CN-122017868-A - Target detection and positioning method for laser radar point cloud and infrared image fusion

CN122017868ACN 122017868 ACN122017868 ACN 122017868ACN-122017868-A

Abstract

The invention relates to the technical field of multi-sensor fusion perception and discloses a target detection and positioning method for laser radar point cloud and infrared image fusion. The method comprises the steps of utilizing a space-time heterogeneous association controller to construct a time window to trigger synchronous acquisition, detecting characteristic conflict of infrared and laser, extracting a penetrability geometric depth vector through waveform layering analysis, calculating a depth gradient to determine a physical true value edge point, constructing a geometric confidence attenuation field to execute morphological contraction on an infrared thermal image to generate an accurate semantic profile matrix, and carrying out orthogonal fusion on the depth vector and the semantic profile to solve the problem of physical decoupling of heterogeneous characteristics caused by a strong scattering medium.

Inventors

Zheng Xumeng
Huang Bingxiao
ZHAO ZHICHENG

Assignees

凯宸能源(浙江)有限公司
陕西凯宸光电科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (9)

1. The target detection and positioning method based on laser radar point cloud and infrared image fusion is characterized by being applied to a perception system provided with a space-time heterogeneous association controller, wherein the perception system comprises a full-waveform laser radar module, an infrared thermal imaging camera module and a history track storage module, and the method comprises the following steps: the method comprises the steps that S1, a space-time heterogeneous association controller calculates a target motion period according to data in a history track storage module and constructs a target expected arrival time window, and only when a system clock is in the target expected arrival time window, the full-waveform laser radar module and the infrared thermal imaging camera module are activated to execute synchronous acquisition to acquire original full-waveform echo data and original infrared thermal image data; Step S2, the space-time heterogeneous association controller performs feature conflict detection on the original infrared thermal image data and the original full-waveform echo data, judges that scattering medium interference exists when detecting that the infrared thermal radiation features and the near-field echo intensity of the laser radar have logic conflict, performs layering analysis on the original full-waveform echo data, and extracts secondary echo signals of a far-field area to generate a penetrability geometric depth vector; Step S3, the time-space heterogeneous association controller calculates the time domain gradient of the penetrability geometric depth vector in the scanning direction to determine a physical truth value edge point, constructs a geometric confidence attenuation field by utilizing the physical truth value edge point, performs morphological contraction operation on the original infrared thermal image data based on the geometric confidence attenuation field, eliminates a thermal diffusion halation region and generates an accurate semantic contour matrix; And S4, carrying out orthogonal fusion on the penetrability geometric depth vector and the accurate semantic profile matrix, and solving the three-dimensional space coordinate of the target.
2. The method according to claim 1, wherein the specific condition for determining that the scattering medium interference exists in the step S2 is: the space-time heterogeneous association controller identifies that a communication region higher than the background temperature exists in the original infrared thermal image data; And searching a laser radar beam corresponding to the communication area, and if the first echo peak intensity of the beam in the preset near field safety distance is larger than a preset scattering threshold value, judging that the infrared characteristic is a transmission signal and the laser characteristic is a scattering signal, and confirming that the current scattering medium interference working condition exists.
3. The method according to claim 2, wherein the specific process of performing hierarchical parsing in step S2 includes: the time-space heterogeneous association controller generates a time domain shielding mask and filters the first echo peak value and a time domain neighborhood signal thereof; And searching a secondary echo peak value with the highest signal-to-noise ratio in the filtered residual waveform data, and converting the flight time of the secondary echo peak value into the penetrability geometric depth vector.
4. The method according to claim 1, wherein the specific process of determining the physical true value edge points in step S3 includes: Calculating depth differences between adjacent light spots in the penetration geometric depth vector; when the modular length of the depth difference value is in step and exceeds a preset physical edge threshold, marking the corresponding position as the physical truth value edge point; and the space-time heterogeneous association controller calculates the minimum Euclidean distance from each pixel point in the original infrared thermal image data to an edge track formed by fitting the edge points with physical truth values.
5. The method according to claim 4, wherein performing a morphological contraction operation in step S3 includes constructing a geometric confidence attenuation field and calculating a thermal radiation overflow entropy; the calculation formula of the geometric confidence attenuation field is as follows: Wherein, the Representing the infrared image pixel coordinates, Representing the geometric confidence of the pixel point, Representing pixel coordinates To a minimum euclidean distance value of an edge trajectory formed by the physical true value edge point fitting, Is a physical radius parameter of the laser spot, Is an edge steepening factor.
6. The method according to claim 5, wherein the specific process of generating the precise semantic profile matrix in step S3 further comprises: Calculating the local gradient modular length of the original infrared thermal image data, and constructing a thermal radiation overflow entropy model by combining the geometric confidence attenuation field: Wherein, the Representing pixel coordinates The heat radiation at the location overflows the entropy, Representing the original infrared thermal image data at pixel coordinates The local gradient mode length at the position is equal to the local gradient mode length, Is a numerical stability constant; Generating the accurate semantic profile matrix by a nonlinear pruning function based on the thermal radiation overflow entropy : Wherein, the The raw pixel intensity of the raw infrared thermographic data, As a geometric weight balance factor, In order to suppress the intensity coefficient of the light, As a function of the cut-off of the values, Is the overflow entropy upper limit.
7. The method according to claim 1, wherein in the step S1, the generating logic of the target expected arrival time window is: A time sequence regression prediction algorithm is applied to the historical target motion time sequence in the historical track storage module, and the entering time and the leaving time of the current period are predicted; Outside the target expected arrival time window, the spatio-temporal heterogeneous associated controller is in a low power conservation mode and logically blocks signal inputs of all sensors.
8. The method according to any one of claims 1 to 7, wherein the target is a rotating blade of a wind power plant and the environmental obstacle is a tower of a wind power plant; The scattering medium interference is optical scattering interference caused by dense fog, water vapor clusters, or rain and snow aerosols.
9. A perception system, comprising: The full waveform laser radar module is used for transmitting laser beams and receiving full waveform echo data; the infrared thermal imaging camera module is used for acquiring infrared thermal image data; the historical track storage module is used for storing the motion time sequence data of the target; And the time-space heterogeneous association controller is in communication connection with the full-waveform laser radar module, the infrared thermal imaging and shooting module and the history track storage module and is used for executing the target detection and positioning method for fusion of the laser radar point cloud and the infrared image according to any one of claims 1 to 8.

