, the step of calculating an intention confidence coefficient comprises: calculating the geometric curvature and the speed vector change rate of the track; Outputting the intention confidence coefficient with the value of 1 if the geometric curvature is larger than a second threshold value or the speed vector change rate is smaller than a third threshold value; otherwise, outputting the intent confidence coefficient with a value of 0 and resetting the historical tolerance index for the target.
  3. 3. The method according to claim 1, wherein in the step S2, the step of generating a drive-off enable signal comprises: calculating a projection path coordinate of the driving-off equipment at the next moment, and performing Boolean AND operation on the projection path coordinate and the safety mask; if the operation result is 0, generating an invalid driving-off enabling signal, wherein the driving-off equipment comprises a laser emitting module, and responding to the invalid driving-off enabling signal, cutting off a power supply loop of the laser emitting module; and if the operation result is 1, generating the effective driving-off enabling signal.
  4. 4. The method according to claim 1, wherein in said step S3, the step of calculating a control gain comprises: Acquiring the current historical accumulated interaction times of a target, and calculating the normalized historical tolerance index; calculating an initial gain based on the relative deviation ratio and the historical tolerance index using a calculation model comprising an exponential function; Judging whether the initial gain is larger than a maximum gain threshold value or not; if yes, setting the control gain to be the maximum gain threshold; otherwise, the control gain is set to the initial gain.
  5. 5. The method of claim 4, wherein the computational model comprising an exponential function is: Wherein, the For the initial gain to be the same as the initial gain, For the relative deviation ratio to be the same, For the said historical tolerance index(s), In order to withstand the compensation weights, Is a aggressive factor.
  6. 6. The method according to claim 1, wherein the step S3 further comprises: adjusting a time-frequency variation parameter of the driving-off signal according to the control gain; The adjusting time-frequency variation parameter is positively correlated with shannon entropy of the signal, including adjusting a jump rate of the sound wave frequency or a beam flicker frequency.
  7. 7. The method according to claim 1, wherein in said step S4, controlling the driving-off device to perform an action in the direction of the negative gradient vector comprises: calculating the gradient direction of the virtual repulsive potential energy field at the current position of the target; And controlling the projection focus of the driving-off device to point to the direction opposite to the gradient direction.
  8. 8. A bird repellent adaptive control system incorporating bird feature AI identification for implementing the method of any of claims 1 to 7, comprising: The intention perception module is configured to extract space-time track characteristics and output an activation instruction; a safety arbitration module configured to perform a boolean operation based on a safety mask and output a drive-off enable signal; a policy generation module configured to calculate a control gain based on the actual latency; and the execution control module is configured to calculate a negative gradient vector of the virtual repulsive potential energy field and control the driving-out device.

Description

Bird-repellent self-adaptive control method fusing bird feature AI identification Technical Field The invention relates to the technical field of airspace safety control, in particular to a bird-repellent self-adaptive control method fusing bird feature AI identification. Background In sensitive areas such as airport runways, high-voltage substations and the like, bird invasion is a main risk source for threatening aviation safety and power transmission stability. The existing bird repelling technology generally adopts physical stimulation means such as sound wave repelling, laser scanning or explosion light. However, such techniques commonly employ open loop control logic, i.e., the system mechanically outputs a fixed pattern or simply random pattern stimulation signal upon detection of a target or triggering at a preset time. Because of the lack of cognition and interaction with the biological intelligence of birds, this open loop control faces a systematic failure problem in that birds have very strong neural adaptability and learning ability, which rapidly develop habituation reactions upon prolonged exposure to repeated stimuli without substantial injury, resulting in an exponential decay in the driving efficacy over time. To combat the adaptation, some of the existing solutions attempt to simply increase the intensity or frequency of the physical stimulus. However, this strategy introduces new security risks while addressing the adaptability problem. The high-intensity indiscriminate stimulus is very likely to trigger the biological stress response of birds, resulting in disordered frightening behavior of the target. In limited space such as airports, unordered frightened bird groups are very easy to invade an airplane take-off and landing channel or strike key equipment, and secondary disasters which are more serious than static stopping and dwelling are caused. Therefore, the prior art is difficult to establish balance between maintaining long-term driving-away effectiveness and guaranteeing driving-away process safety, and cannot cope with complex game challenges brought by targets with cognitive learning ability. Disclosure of Invention The invention provides a bird-repellent self-adaptive control method integrating bird feature AI identification, which aims to solve a pair of contradictions faced by the conventional bird-repellent device in long-term use, wherein on one hand, birds are not afraid after being used to fixed stimulation, so that the device is invalid, and on the other hand, if the stimulation intensity is simply increased, frightened birds are easy to randomly fly out, and on the other hand, the frightened birds possibly collide with an airplane or the device in an airport and other areas, so that safety accidents are caused. In view of the above problems, the present invention provides a bird repellent adaptive control method fusing bird feature AI identification, comprising the steps of: s1, extracting space-time track characteristics of target birds, and calculating an intention confidence coefficient, wherein when the intention confidence coefficient is larger than a first threshold value, an activation instruction is generated; s2, responding to the activation instruction, acquiring historical tolerance indexes of a safety mask and a target, and generating a driving-off enabling signal, wherein the generating logic of the driving-off enabling signal comprises Boolean operation based on the safety mask; Step S3, acquiring the actual latency of the target from stimulus to displacement, calculating a relative deviation ratio based on the difference between the actual latency and a safe escape time threshold, and calculating a control gain based on the relative deviation ratio and the historical tolerance index; And S4, mapping the control gain into the peak height of the virtual repulsive potential energy field, calculating a negative gradient vector of the virtual repulsive potential energy field, and controlling the driving-off equipment to execute actions along the direction of the negative gradient vector. Further, in the step S1, the step of calculating the intention confidence coefficient includes: calculating the geometric curvature and the speed vector change rate of the track; Outputting the intention confidence coefficient with the value of 1 if the geometric curvature is larger than a second threshold value or the speed vector change rate is smaller than a third threshold value; otherwise, outputting the intent confidence coefficient with a value of 0 and resetting the historical tolerance index for the target. Further, in the step S2, the step of generating the driving enable signal includes: calculating a projection path coordinate of the driving-off equipment at the next moment, and performing Boolean AND operation on the projection path coordinate and the safety mask; if the operation result is 0, generating an invalid driving-off enabling signal, wherein the drivin