A refrigerator camera adjusting system and method based on remote monitoring

Abstract

This invention discloses a remote monitoring-based camera adjustment system and method for refrigerators, relating to the field of camera device control technology in remote monitoring of refrigerators. The system includes: constructing a first compensation dataset; responding to a remotely input gimbal target angle command, querying the first compensation dataset to obtain the optical path offset corresponding to the target angle, and calculating the compensation deflection; obtaining the execution drive angle; after the gimbal actuator rotates to its position, determining the target depth value corresponding to the field of view center, retrieving a preset object distance-focal length mapping relationship, and generating a first drive displacement; controlling the gimbal drive device to move to the target focusing position; the gimbal drive device performing an over-stepping action along a preset direction, and then reversing by an equal step, ensuring that the meshing surfaces within the transmission mechanism remain in contact in a single direction; simultaneously performing a hard zeroing operation to refresh the absolute coordinate reference zero point of the pre-compensation dataset. This invention significantly shortens the focusing convergence time.

Description

Technical Field

[0001] This invention relates to the field of camera device control technology in remote monitoring of freezers, specifically a freezer camera adjustment system and method based on remote monitoring. Background Technology

[0002] With the rapid development of cold chain logistics and unmanned retail, remote video monitoring inside refrigerated display cases has become a standard feature for inventory management, anti-theft monitoring, and equipment operation status inspection. Existing refrigerated display case monitoring cameras are typically installed on the top of the display case or inside the glass door, equipped with a two-axis pan-tilt actuator to remotely adjust the horizontal azimuth and vertical tilt angles. They also feature independent focusing motors to drive the internal lens assembly for automatic or semi-automatic focusing.

[0003] In terms of gimbal control, existing technologies generally use an image sharpness evaluation function as a feedback signal, driving the gimbal motor through closed-loop control until the image sharpness meets a preset threshold. For focus control, existing technologies generally employ a "hill-climbing search method" or its improved version, which involves driving the focus motor to move tentatively in both forward and reverse directions, comparing the sharpness evaluation values ​​at each position, and gradually approaching the extreme point of the evaluation function. Furthermore, to avoid position drift caused by gear backlash after long-term operation of the stepper motor, some existing systems perform a gimbal zeroing operation every time they are powered on, calibrating the mechanical zero point by running to the physical limit switch.

[0004] The aforementioned existing technical solutions can basically meet the needs of use in conventional indoor monitoring scenarios, but when applied to the special enclosed scenario of a freezer, the following shortcomings are exposed:

[0005] Firstly, the inner surface of the transparent protective layer of the freezer is covered with a dense layer of condensation droplets or ice crystal particles due to the interaction between the low temperature and the external temperature and humidity. This dielectric layer forms a random microlens array. When the pan-tilt angle changes, the incident angle of light passing through the micro-curved surface of the droplets changes accordingly, resulting in an additional optical path offset that is non-linear with the angle. However, existing pan-tilt control methods only adjust based on image sharpness feedback, ignoring the actual impact of this optical path offset on the imaging position. This leads to a significant lateral drift in the actual imaging position of the target object on the CMOS imager after the pan-tilt physically rotates into place.

[0006] Secondly, the imaging beam passing through the microlens array of condensed water droplets undergoes wavefront distortion, causing the sharpness evaluation function curve to exhibit multiple local spurious maxima within the effective depth of focus. Under these circumstances, the traditional hill-climbing search method cannot distinguish between the true focal plane and the spurious peaks. The focusing motor repeatedly attempts to converge between multiple spurious peaks, resulting in long focusing convergence times or even failure to converge. At the same time, the frequent reversals of the focusing motor exacerbate the wear of the transmission pair.

[0007] Third, the frequent forward and reverse rotation of the gimbal motor leads to the accumulation of meshing clearance in the worm gear transmission pair, causing a continuous increase in the deviation between the encoder reading and the actual mechanical angle. The drift of the gimbal angle directly changes the light incident angle condition, rendering the calibrated optical path offset-angle correspondence invalid. The gimbal error and focusing error are amplified through cross-coupling via the optical path, and the existing system only performs a hard zeroing once upon power-up, which cannot eliminate the real-time accumulated error caused by frequent reversals within a single operating cycle.

[0008] In summary, the existing technology's approach of separating gimbal angle control from focal length control results in a situation where, in remote control scenarios involving high humidity and low temperature in freezers, the optical coupling effect of the condensation medium layer prevents the three core indicators—positioning accuracy, focusing speed, and long-term stability—from simultaneously meeting engineering requirements. Summary of the Invention

[0009] The purpose of this invention is to provide a remote monitoring-based freezer camera adjustment system and method to solve the problems raised in the prior art.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method for adjusting a freezer camera based on remote monitoring, specifically including the following steps:

[0011] When the freezer is in a set working condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle node and the corresponding optical path offset are associated and stored to construct the first compensation dataset.

[0012] In response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and the compensation deflection is calculated based on the optical path offset; the combined result of the target angle and the compensation deflection is used as the execution drive angle to control the gimbal actuator to perform a one-time position rotation.

[0013] After the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined based on the current gimbal actuator. The preset object distance-focal length mapping relationship is retrieved based on the target depth value to generate a first driving displacement. The gimbal drive device is then controlled to move to the target focusing position corresponding to the first driving displacement.

[0014] Upon completion of each focus lock command or each power-on initialization, the gimbal drive device is controlled to perform an overstepping action in a preset direction, and then reverses to perform an equal step, so that the meshing surfaces in the transmission mechanism remain in contact in a single direction; at the same time, a hard zeroing operation is performed to refresh the absolute coordinate reference zero point of the pre-compensation dataset.

[0015] Furthermore, under actual temperature and humidity conditions inside the freezer, the optical path offset of light rays at different incident angles is unknown due to condensation media such as water droplets or ice crystals adhering to the surface of the transparent protective layer made of glass door. This makes it impossible to establish an accurate input-output mapping model during subsequent dynamic adjustment. This step quantifies the physical disturbance into deterministic data through offline calibration, providing a priori benchmark for subsequent open-loop pre-compensation.

[0016] With the freezer in a set operating condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle nodes are associated with the corresponding optical path offsets and stored to construct the first compensation dataset, specifically:

[0017] Once the freezer reaches the set stable operating conditions and the internal target remains stationary, perform an offline calibration.

[0018] The gimbal actuator is controlled to traverse the entire mechanical travel range of its horizontal azimuth and vertical pitch angles, and values ​​are taken according to a fixed sampling step size. At each discrete angle node, an airborne near-infrared ranging spot is projected onto the target. By receiving the reflected echo and calculating the transmission-reception time difference, the total round-trip time of the actual light from the front lens of the camera to the target at that angle is measured, and the difference is made with the theoretical travel time in vacuum to obtain the optical path offset at that discrete angle node.

