Full-automatic production line coupling method for solid-state laser radar RX lens

Abstract

In order to solve the existing technical problem, the invention provides a solid-state laser radar RX lens full-automatic production line coupling method, which is divided into a decision-making layer and an execution mechanism, in the decision-making layer, Z-axis optimal position decision-making, principal point position decision-making and MTF maximum value position finding decision-making are performed in sequence, and after an optimal position point is obtained in the decision-making layer, the optimal position point is determined according to a displacement record. Z-axis optimal position coupling positioning, main point position coupling positioning and MTF maximum value position coupling positioning are sequentially carried out on an executing mechanism, after MTF maximum value position coupling positioning is finished, a decision layer verifies whether coupling passes or not, after verification passes, the executing mechanism carries out position curing through a UV lamp, after curing is finished, the executing mechanism quits, and coupling of the RX lens is completed. In the invention, the coupling precision can be improved along with the improvement of the precision of the execution layer, the decision-making layer and the execution mechanism belong to a separated architecture, so that the later production line equipment upgrading is facilitated, and the execution mechanism always limits the position in the curing process and releases the position limitation until the curing is finished. RX position offset caused by glue curing shrinkage and thermal drift is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of lidar manufacturing technology, specifically to a fully automated production line coupling method for solid-state lidar RX lenses. Background Technology

[0002] With the rapid development of autonomous driving, intelligent transportation, and robotic perception systems, LiDAR, as a core environmental perception sensor, has become crucial to industry competition in terms of performance, reliability, and mass production capabilities. Among them, solid-state LiDAR, due to its advantages such as no mechanical rotating parts, small size, long lifespan, and strong vibration resistance, is gradually replacing traditional mechanical LiDAR and becoming the mainstream technology for vehicle-mounted forward-facing main radar.

[0003] Currently, mainstream solid-state solutions generally use VCSEL (Vertical-Cavity Surface-Emitting Laser) arrays as the emission source, combined with SPAD (Single-Photon Avalanche Diode) arrays to achieve single-photon level detection. Three-dimensional imaging is achieved through Flash array illumination or OPA optical phased array scanning. VCSEL arrays feature low threshold, high modulation bandwidth, and the ability to be integrated in two dimensions, while SPADs offer single-photon sensitivity, high temporal resolution (<100 ps), and strong resistance to sunlight interference. The combination of these two technologies gives solid-state radar significant advantages in short-to-medium range, high-resolution detection.

[0004] However, the beam emitted by a VCSEL array has characteristics such as a large divergence angle, significant astigmatism, and poor inter-array consistency. Efficient collimation, astigmatism correction, and field-of-view shaping are required through a TX lens group to ensure uniform illumination, absence of dark areas, and low sidelobes within the SPAD receiving field of view. Since the SPAD array is extremely sensitive to the angle and spatial distribution of the incident light, the optical performance of the TX lens directly determines the system's signal-to-noise ratio (SNR), telemetry capability, angular resolution, and point cloud density, becoming one of the key bottlenecks of the entire system.

[0005] In existing technologies, the assembly and coupling of RX lenses mainly rely on manual or semi-automatic equipment, typically employing an "assemble first, then adjust" alignment strategy. Some manufacturers have introduced active alignment technology from the field of optical communication, adjusting the lens position by monitoring optical power feedback to improve coupling efficiency. However, these methods reveal the following significant problems when facing the high integration, multi-beam, large field of view, and automotive-grade reliability requirements of pure solid-state LiDAR TX lens assemblies:

[0006] 1. Insufficient Coupling Accuracy: Difficult to Meet Wavefront Alignment Requirements. Pure solid-state TX lenses typically require simultaneous collimation and shaping of thousands of light beams. Their optical performance evaluation criteria have been upgraded from the traditional "optimal single-point power" to "minimum wavefront error across the entire field of view." Existing coupling methods based on optical power detection can only reflect local light intensity distribution and cannot obtain overall wavefront information. This results in a large far-field divergence angle and poor sidelobe suppression in the coupled system, affecting radar rangefinding capability and angular resolution.

[0007] 2. Optical axis shift caused by adhesive curing shrinkage and thermal drift: Current coupling processes generally use UV adhesives or epoxy adhesives for lens fixation. During the curing process, the adhesive exhibits a shrinkage displacement of 1 μm to 3 μm. Furthermore, during subsequent automotive-grade temperature cycling (-40 ℃ to +85 ℃), the mismatch between the thermal expansion coefficient of the adhesive layer and the glass / metal structure causes the lens optical axis to drift by 10 µrad to 30 µrad, far exceeding the requirements of automotive-grade LiDAR for TX optical axis stability (<5 µrad).