Description

Target detection and positioning method for laser radar point cloud and infrared image fusion Technical Field The invention relates to the technical field of multi-sensor fusion perception, in particular to a target detection and positioning method for laser radar point cloud and infrared image fusion. Background In clearance monitoring applications for periodically moving targets such as wind turbine generator set blades, the sensing system needs to maintain high-precision three-dimensional space measurement capability in an all-weather environment. However, in strong scattering aerosol environments such as dense fog, water clusters, rain and snow, physical degradation of the optical transmission channel occurs, resulting in nonlinear decoupling of the spatial and temporal dimensions of the characteristic data acquired by a single sensor. Specifically, mie scattering effect caused by the suspended particles can cause a laser radar receiving end to generate high-intensity near-field echo noise, a noise substrate can mask weak echo signals of a far-field target to cause blockage and distortion of a geometric ranging channel, and meanwhile, thermal radiation diffusion effect of an atmosphere medium can cause nonlinear halation expansion of a physical boundary of the target in infrared imaging to widen a point diffusion function of the infrared image and cannot reflect the accurate geometric profile of the target. The existing fusion sensing technology is generally based on the assumption of clean atmosphere, and relies on independent confidence of a single sensor for weighting, so that effective three-dimensional space constraint cannot be established under the working condition that the two sensors are limited by a physical rule and fail, and an accurate target positioning result cannot be output by a monitoring system under extreme meteorological conditions. Disclosure of Invention The invention provides a target detection and positioning method for laser radar point cloud and infrared image fusion, and aims to solve the technical problem that high-precision positioning cannot be realized due to physical decoupling of heterogeneous sensor characteristics in space-time and semantic dimensions caused by interference of strong scattering media. In view of the above problems, the invention provides a target detection and positioning method for laser radar point cloud and infrared image fusion, which is applied to a perception system configured with a space-time heterogeneous association controller, wherein the perception system comprises a full-waveform laser radar module, an infrared thermal imaging and shooting module and a history track storage module, and the method comprises the following steps: the method comprises the steps that S1, a space-time heterogeneous association controller calculates a target motion period according to data in a history track storage module and constructs a target expected arrival time window, and only when a system clock is in the target expected arrival time window, the full-waveform laser radar module and the infrared thermal imaging camera module are activated to execute synchronous acquisition to acquire original full-waveform echo data and original infrared thermal image data; Step S2, the space-time heterogeneous association controller performs feature conflict detection on the original infrared thermal image data and the original full-waveform echo data, judges that scattering medium interference exists when detecting that the infrared thermal radiation features and the near-field echo intensity of the laser radar have logic conflict, performs layering analysis on the original full-waveform echo data, and extracts secondary echo signals of a far-field area to generate a penetrability geometric depth vector; Step S3, the time-space heterogeneous association controller calculates the time domain gradient of the penetrability geometric depth vector in the scanning direction to determine a physical truth value edge point, constructs a geometric confidence attenuation field by utilizing the physical truth value edge point, performs morphological contraction operation on the original infrared thermal image data based on the geometric confidence attenuation field, eliminates a thermal diffusion halation region and generates an accurate semantic contour matrix; And S4, carrying out orthogonal fusion on the penetrability geometric depth vector and the accurate semantic profile matrix, and solving the three-dimensional space coordinate of the target. Further, the specific conditions for determining that the scattering medium interference exists in the step S2 are as follows: the space-time heterogeneous association controller identifies that a communication region higher than the background temperature exists in the original infrared thermal image data; And searching a laser radar beam corresponding to the communication area, and if the first echo peak intensity of the beam in the