[0019] The horizontal azimuth angle, vertical pitch angle, and optical path offset corresponding to the discrete angle node are used as a triplet and written into a two-dimensional data table in non-volatile memory to form the first compensation dataset.

[0020] Furthermore, in response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and a compensation deflection amount for compensating for optical path refraction deviation is calculated based on the optical path offset; the combined result of the target angle and the compensation deflection amount is used as the execution drive angle to control the gimbal actuator to perform a one-time position rotation, specifically:

[0021] Based on the remotely input gimbal target angle, the bilinear interpolation model is invoked to calculate the optical path offset corresponding to the gimbal target angle;

[0022] Based on the difference approximation of Fermat's principle, the amount of compensation deflection required to offset the refraction deviation at the current viewing angle is calculated.

[0023] Instead of directly driving to the target angle of the gimbal, the calculated and synthesized driving angle is used; the pulse width modulation driver is controlled to drive the gimbal actuator to rotate continuously into position in one go with the synthesized driving angle as the target.

[0024] In one embodiment of the present invention, the target horizontal azimuth angle is denoted as α. x The target's vertical pitch angle is β y The target's horizontal azimuth angle and vertical pitch angle together constitute the gimbal target angle (α). tgt ,β tgt );

[0025] The partial derivative of the optical path offset relative to the target's horizontal azimuth angle can be obtained as ∂ α Δ(α x ,β y The partial derivative of the vertical pitch angle relative to the target is ∂ β Δ(α x ,β y );

[0026] It should be noted that, due to Δ(α) x ,β y The interpolation between discrete angle nodes is linear, and its partial derivatives can be accurately calculated through the differences between neighboring nodes; taking the target horizontal azimuth as an example, if α x Located at node α i With α i+1 Between, then: ∂ α Δ(α x ,β y )=(Δ i+1,j -Δ i,j ) / Δα; where Δ i+1,j Indicates at the angle node (α) i+1 ,β j The optical path offset measured at ( ) point;

[0027] Optionally, when the gimbal target angle falls on the boundary, a one-sided difference is taken; similarly, the partial derivative of the optical path offset relative to the target's vertical pitch angle can be obtained as ∂ β Δ(α x ,β y );

[0028] According to the optical path perturbation theorem in geometric optics, the mechanical compensation angle is proportional to the optical path gradient.

[0029] Let the horizontal compensation coefficient be K. α The vertical compensation coefficient is K. β Then the compensation deflection amount (δ) α ,δ β Characterized as:

[0030] δα =-K α ∙∂ α Δ(α x ,β y );δ β =-K β ∙∂ β Δ(α x ,β y );

[0031] Calculate the horizontal execution drive angle: α cmd =α x +δ α Vertical execution drive angle: β cmd =β y +δ β ;

[0032] The synthetic execution driving angle is (α) cmd ,β cmd ).

[0033] The synthesized drive angle is sent to the stepper microstepper of the gimbal actuator; if the step angle of the gimbal stepper motor is denoted as θ step The number of pulses to be output in the horizontal direction is scalar P. α for:

[0034] P α =round[(α cmd -α cur ) / θ step In the formula, α cur Indicates the current horizontal azimuth angle; round indicates rounding to the nearest integer pulse.

[0035] Similarly, the scalar P representing the number of pulses to be output in the vertical direction can be obtained. β ;

[0036] Drive the gimbal actuator according to the number of pulses P α and P β Direct drive enables one-step completion.

[0037] It should be noted that the horizontal compensation coefficient K α With vertical compensation coefficient K β The values ​​are calculated from the equivalent focal length of the camera lens assembly and the thickness of the transparent protective layer. These are all factory-fixed known quantities and are considered constants here.

[0038] This step adjusts the gimbal control from a mechanical angle closed loop to an effective optical angle closed loop, ensuring that even if the mechanical angle closed loop fails to reach the target angle, the optical target object is precisely centered in the field of view. This open-loop, one-step method further compresses the gimbal response time and eliminates overshoot and oscillations inherent in traditional closed-loop control.

[0039] Furthermore, after the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined based on the current pitch attitude of the gimbal actuator. A preset object distance-focal length mapping relationship is retrieved based on the target depth value to generate a first driving displacement of the focusing lens group. The drive motor is controlled to move from the current focusing position along a single direction to the target focusing position corresponding to the first driving displacement, and upon arrival, a one-way cutoff lock is executed. That is, only one-way fine-tuning or direct locking is performed in the moving direction, prohibiting retraction in the opposite direction. Specifically:

[0040] After the gimbal actuator rotates to its position and comes to a complete stop, read the current vertical pitch angle feedback value of the gimbal.

[0041] Based on the vertical installation height from the gimbal rotation center to the shelf plane, the target depth value between the current optical axis and the shelf plane is calculated using trigonometric functions.

[0042] Based on the Gaussian imaging formula, the object distance-focusing motor step count correspondence table is pre-calibrated at the factory.

[0043] The first driving displacement is obtained by looking up the calculated target depth value in a table.

[0044] The focusing motor is instructed to advance from its current position along a preset direction to the first driving displacement. Upon reaching the first displacement, it takes one more microstep forward from that position and compares the high-frequency energy values ​​of the image before and after taking that microstep. If the energy value increases, it continues to advance one microstep forward. If the energy value decreases or remains unchanged, the focusing lock command is immediately executed to prevent the motor from reversing and the current position is recorded as the focusing position.

[0045] The preset direction is uniformly set to the direction away from the camera lens, that is, the direction of increasing focus steps.

[0046] This step replaces blind image search with triangulation, which greatly reduces focusing time; forced unidirectional motion completely avoids nonlinear hysteresis caused by gear reversal backlash, and eliminates the possibility of hovering between multiple pseudo-extremes, greatly improving the one-time success rate of focusing.

[0047] Furthermore, upon completion of each focus lock command or power-on initialization, the drive motor is controlled to perform an over-stepping action along a preset fixed direction, followed by a reverse retraction of an equal amount of steps, ensuring that the meshing surfaces within the transmission mechanism remain in contact in a single direction. Simultaneously, the gimbal actuator is periodically controlled to run to the physical limit switch to perform a hard zeroing operation, thereby refreshing the absolute coordinate reference zero point of the pre-compensation dataset. Specifically:

[0048] After each focus lock command is completed and before the gimbal enters standby power-saving mode, the drive motor is controlled to output an additional pulse train exceeding the target angle along the preset direction. The number of pulses is twice the number of pulses corresponding to the maximum backlash of the transmission gear set as determined in advance. Then, the motor is immediately controlled to retract an equal number of pulses in the opposite direction.