[0008] 3. Low level of automation and poor consistency: Low yield. Existing semi-automatic coupling equipment relies on manual intervention for lens pre-positioning and glue volume control. The coupling cycle is long (>3 min per lens) and is easily affected by the operator's experience, resulting in poor coupling consistency between batches and a yield of less than 85%, which is difficult to meet the annual production capacity demand of millions of automotive LiDARs.

[0009] To address the aforementioned issues, this invention provides a fully automated production line coupling method for solid-state lidar RX lenses. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention provides a fully automated coupling method for solid-state lidar RX lenses on a production line. The method comprises a decision-making layer and an execution mechanism. In the decision-making layer, the optimal Z-axis position is determined sequentially, followed by the principal point position and the MTF maximum value. After the optimal position is determined, the execution mechanism sequentially performs Z-axis optimal position coupling, principal point position coupling, and MTF maximum value position coupling based on displacement records. After the MTF maximum value position coupling is completed, the decision-making layer verifies the coupling. If successful, the execution mechanism uses a UV lamp to solidify the position. After solidification, the execution mechanism releases the position constraint, completing the coupling of the RX lens. In this invention, the coupling accuracy can be improved with the accuracy of the execution layer. The decision-making layer and execution mechanism are separate structures, facilitating future production line equipment upgrades. Furthermore, the execution mechanism continuously constrains the position during the solidification process until solidification, effectively avoiding RX position shifts caused by adhesive curing shrinkage and thermal drift.

[0011] To achieve the above objectives, the present invention provides the following technical solution:

[0012] This invention provides a fully automated production line coupling method for solid-state lidar RX lenses, specifically including a decision-making layer and an execution mechanism. In the decision-making layer, the optimal Z-axis position decision, the principal point position decision, and the MTF maximum value decision are performed sequentially. After the optimal position point is obtained in the decision-making layer, the optimal Z-axis position coupling positioning, the principal point position coupling positioning, and the MTF maximum value position coupling positioning are performed sequentially in the execution mechanism based on the displacement record.

[0013] In the decision-making layer, the optimal Z-axis position is determined first. The decision algorithm uses the Sobel energy gradient algorithm, which is more sensitive to boundaries. The Sobel operator is used to calculate the horizontal and vertical gradients, and the sum of squared magnitudes is used as the sharpness. During the movement of the actuator, the decision-making layer collects the imaging data matrix of the current frame. The energy of a single data matrix is... Perform normalization processing, based on The energy value is mapped to a grayscale value G, resulting in a grayscale image I. Where Gx and Gy are the results of the Sobel convolution, and finally based on The response matrix L is obtained. The sharpness score is defined as the sum of L. The decision layer calculates the sharpness score in real time and plots the sharpness curve. After the peak appears, the first derivative will undergo a positive to negative transition. This transition point is recorded to complete the optimal Z-axis position decision.

[0014] The actuator locates the aforementioned jump point based on the displacement record, fixes the Z-axis position, and determines whether to defocus the Z-axis based on the device characteristics.

[0015] In the decision-making layer, after determining the optimal Z-axis position, the lens needs to be placed at the center of the imaging plane. Therefore, it is necessary to find the principal point position. Two equal and perpendicular black lines are pre-defined on the imaging plane, and their perpendicular intersection is the principal point. The decision-making layer reserves four diagonally selected areas in the XY direction. The threshold segmentation method is used to find the black vertical lines inside the areas. The diagonal selected areas of the four diagonally selected areas are the same length as the imaging plane. As long as the black line position is found in two of the four diagonally selected areas, the actuator needs to move in the direction of the centroid of the four diagonally selected areas according to the correspondence. After balancing the centroids of the four diagonally selected areas, the principal point position decision is completed.

[0016] The actuator places the RX lens within the principal position area based on the displacement record.

[0017] In the decision layer, after the principal point position is determined, the decision algorithm adopts the MTF value maximization method. Starting from the principal point position, it calculates the MTF values ​​of the four diagonal regions and one region around the principal point. When all five selected areas satisfy MTF>=0.7, the position of the maximum MTF value is determined.