[0049] This back-and-forth motion ensures that the gear set always comes into close contact with the meshing surface from the same side, eliminating the backlash on both sides.

[0050] During each system power-on initialization, the control gimbal actuator first moves at a low speed toward the positive limit switch in the horizontal direction. After triggering a level change, the position is marked as 0. Then it moves toward the negative limit switch in the vertical direction. After triggering a level change, the position is marked as 0.

[0051] The calibrated position is recorded as the new physical zero point, and the origin of the stored first compensation dataset is realigned with the new physical zero point.

[0052] A remote monitoring-based freezer camera control system includes:

[0053] The calibration and storage module, whose input is connected to the ranging unit of the camera device, is used to traverse the rotatable angle range of the gimbal actuator under the set working conditions of the freezer, measure and record the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer at each discrete angle node, and associate each angle node with the corresponding optical path offset to store as the first compensation dataset.

[0054] An angle pre-compensation drive module has its input terminals connected to the output terminal of the calibration and storage module and the remote command receiving terminal, respectively. It is used to respond to the target angle command input remotely, query the first compensation dataset to obtain the corresponding optical path offset, calculate the compensation deflection amount accordingly, and use the combined result of the target angle and the compensation deflection amount as the execution drive angle to drive the gimbal actuator to perform a one-time position rotation. Its control output terminal is connected to the gimbal drive device.

[0055] The depth linkage focusing module has its input end connected to the angle feedback output end of the gimbal drive device. After the gimbal drive device rotates to the position, it calculates the target depth value based on the current pitch angle feedback value, and generates the first driving displacement of the focusing lens group according to the preset object distance-focal length mapping relationship. It drives the gimbal focusing device to move from the current position to the target focusing position in a single direction, and performs a one-way cut-off lock after reaching the target position. Its control output end is connected to the gimbal focusing device.

[0056] The backlash elimination and reference maintenance module has its input terminal connected to the system power-on signal terminal and the locking completion signal output terminal of the depth linkage focusing module. Its control output terminal is connected to the gimbal focusing device and the gimbal driving device respectively. It is used to control the gimbal focusing device to perform an over-step action along a preset fixed direction and then reverse and retreat by an equal amount of steps after each focus locking is completed, so that the meshing surface of the transmission mechanism is kept in contact in a single direction. It is also used to control the gimbal driving device to run to the physical limit switch to perform a hard zeroing operation when the system is powered on, so as to refresh the absolute coordinate reference zero point of the pre-compensation dataset.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] 1. This invention transforms the complex refraction disturbance of the condensation medium layer at different incident angles into a fixed two-dimensional lookup table physical prior; when responding to remote target angle commands, it does not rely on real-time image feedback to drive the gimbal to approach the target, but instead calculates the compensation deflection required to offset the optical path refraction deviation, and directly drives the gimbal actuator to reach the pre-compensated synthetic mechanical angle in one step; this enables the gimbal to achieve a one-time accurate positioning of the effective optical angle under any thickness distribution of the condensation medium layer, significantly shortening the response time and being unaffected by changes in the density and distribution of condensation droplets;

[0059] 2. Based on the physical feedback value of the current pitch angle of the gimbal, the present invention directly calculates the target depth value at the center of the current field of view using fixed installation geometric parameters, and obtains the theoretical displacement of the focusing lens group by looking up the table. The present invention stipulates that the focusing motor advances to the target position only in a single direction and performs unidirectional cutoff locking after reaching the target position. This strategy completely avoids the multi-peak pseudo-extreme trap caused by the condensation medium layer, so that the focusing convergence time is stable at the millisecond level of a single stroke of the motor, and there is no reciprocating oscillation wear.

[0060] 3. This invention designs differentiated mechanical error elimination strategies for the gimbal motor and the focusing motor respectively. For the focusing motor, it eliminates the backlash uncertainty caused by occasional retraction requirements after frequent unidirectional advance. For the gimbal motor, it ensures that the gimbal angle drift error and the focusing screw backlash error are clamped within their respective reference domains, and the two no longer generate cross-coupling amplification through the optical path, significantly extending the maintenance-free cycle. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a remote monitoring-based method for adjusting a freezer camera according to the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example: Figure 1 As shown, the present invention provides a technical solution, which specifically includes the following steps:

[0064] When the freezer is in a set working condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle node and the corresponding optical path offset are associated and stored to construct the first compensation dataset.

[0065] In response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and the compensation deflection is calculated based on the optical path offset; the combined result of the target angle and the compensation deflection is used as the execution drive angle to control the gimbal actuator to perform a one-time position rotation.

[0066] After the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined based on the current gimbal actuator. The preset object distance-focal length mapping relationship is retrieved based on the target depth value to generate a first driving displacement. The gimbal drive device is then controlled to move to the target focusing position corresponding to the first driving displacement.

[0067] Upon completion of each focus lock command or each power-on initialization, the gimbal drive device is controlled to perform an overstepping action in a preset direction, and then reverses to perform an equal step, so that the meshing surfaces in the transmission mechanism remain in contact in a single direction; at the same time, a hard zeroing operation is performed to refresh the absolute coordinate reference zero point of the pre-compensation dataset.

[0068] In this embodiment, the gimbal focusing device is a focusing motor; the gimbal driving device is a drive motor.

[0069] Furthermore, under actual temperature and humidity conditions inside the freezer, the optical path offset of light rays at different incident angles is unknown due to condensation media such as water droplets or ice crystals adhering to the surface of the transparent protective layer made of glass door. This makes it impossible to establish an accurate input-output mapping model during subsequent dynamic adjustment. This step quantifies the physical disturbance into deterministic data through offline calibration, providing a priori benchmark for subsequent open-loop pre-compensation.

[0070] With the freezer in a set operating condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle nodes are associated with the corresponding optical path offsets and stored to construct the first compensation dataset, specifically:

[0071] Once the freezer reaches the set stable operating conditions and the internal target remains stationary, perform an offline calibration.

[0072] The gimbal actuator is controlled to traverse the entire mechanical travel range of its horizontal azimuth and vertical pitch angles, and values ​​are taken according to a fixed sampling step size. At each discrete angle node, an airborne near-infrared ranging spot is projected onto the target. By receiving the reflected echo and calculating the transmission-reception time difference, the total round-trip time of the actual light from the front lens of the camera to the target at that angle is measured, and the difference is made with the theoretical travel time in vacuum to obtain the optical path offset at that discrete angle node.

[0073] The horizontal azimuth angle, vertical pitch angle, and optical path offset corresponding to the discrete angle node are used as a triplet and written into a two-dimensional data table in non-volatile memory to form the first compensation dataset.