[0018] After the MTF maximum value position is determined, the actuator confirms the final position as the MTF maximum value position point, turns on the UV lamp to cure the lens, and after the curing operation is completed, the actuator exits, acquires a raw frame, judges the MTF value error after curing, and terminates the coupling after the judgment is passed. Attached Figure Description

[0019] Figure 1 This is a general framework diagram of the present invention. Detailed Implementation

[0020] This invention provides a fully automated coupling method for solid-state lidar RX lenses on a production line. It consists of a decision-making layer and an execution mechanism. In the decision-making layer, the optimal Z-axis position is determined sequentially, followed by the principal point position and the MTF maximum value. After the optimal position is determined, the execution mechanism sequentially performs Z-axis optimal position coupling, principal point position coupling, and MTF maximum value position coupling based on displacement records. After the MTF maximum value position coupling is completed, the decision-making layer verifies the coupling. If successful, the execution mechanism uses a UV lamp to solidify the position. After solidification, the execution mechanism releases the position constraint, completing the coupling of the RX lens. In this invention, the coupling accuracy can be improved with the accuracy of the execution layer. The decision-making layer and execution mechanism are separate, facilitating future production line equipment upgrades. Furthermore, the execution mechanism continuously constrains the position during solidification until it is released, effectively avoiding RX position shift caused by adhesive curing shrinkage and thermal drift.

[0021] To achieve the above objectives, the present invention provides the following technical solution:

[0022] This invention provides a fully automated production line coupling method for solid-state lidar RX lenses, specifically including a decision-making layer and an execution mechanism. In the decision-making layer, the optimal Z-axis position decision, the principal point position decision, and the MTF maximum value decision are performed sequentially. After the optimal position point is obtained in the decision-making layer, the optimal Z-axis position coupling positioning, the principal point position coupling positioning, and the MTF maximum value position coupling positioning are performed sequentially in the execution mechanism based on the displacement record.

[0023] In the decision-making layer, the optimal Z-axis position is determined first. The decision algorithm uses the Sobel energy gradient algorithm, which is more sensitive to boundaries. The Sobel operator is used to calculate the horizontal and vertical gradients, and the sum of squared magnitudes is used as the sharpness. During the movement of the actuator, the decision-making layer collects the imaging data matrix of the current frame. The energy of a single data matrix is... Perform normalization processing, based on The energy value is mapped to a grayscale value G, resulting in a grayscale image I. Where Gx and Gy are the results of the Sobel convolution, and finally based on The response matrix L is obtained. The sharpness score is defined as the sum of L. The decision layer calculates the sharpness score in real time and plots the sharpness curve. After the peak appears, the first derivative will undergo a positive to negative transition. This transition point is recorded to complete the optimal Z-axis position decision.

[0024] The actuator locates the aforementioned jump point based on the displacement record, fixes the Z-axis position, and determines whether to defocus the Z-axis based on the device characteristics.

[0025] In the decision-making layer, after determining the optimal Z-axis position, the lens needs to be placed at the center of the imaging plane. Therefore, it is necessary to find the principal point position. Two equal and perpendicular black lines are pre-defined on the imaging plane, and their perpendicular intersection is the principal point. The decision-making layer reserves four diagonally selected areas in the XY direction. The threshold segmentation method is used to find the black vertical lines inside the areas. The diagonal selected areas of the four diagonally selected areas are the same length as the imaging plane. As long as the black line position is found in two of the four diagonally selected areas, the actuator needs to move in the direction of the centroid of the four diagonally selected areas according to the correspondence. After balancing the centroids of the four diagonally selected areas, the principal point position decision is completed.

[0026] The actuator places the RX lens within the principal position area based on the displacement record.

[0027] In the decision layer, after the principal point position is determined, the decision algorithm adopts the MTF value maximization method. Starting from the principal point position, it calculates the MTF values ​​of the four diagonal regions and one region around the principal point. When all five selected areas satisfy MTF>=0.7, the position of the maximum MTF value is determined.

[0028] After the MTF maximum value position is determined, the actuator confirms the final position as the MTF maximum value position point, turns on the UV lamp to cure the lens, and after the curing operation is completed, the actuator exits, acquires a raw frame, judges the MTF value error after curing, and terminates the coupling after the judgment is passed.

[0029] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fully automated production line coupling method for solid-state lidar RX lenses, characterized in that, Specifically, it includes a decision-making layer and an execution mechanism. In the decision-making layer, the optimal Z-axis position decision, the principal point position decision, and the MTF maximum value decision are performed sequentially. After the optimal position point is obtained in the decision-making layer, the execution mechanism performs the optimal Z-axis position coupling positioning, the principal point position coupling positioning, and the MTF maximum value position coupling positioning sequentially based on the displacement record.

Images