[0074] In one embodiment of the present invention, the stable operating conditions set for the freezer are denoted as preset temperature T0 and preset humidity H0;

[0075] T0 and H0 can be set according to the actual situation.

[0076] Let the mechanically adjustable range of the horizontal azimuth angle be [α]. min ,α max The mechanically adjustable range of the vertical pitch angle is [β]. min ,β max ]; where α min and α max These represent the upper and lower adjustable limits of the horizontal azimuth angle of the gimbal actuator, respectively; β min and β max These represent the upper and lower adjustable limits of the vertical pitch angle of the gimbal actuator, respectively.

[0077] Discretize the above interval with fixed step sizes Δα and Δβ, and denote the number of angle nodes sampled as M and N, respectively;

[0078] Discrete angle sequences were collected separately, and the horizontal azimuth sequence was denoted as A={α1,α2,…,α...}. i ,…,α M}, where α i =α min+(i-1)∙Δα, where i is the horizontal angle node index number, i=1,2,…,M; α i This represents the i-th horizontal angle scalar value in the sequence; Δα represents the minimum sampling interval of the horizontal azimuth angle.

[0079] Vertical pitch angle sequence B={β1,β2,…,β j ,…,β N}, where β j =β min +(j-1)∙Δβ, where j is the index number of the vertical angle node, j=1,2,…,N; β j This represents the j-th vertical angle scalar value in the sequence; Δβ represents the minimum sampling interval for the vertical pitch angle.

[0080] At angle node (α) i ,β j The optical path offset measured at point ) is denoted as Δ. i,j The unit is millimeters; the first compensation dataset is a two-dimensional matrix D with dimension M×N, represented as:

[0081] ; where Δ M,N Represents the angle node (α) M ,β N The optical path offset measured at ( ) point;

[0082] It should be noted that Δ i,j The subscript i corresponds to the i-th element of the horizontal azimuth sequence, and the subscript j corresponds to the j-th element of the vertical pitch sequence.

[0083] Preferably, to support querying any non-discrete target angle, a bilinear interpolation model is defined; for any target horizontal azimuth angle α x Vertical pitch angle β of the target y When α i <α x <α i+1 And β j <β y <β j+1 Its corresponding optical path offset scalar Δ(α) x ,β y The computational representation is as follows:

[0084] First, define the horizontal weight λ. α =(α x -α i ) / (α i+1 -α i Vertical weight λ β =(β y -βj ) / (β j+1 -β j Then: Δ(α) x ,β y )=(1-λ α )(1-λ β )∙Δ i,j +λ α (1-λ β )∙Δ i+1,j +λ β (1-λ α )∙Δ i,j+1 +λ α λ β ∙Δ i+1,j+1 In the formula, Δ i+1,j+1 Indicates at the angle node (α) i+1 ,β j+1 The optical path offset measured at ( ) point;

[0085] The bilinear interpolation function described above ensures a smooth mapping from a discrete matrix to a continuous angular domain. Interpolation can be performed using other methods such as linear interpolation or nonlinear fitting, which are not limited here.

[0086] It is important to note that the calibration process must be performed only after the freezer compressor is running stably and the condensation on the glass door surface has reached dynamic equilibrium, to ensure Δ i,j Numerical stability and repeatability; when based on the bilinear interpolation function of this application, the sampling step size Δα and Δβ should not be greater than three times the minimum controllable step angle of the gimbal actuator to ensure interpolation accuracy.

[0087] Furthermore, when a remote command is issued to the target angle, if the gimbal is directly driven to rotate to that mechanical angle, the direction of light refraction through the attached medium layer will change due to the change in the incident angle. The actual imaging position of the target object will drift laterally and deviate from the CMOS center field of view. By reverse pre-deflection, the light can be made to fall exactly on the target after refraction, achieving precise positioning.

[0088] In response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and a compensation deflection amount for compensating for optical path refraction deviation is calculated based on the optical path offset; the combined result of the target angle and the compensation deflection amount is used as the execution drive angle to control the gimbal actuator to perform a one-time position rotation, specifically:

[0089] Based on the remotely input gimbal target angle, the bilinear interpolation model is invoked to calculate the optical path offset corresponding to the gimbal target angle;

[0090] Based on the difference approximation of Fermat's principle, the amount of compensation deflection required to offset the refraction deviation at the current viewing angle is calculated.

[0091] Instead of directly driving to the target angle of the gimbal, the calculated and synthesized driving angle is used; the pulse width modulation driver is controlled to drive the gimbal actuator to rotate continuously into position in one go with the synthesized driving angle as the target.

[0092] In one embodiment of the present invention, the target horizontal azimuth angle is denoted as α. x The target's vertical pitch angle is β y The target's horizontal azimuth angle and vertical pitch angle together constitute the gimbal target angle (α). tgt ,β tgt );

[0093] The partial derivative of the optical path offset relative to the target's horizontal azimuth angle can be obtained as ∂ α Δ(α x ,β y The partial derivative of the vertical pitch angle relative to the target is ∂ β Δ(α x ,β y );

[0094] It should be noted that, due to Δ(α) x ,β y The interpolation between discrete angle nodes is linear, and its partial derivatives can be accurately calculated through the differences between neighboring nodes; taking the target horizontal azimuth as an example, if α x Located at node α i With α i+1 Between, then: ∂ α Δ(α x ,β y )=(Δ i+1,j -Δ i,j ) / Δα; where Δ i+1,j Indicates at the angle node (α) i+1 ,β j The optical path offset measured at ( ) point;

[0095] Optionally, when the gimbal target angle falls on the boundary, a one-sided difference is taken; similarly, the partial derivative of the optical path offset relative to the target's vertical pitch angle can be obtained as ∂ β Δ(α x ,β y );

[0096] According to the optical path perturbation theorem in geometric optics, the mechanical compensation angle is proportional to the optical path gradient.

[0097] Let the horizontal compensation coefficient be K. α The vertical compensation coefficient is K. β Then the compensation deflection amount (δ)α ,δ β Characterized as:

[0098] δ α =-K α ∙∂ α Δ(α x ,β y );δ β =-K β ∙∂ β Δ(α x ,β y );

[0099] Calculate the horizontal execution drive angle: α cmd =α x +δ α Vertical execution drive angle: β cmd =β y +δ β ;

[0100] The synthetic execution driving angle is (α) cmd ,β cmd ).

[0101] The synthesized drive angle is sent to the stepper microstepper of the gimbal actuator; if the step angle of the gimbal stepper motor is denoted as θ step The number of pulses to be output in the horizontal direction is scalar P. α for:

[0102] P α =round[(α cmd -α cur ) / θ step In the formula, α cur Indicates the current horizontal azimuth angle; round indicates rounding to the nearest integer pulse.

[0103] Similarly, the scalar P representing the number of pulses to be output in the vertical direction can be obtained. β ;

[0104] Drive the gimbal actuator according to the number of pulses P α and P β Direct drive enables one-step completion.

[0105] It should be noted that the horizontal compensation coefficient K α With vertical compensation coefficient K β The values ​​are calculated from the equivalent focal length of the camera lens assembly and the thickness of the transparent protective layer. These are all factory-fixed known quantities and are considered constants here.

[0106] It is important to note that this step must be strictly performed in an open-loop manner, meaning that no real-time correction signal should be inserted during the operation of the motor due to a brief blurring of the image; the horizontal and vertical compensation coefficients need to be finely adjusted and corrected by a calibration test at a single standard angle before leaving the factory to accommodate differences between lens batches.

[0107] This step adjusts the gimbal control from a mechanical angle closed loop to an effective optical angle closed loop, ensuring that even if the mechanical angle closed loop fails to reach the target angle, the optical target object is precisely centered in the field of view. This open-loop, one-step method further compresses the gimbal response time and eliminates overshoot and oscillations inherent in traditional closed-loop control.

[0108] Furthermore, after the gimbal rotates, the object distance changes significantly due to the tilt of the optical axis. If traditional hill-climbing search focusing is initiated at this time, the evaluation function curve caused by the attached medium layer will exhibit multiple spurious maxima, leading to the focusing motor repeatedly probing in both directions. This is not only extremely time-consuming but also rapidly wears down the mechanical transmission components. This step predicts the theoretical focal length through geometric depth calculation and enforces the unidirectional movement of the motor, thereby cutting off oscillations, reducing operation time, and maintaining the mechanical transmission components.

[0109] After the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined based on the current pitch attitude of the gimbal actuator. A preset object distance-focal length mapping relationship is retrieved based on the target depth value to generate the first driving displacement of the focusing lens group. The drive motor is controlled to move from the current focusing position along a single direction to the target focusing position corresponding to the first driving displacement, and upon arrival, a one-way cutoff lock is executed. That is, only one-way fine-tuning or direct locking is performed in the moving direction, prohibiting retraction in the opposite direction. Specifically:

[0110] After the gimbal actuator rotates to its position and comes to a complete stop, read the current vertical pitch angle feedback value of the gimbal.

[0111] Based on the vertical installation height from the gimbal rotation center to the shelf plane, the target depth value between the current optical axis and the shelf plane is calculated using trigonometric functions.

[0112] Based on the Gaussian imaging formula, the object distance-focusing motor step count correspondence table is pre-calibrated at the factory.

[0113] The first driving displacement is obtained by looking up the calculated target depth value in a table.

[0114] The focusing motor is instructed to advance from its current position along a preset direction to the first driving displacement. Upon reaching the first displacement, it takes one more microstep forward from that position and compares the high-frequency energy values ​​of the image before and after taking that microstep. If the energy value increases, it continues to advance one microstep forward. If the energy value decreases or remains unchanged, the focusing lock command is immediately executed to prevent the motor from reversing and the current position is recorded as the focusing position.

[0115] The preset direction is uniformly set to the direction away from the camera lens, that is, the direction of increasing focus steps.

[0116] In one embodiment of the present invention, after the gimbal actuator has rotated to its position and come to a complete stop, the current vertical pitch angle feedback value β of the gimbal is read by a potentiometer / encoder. fb ; β is defined when the horizontal viewing angle is specified fb =0, downward tilt is considered a negative angle;

[0117] Let H be the vertical installation height from the center of rotation of the gimbal to the plane of the shelf. mount ;

[0118] Let the target depth value between the current optical axis and the shelf plane be Z, and take the unit as millimeters;

[0119] Based on the tangent relationship of a right triangle, the target depth value Z between the current optical axis and the shelf plane is uniquely determined as follows:

[0120] Z=H mount / tan(|β fb |+θ0); where θ0 represents the fixed installation angle correction term between the camera optical axis and the zero position of the gimbal rotation mechanism;

[0121] Let the factory-calibrated object distance-focusing motor step count correspondence table be {S1,S2,…,S...} K The corresponding object distance sequence is {Z1, Z2, ..., Z}. K};Z K S represents the depth value of the Kth target; K Z represents K The corresponding driving displacement;

[0122] For the calculated continuous target depth value Z, when Z can correspond to any target depth value calibrated by the factory, the corresponding driving displacement is obtained directly by looking up the table as the first driving displacement S.

[0123] Optionally, when Z q <Z <Z q+1 Then, the first driving displacement is obtained by linear interpolation: S = S q +[(ZZ q ) / (Z q+1 -Z q )]∙(Sq+1 -S q );in the formula, q represents the element serial number of the object distance-focusing motor step number correspondence table, q∈[1,K]; K represents the total number of elements of the object distance-focusing motor step number correspondence table; S q represents the q-th element in the object distance-focusing motor step number correspondence table;

[0124] the current actual position of the focusing motor is recorded as S cur ;

[0125] the system compares S with S cur , when S>S cur , the forward target step number S is recorded as target =S; if S≤S cur , then in order to enforce the forward-only unidirectional principle, the system does not adopt retraction, but sets the forward target step number to S target =S cur +S0, where S0 is a preset forward safety offset constant, and then the motor advances forward to S in one go target .

[0126] After reaching S target , the process enters a unidirectional cutoff verification phase, and a micro-step perturbation increment g is defined, where g is a very small positive scalar;

[0127] the focusing motor moves forward by g steps, and calculates the image high-frequency energy values E1 and E2 before and after the stepping; if E2≥E1, then let S target =S target +g, and repeat the iteration; if E2<E1, immediately terminate and lock, and the final locked position scalar S final is taken as the current S target value.

[0128] Throughout the whole process, the focusing motor only outputs forward pulses, and its pulse direction control signal is always at high level / low level and never flips.

[0129] It should be noted that the vertical installation height from the pan-tilt rotation center to the shelf plane must be accurately measured; if there are multiple shelf layers in the freezer, the height of the target layer currently aimed at by the camera is taken; the micro-step increment g in the unidirectional cutoff verification must be smaller than the dead zone threshold of the focusing lens group, that is, the maximum stroke where the motor moves but the picture sharpness does not change, so as to avoid introducing mechanical idle travel interference; this step is not allowed to process pan-tilt rotation commands during the focusing process, so as to avoid centrifugal inertia disturbance.

[0130] This step replaces the blind image search with triangular ranging, which greatly reduces the focusing time; the forced unidirectional movement completely avoids the nonlinear return difference caused by the gear reverse clearance, and at the same time eliminates the possibility of wandering between multi-peak pseudo-extreme values, which greatly improves the one-time success rate of focusing.

[0131] It should be noted that the current vertical pitch angle feedback value of the gimbal comes from the mechanical angle of the rotating bracket base read by its internal potentiometer or magnetic encoder. However, when the camera module containing the lens and CMOS sensor is fixed to the rotating bracket by screws or glue, the principal optical axis of its optical lens cannot be perfectly parallel to the mechanical zero-position normal of the bracket, usually with a tolerance of [0.5, 1.5] degrees. Secondly, in the negative temperature environment of a freezer, the expansion coefficients of metal and plastic are different, causing the front lens barrel of the camera to have a slight static droop or warping relative to the rotation center of the gimbal. This deformation exists at the factory and is a fixed offset, which cannot be detected by the rotary encoder of the gimbal. If θ0=0, the system will be at the mechanical zero position β. fb When θ = 0, the optical axis is considered to be absolutely horizontal, and the depth Z is calculated to be infinite according to the formula, causing the focusing motor to be pushed to the far end of the limit, and the image becomes completely blurred; therefore, it is necessary to calibrate at the factory to obtain θ0.

[0132] Preferably, in one embodiment of the present invention, θ0 is obtained by calibration using the two-point equal-height alignment method:

[0133] Inside the freezer, on the shelf surface, two identical high-contrast targets, P1 and P2, of known size, are placed one in front of the other along the horizontal direction of the optical axis; both targets have the same height, H. target And their horizontal distance is known to be L; for example, take L = 500 mm;

[0134] The gimbal actuator is forcibly driven to the zero point of its internal mechanical potentiometer via remote command, thus causing the feedback reading β to... fb =0.

[0135] Record the difference in the vertical pixel coordinates of the center points of the two targets in the camera view at this moment;

[0136] Because of the unknown θ0, when the mechanical reading is 0, the real optical axis is not horizontal, but rather obliquely points at the shelf plane at a small angle θ0.

[0137] Let y1 be the imaging height of the optical axis hitting the high-contrast target P1 at the actual tilt angle, in pixels, and y2 be the imaging height of the optical axis hitting the high-contrast target P2.

[0138] Using pinhole projection from a camera, the pixel height difference Δy = y2 - y1 is converted into the actual physical height difference Δh, with the unit being millimeters; the conversion factor k p This can be directly provided by the camera's factory specifications, namely the ratio of focal length to pixel size;

[0139] Calculate the actual physical height difference Δh=k p ∙Δy;

[0140] Since the two high-contrast targets are at the same height, the actual physical height difference Δh is entirely caused by the tilt of the optical axis; according to the principle of similar triangles, the falling / rise slope of the optical axis over the horizontal distance L is:

[0141] tan(θ0)=Δh / L;

[0142] Accordingly, the absolute value of the correction term for the fixed installation angle between the optical axis of the camera and the mechanical zero position of the pan-tilt rotation can be obtained;

[0143] If the imaging height of the high-contrast target P2 is higher than that of P1, that is, y2>y1, it indicates that the optical axis is actually tilted upward, at this time Δh>0, and θ0 takes a positive value.

[0144] If y2<y1, it indicates that the optical axis is actually tilted downward, and θ0 takes a negative value.

[0145] The calculated signed θ0 is burned into the non-volatile register as the correction term for the fixed installation angle between the optical axis of the ex-factory camera and the mechanical zero position of the pan-tilt rotation.

[0146] Further, the focus motor and the pan-tilt motor perform unidirectional propulsion for a long time, which avoids real-time oscillation. However, once an abnormal working condition such as power failure or remote emergency stop is encountered, the meshing surface of the transmission gear is in a random state when the system is reset, and long-term unidirectional force will accumulate tiny elastic deformation, resulting in slow drift of the mechanical zero position. This step maintains long-term coordinate accuracy through active mechanical backlash elimination actions and periodic hard reference calibration.

[0147] Each time the focus lock command is completed or at each power-on initialization, the drive motor is controlled to perform an over-step action along a preset fixed direction, and then retreat by an equal number of steps in the reverse direction, so that the meshing surfaces in the transmission mechanism are kept in unidirectional contact; at the same time, the pan-tilt actuator is periodically controlled to run to the physical limit switch to complete the hard zeroing operation, so as to refresh the absolute coordinate reference zero point of the pre-compensation data set, specifically as follows:

[0148] After each focus lock command is completed and before the pan-tilt enters the standby power-saving state, the drive motor is controlled to output an additional pulse train exceeding the target angle along the preset direction, and the number of pulses is twice the number of pulses corresponding to the maximum idle travel backlash of the transmission gear set measured in advance; then the motor is immediately controlled to retreat by an equal number of pulses in the opposite direction;

[0149] This back-and-forth movement in one forward and one reverse direction keeps the gear set always against the meshing surface from the same side, eliminating the gaps on both sides.

[0150] During each system power-on initialization, the control gimbal actuator first moves at a low speed toward the positive limit switch in the horizontal direction. After triggering a level change, the position is marked as 0. Then it moves toward the negative limit switch in the vertical direction. After triggering a level change, the position is marked as 0.

[0151] The calibrated position is recorded as the new physical zero point, and the origin of the stored first compensation dataset is realigned with the new physical zero point.

[0152] In one embodiment of the present invention, the maximum backlash of the transmission gear set is defined as the number of idle pulses of the motor output shaft, expressed as G. back The unit is pulse, and this value is the factory measurement constant.

[0153] At the end of each focus lock, the current focus motor position is S. final At the same time, the drive motor is controlled to output an additional pulse train exceeding the target angle in a direction away from the camera lens;

[0154] First, output the positive excess pulse number P. extra =2×G back The motor is temporarily moved to the virtual position S. temp =S final +P extra ;

[0155] Next, the number of reverse pulses P is output. return =P extra The motor eventually stopped at physical position S. final ′=S temp -P return ;

[0156] Due to the nonlinearity of mechanical transmission clearance, although S final Numerically, ′ is equal to S. final However, in terms of physical meshing, the driving surfaces of all gears are forced to the same side, usually the positive direction side, eliminating the idle travel G during forward and reverse switching. back .

[0157] For gimbal angle maintenance, define the angle reset mapping relationship after hard zeroing; let C be the original encoder reading corresponding to the limit switch trigger level. α,limit and C β,limit Then reset the current horizontal azimuth to: α cur =0; Current vertical pitch angle reset to β cur =0;

[0158] Subsequently, the horizontal azimuth sequence A and vertical pitch sequence B of the first compensation dataset D are corrected by origin shift and updated as follows:

[0159] A={α1-α0,α2-α0,…,α i -α0,…,α M -α0};

[0160] B={β1-β0,β2-β0,…,β j -β0,…,β N -β0}, where α0 and β0 are the mechanical deviation constants measured in this hard zeroing;

[0161] It should be noted that the additional pulse train that controls the drive motor to output beyond the target angle in the direction away from the camera lens must be executed when the motor is completely stationary and has no acceleration or deceleration inertia; otherwise, the additional pulses will cause overshoot.

[0162] The operating speed for gimbal angle maintenance operations must be 30% lower than the normal operating speed to avoid mechanical impact damaging the limit switches. If the system experiences an unexpected power outage during operation, the maintenance should be abandoned, and a hard reset should be performed first upon the next power-on.

[0163] This step addresses the inherent position memory loss issue in open-loop stepper systems. Active backlash elimination ensures a high degree of consistency in gear engagement at each stop, translating to a pulse count P. α P β The calculation provides an accurate starting baseline; the periodic hard zeroing uses physical limits as absolute anchor points, completely eliminating the cumulative error caused by temperature changes or slight slippage after long-term operation, ensuring the long-term effectiveness of the pre-compensation matrix and extending the maintenance-free cycle of the equipment.

[0164] A remote monitoring-based freezer camera control system includes:

[0165] The calibration and storage module, whose input is connected to the ranging unit of the camera device, is used to traverse the rotatable angle range of the gimbal actuator under the set working conditions of the freezer, measure and record the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer at each discrete angle node, and associate each angle node with the corresponding optical path offset to store as the first compensation dataset.

[0166] An angle pre-compensation drive module has its input terminals connected to the output terminal of the calibration and storage module and the remote command receiving terminal, respectively. It is used to respond to the target angle command input remotely, query the first compensation dataset to obtain the corresponding optical path offset, calculate the compensation deflection amount accordingly, and use the combined result of the target angle and the compensation deflection amount as the execution drive angle to drive the gimbal actuator to perform a one-time position rotation. Its control output terminal is connected to the gimbal drive device.

[0167] The depth linkage focusing module has its input end connected to the angle feedback output end of the gimbal drive device. After the gimbal drive device rotates to the position, it calculates the target depth value based on the current pitch angle feedback value, and generates the first driving displacement of the focusing lens group according to the preset object distance-focal length mapping relationship. It drives the gimbal focusing device to move from the current position to the target focusing position in a single direction, and performs a one-way cut-off lock after reaching the target position. Its control output end is connected to the gimbal focusing device.

[0168] The backlash elimination and reference maintenance module has its input terminal connected to the system power-on signal terminal and the locking completion signal output terminal of the depth linkage focusing module. Its control output terminal is connected to the gimbal focusing device and the gimbal driving device respectively. It is used to control the gimbal focusing device to perform an over-step action along a preset fixed direction and then reverse and retreat by an equal amount of steps after each focus locking is completed, so that the meshing surface of the transmission mechanism is kept in contact in a single direction. It is also used to control the gimbal driving device to run to the physical limit switch to perform a hard zeroing operation when the system is powered on, so as to refresh the absolute coordinate reference zero point of the pre-compensation dataset.

[0169] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for adjusting a freezer camera based on remote monitoring, characterized in that: Specifically, the steps include the following: When the freezer is in a set working condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle node and the corresponding optical path offset are associated and stored to construct the first compensation dataset. In response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and the compensation deflection is calculated based on the optical path offset; the combined result of the target angle and the compensation deflection is used as the execution drive angle to control the gimbal actuator to perform a one-time position rotation. After the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined according to the current gimbal actuator. Based on the target depth value, the preset object distance-focal length mapping relationship is retrieved to generate the first driving displacement. Control the gimbal drive device to move to the target focusing position corresponding to the first drive displacement; Upon completion of each focus lock command or each power-on initialization, the gimbal drive device is controlled to perform an overstepping action in a preset direction, and then reverses to perform an equal step, so that the meshing surfaces in the transmission mechanism remain in contact in a single direction; at the same time, a hard zeroing operation is performed to refresh the absolute coordinate reference zero point of the pre-compensation dataset.

2. The method for adjusting a freezer camera based on remote monitoring according to claim 1, characterized in that: With the freezer in a set operating condition, the rotatable angle range of the gimbal actuator is traversed, and the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer is measured and recorded at each discrete angle node. The angle nodes are associated with the corresponding optical path offsets and stored to construct the first compensation dataset, specifically: Once the freezer reaches the set stable operating conditions and the internal target remains stationary, perform an offline calibration. The gimbal actuator is controlled to traverse the entire mechanical travel range of its horizontal azimuth and vertical pitch angles, and values ​​are taken according to a fixed sampling step size. At each discrete angle node, an airborne near-infrared ranging spot is projected onto the target. By receiving the reflected echo and calculating the transmission-reception time difference, the total round-trip time of the actual light from the front lens of the camera to the target at that angle is measured, and the difference is made with the theoretical travel time in vacuum to obtain the optical path offset at that discrete angle node. The horizontal azimuth angle, vertical pitch angle, and optical path offset corresponding to the discrete angle node are used as a triplet and written into a two-dimensional data table to form the first compensation dataset.

3. The method for adjusting a freezer camera based on remote monitoring according to claim 2, characterized in that: In response to a remotely input gimbal target angle command, the first compensation dataset is queried to obtain the optical path offset corresponding to the target angle, and a compensation deflection is calculated based on the optical path offset. The combined result of the target angle and the compensation deflection is used as the execution drive angle to control the gimbal actuator to perform a single-position rotation, specifically: Based on the remotely input gimbal target angle, the optical path offset corresponding to the gimbal target angle is calculated. Based on the difference approximation, the amount of compensation deflection required to offset the refraction deviation at the current viewing angle is calculated. Calculate the angle of synthetic execution; The pulse width modulation driver is controlled to drive the gimbal actuator to rotate continuously into position in one go, with the synthesized drive angle as the target.

4. The method for adjusting a freezer camera based on remote monitoring according to claim 3, characterized in that: After the gimbal actuator rotates to its position, the target depth value corresponding to the center of the field of view is determined based on the current gimbal actuator. A preset object distance-focal length mapping relationship is retrieved based on the target depth value to generate a first driving displacement. The gimbal drive device is then controlled to move to the target focusing position corresponding to the first driving displacement, specifically as follows: After the gimbal actuator rotates to its position and comes to a complete stop, read the current vertical pitch angle feedback value; Calculate the target depth value between the current optical axis and the shelf plane by combining the vertical installation height from the rotation center of the gimbal to the shelf plane; The output distance-focusing motor step count correspondence table is pre-calibrated at the factory; The first driving displacement is obtained by looking up the calculated target depth value in a table. The gimbal focusing device is moved from its current position along a preset direction to the first driving displacement. Upon arrival, take one more microstep forward from that position and compare the high-frequency energy values ​​of the image before and after taking that microstep. If the energy value increases, continue to advance one microstep forward. If the energy value decreases or remains unchanged, immediately execute the focus lock command, prohibit the motor from reversing, and record the current position as the focus position.

5. The method for adjusting a freezer camera based on remote monitoring according to claim 4, characterized in that: Upon completion of each focus lock command or power-on initialization, the gimbal drive device is controlled to perform an over-stepping action in a preset direction, followed by a reverse retraction of an equal step, ensuring that the meshing surfaces within the transmission mechanism remain in contact in a single direction; simultaneously, a hard zeroing operation is performed to refresh the absolute coordinate reference zero point of the pre-compensation dataset, specifically: After each focus lock command is completed and before the gimbal enters standby power-saving mode, the gimbal drive device is controlled to output an additional pulse train exceeding the target angle along the preset direction. The number of pulses is twice the number of pulses corresponding to the maximum backlash of the transmission gear set as measured in the pre-determined sequence. Then, the motor is immediately controlled to retract an equal number of pulses in the opposite direction. During each system power-on initialization, the control gimbal actuator first moves at a low speed toward the positive limit switch in the horizontal direction. After triggering a level change, the position is marked as 0. Then it moves toward the negative limit switch in the vertical direction. After triggering a level change, the position is marked as 0. The calibrated position is recorded as the new physical zero point, and the origin of the stored first compensation dataset is realigned with the new physical zero point.

6. The method for adjusting a freezer camera based on remote monitoring according to claim 2, characterized in that: Let the mechanically adjustable range of the horizontal azimuth angle be [α]. min ,α max The mechanically adjustable range of the vertical pitch angle is [β]. min ,β max ]; where α min and α max These represent the upper and lower adjustable limits of the horizontal azimuth angle of the gimbal actuator, respectively; β min and β max These represent the upper and lower adjustable limits of the vertical pitch angle of the gimbal actuator, respectively. Discretize the above interval with fixed step sizes Δα and Δβ, and denote the number of angle nodes sampled as M and N, respectively; Discrete angle sequences were collected separately, and the horizontal azimuth sequence was denoted as A={α1,α2,…,α...}. i ,…,α M }, where α i =α min +(i-1)∙Δα, where i is the horizontal angle node index number, i=1,2,…,M; α i This represents the i-th horizontal angle scalar value in the sequence; Δα represents the minimum sampling interval of the horizontal azimuth angle. Vertical pitch angle sequence B={β1,β2,…,β j ,…,β N }, where β j =β min +(j-1)∙Δβ, where j is the index number of the vertical angle node, j=1,2,…,N; β j This represents the j-th vertical angle scalar value in the sequence; Δβ represents the minimum sampling interval for the vertical pitch angle. At angle node (α) i ,β j The optical path offset measured at point ) is denoted as Δ. i,j The first compensation dataset is a two-dimensional matrix D with dimension M×N, represented as: ; where Δ M,N Represents the angle node (α) M ,β N The optical path offset measured at ().

7. The method for adjusting a freezer camera based on remote monitoring according to claim 5, characterized in that: After the gimbal actuator has rotated to its position and come to a complete stop, the vertical pitch angle feedback value β is read. fb ; β is defined when the horizontal viewing angle is specified fb =0, downward tilt is considered a negative angle; Let H be the vertical installation height from the center of rotation of the gimbal to the plane of the shelf. mount ; Let the target depth value between the current optical axis and the shelf plane be Z; The target depth value Z between the current optical axis and the shelf plane is uniquely determined as: Z = H mount / tan(|β fb |+θ0); where θ0 represents the fixed installation angle correction term between the camera optical axis and the zero position of the gimbal rotation mechanism; Let the factory-calibrated object distance-focusing motor step count correspondence table be {S1,S2,…,S...} K The corresponding object distance sequence is {Z1, Z2, ..., Z}. K };Z K S represents the depth value of the Kth target; K Z represents K The corresponding driving displacement; For the calculated continuous target depth value Z, when Z can correspond to any target depth value calibrated by the factory, the corresponding driving displacement is obtained directly by looking up the table as the first driving displacement S. When Z q <Z <Z q+1 Then, linear interpolation yields the first driving displacement. Record the current actual position of the gimbal focusing device as S. cur ; System comparison S with S cur When S>S cur Then record the number of forward target steps S. target =S; if S≤S cur Set the number of forward target steps to S. target =S cur +S0, where S0 is a preset positive safety offset constant, and then the gimbal drive moves forward to S in one go. target .

8. The method for adjusting a freezer camera based on remote monitoring according to claim 7, characterized in that: Also includes: Arrival in S target Then, the process enters the one-way cutoff verification phase, where a microstep perturbation increment g is defined. The pan-tilt focus device moves forward g steps, and calculates the image high-frequency energy values E1 and E2 before and after the stepping; if E2≥E1, set S target =S target +g, and repeat the iteration; if E2<E1, terminate immediately and lock, and the final locked position scalar S final is taken as the current S target value.

9. A method for adjusting a freezer camera based on remote monitoring according to claim 7, characterized in that: The fixed installation angle correction term between the camera optical axis and the zero position of the gimbal rotation mechanism is obtained by calibration using the two-point equal height alignment method.

10. A freezer camera adjustment system based on remote monitoring, employing the freezer camera adjustment method based on remote monitoring as described in any one of claims 1-9, characterized in that: The calibration and storage module is used to traverse the rotatable angle range of the gimbal actuator under the set working conditions of the freezer, measure and record the optical path offset generated when the imaging optical path of the camera passes through the medium attached to the surface of the transparent protective layer of the freezer at each discrete angle node, and associate each angle node with the corresponding optical path offset to store as a first compensation dataset. Angle pre-compensation drive module is used to respond to the target angle command input remotely, query the first compensation dataset to obtain the corresponding optical path offset, calculate the compensation deflection, and use the combined result of the target angle and the compensation deflection as the execution drive angle, so that the gimbal drive device can perform a one-time position rotation. The depth linkage focusing module is used to calculate the target depth value based on the current pitch angle feedback value after the gimbal drive device rotates to the position, and generate the first driving displacement of the focusing lens group according to the preset object distance-focal length mapping relationship, drive the gimbal focusing device to move from the current position to the target focusing position, and perform one-way cut-off locking after arrival. The backlash elimination and reference maintenance module is used to control the gimbal focusing device to perform an over-step action in a preset direction after each focus lock is completed, and then reverse to retreat by an equal amount of steps, so that the meshing surface of the transmission mechanism remains in contact in a single direction. It is also used to control the gimbal drive device to perform a hard zeroing operation when the system is powered on.

Images