AGV docking positioning method based on visual positioning technology

Abstract

The present invention relates to the field of AGV positioning technology, and discloses an AGV trolley docking positioning method based on visual positioning technology. The method includes capturing an image of the target docking area, obtaining a visual image containing docking marks and feature point distribution information, and identifying and obtaining initial positioning parameters; comparing the initial positioning parameters with a standard range, and generating a positioning deviation signal if the range exceeds; tracking feature point displacement data based on the deviation signal, obtaining a displacement change sequence and feature point offset within a tracking period, and comparing them with a threshold value. When the feature point offset is greater than or equal to the threshold value, a high-level offset signal is generated; based on the high-level offset signal, a coordinate conversion coefficient and an angle compensation value are obtained, and a comprehensive positioning correction value is obtained by superposition calculation; at the same time, a dynamic adjustment coefficient is obtained, and the current driving speed parameter and the dynamic adjustment coefficient are integrated to calculate the adjusted driving parameter, thereby completing the driving state adjustment. The dynamic adjustment coefficient is obtained by taking the weighted sum of all feature point offsets and taking the average value.

Description

Technical Field

[0001] The present invention relates to the field of AGV positioning technology, in particular to an AGV trolley docking and positioning method based on visual positioning technology. Background Art

[0002] In the field of industrial automation, AGVs serve as core equipment for material transfer and equipment docking. The accuracy of their docking and positioning is directly related to the smoothness of the production process. Traditional AGV positioning methods, such as laser navigation, which relies on preset reflectors, are prone to signal obstruction in environments with multiple obstacles, resulting in positioning jumps. Magnetic navigation, however, is limited by the layout of magnetic strips on the floor, making it difficult to adapt to the rapid adjustments required for production lines. As the cost of vision sensors decreases, visual positioning technology has gradually become the mainstream choice for AGV positioning, but practical applications still face numerous challenges.

[0003] Existing visual positioning methods, when handling docking scenarios, often determine the initial position through single feature point recognition, lacking real-time verification of positioning parameters. When the AGV approaches the docking area, if the initial positioning parameters are outside a reasonable range, the system struggles to detect deviations in a timely manner, often leading to positional offsets during the initial docking phase. While some technologies incorporate deviation detection, these tracking methods often rely on fixed-cycle sampling to track feature point displacements. This is unable to capture continuous displacement changes in dynamic environments, resulting in lag in offset calculations.

[0004] Traditional methods for offset processing often rely on a single threshold, ignoring the varying impacts of different feature point offsets on overall positioning. When large offsets occur, simple coordinate corrections alone are insufficient to compensate for the cumulative error caused by angular deviations, leading to frequent lateral offsets during docking. Furthermore, driving parameter adjustments often rely on fixed proportional coefficients, failing to consider the combined impact of feature point offsets. This often results in over- or under-adjustments. Over-adjustments can cause the AGV to oscillate during sudden stops, while under-adjustments prevent the deviations from being corrected, ultimately leading to docking failures.

[0005] In complex industrial environments, factors such as lighting variations and dust cover can lead to unstable feature point recognition. Existing technologies lack the ability to dynamically track feature point offsets, making it difficult to distinguish between temporary interference and persistent offsets. Misjudgments often trigger unnecessary adjustments, increasing energy consumption and prolonging docking times. These combined issues have resulted in low efficiency and poor stability for AGVs in high-precision docking scenarios, limiting their application in demanding applications such as electronics manufacturing and warehousing and logistics, where docking accuracy is critical. Summary of the Invention

[0006] The purpose of the present invention is to provide an AGV trolley docking and positioning method based on visual positioning technology to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides an AGV docking and positioning method based on visual positioning technology, the method comprising:

[0008] Step 1: Capture an image of the target docking area to obtain a visual image containing docking marks, wherein the visual image includes feature point distribution information, and obtain initial positioning parameters by identifying the feature point distribution information;

[0009] Step 2: Based on the initial positioning parameters, the initial positioning parameters are compared with the standard positioning parameter range. If the initial positioning parameters exceed the standard positioning parameter range, a positioning deviation signal is generated;

[0010] Step 3: Based on the positioning deviation signal, track the displacement data of the feature points in the visual image. During the tracking period, obtain the displacement change sequence, obtain the feature point offset, and compare it with the offset threshold to generate the offset impact level signal;

[0011] If the feature point offset is greater than or equal to the offset threshold, a high-level offset signal is generated;

[0012] Step 4: Based on the high-level offset signal, obtain the coordinate conversion coefficient and angle compensation value, and superimpose the coordinate conversion coefficient and the angle compensation value to obtain the comprehensive positioning correction value;

[0013] Step 5: Based on the high-level offset signal, the dynamic adjustment coefficient is obtained, and the current driving speed parameter is integrated with the dynamic adjustment coefficient to obtain the required adjustment driving parameter, thus completing the adjustment of the AGV driving state;

[0014] The dynamic adjustment coefficient is obtained as follows:

[0015] The weighted sum of all feature point offsets is taken to get the average value to obtain the dynamic adjustment coefficient.

[0016] Preferably, the initial positioning parameters are obtained as follows:

[0017] The docking area in the visual image is divided into several positioning sub-areas, the number of feature points in each positioning sub-area is obtained, and the number of feature points in each positioning sub-area is weighted and averaged to obtain the feature point density value;

[0018] The ratio of the feature point density value to the standard density value is calculated to obtain the initial positioning parameters.

[0019] Preferably, the feature point offset is obtained in the following manner:

[0020] The horizontal offset component and the vertical offset component are obtained, and vector synthesis of the horizontal offset component and the vertical offset component is performed to obtain the feature point offset.

[0021] Preferably, the lateral offset component is obtained in the following manner:

[0022] Calculate the ratio of the image frame interval value and the feature point pixel displacement value within the same tracking period to obtain the displacement synchronization coefficient, select the period where the displacement synchronization coefficient meets the preset conditions, and mark it as the valid tracking period;

[0023] The duration of the effective tracking period is obtained, and the ratio of the duration of the effective tracking period to the total duration of the tracking cycle is calculated to obtain the lateral offset component.

[0024] Preferably, the image frame interval value and the feature point pixel displacement value are obtained in the following manner:

[0025] During the tracking period, the tracking period is divided into several time segments, and a time-displacement coordinate system is established. The horizontal dimension represents the time segment, and the vertical dimension represents the pixel position of the feature point corresponding to each time segment. The obtained pixel position of the feature point is substituted into the time-displacement coordinate system to draw the pixel displacement change curve;

[0026] Based on the pixel displacement change curve, the pixel displacement peak point and the pixel displacement valley point are obtained, and the coordinates of the adjacent pixel displacement peak point and the pixel displacement valley point are calculated to obtain the pixel displacement value of the feature point;

[0027] During the tracking period, the tracking period is divided into several time segments, and a time-frame number coordinate system is established. The horizontal dimension represents the time segment, and the vertical dimension represents the image acquisition frame number corresponding to each time segment. The obtained image acquisition frame number is substituted into the time-frame number coordinate system to draw a frame number change curve.

[0028] Based on the frame number change curve, the frame number increasing points and the frame number decreasing points are obtained, and the coordinates of adjacent frame number increasing points and frame number decreasing points are calculated to obtain the image frame interval value.

[0029] Preferably, the longitudinal offset component is obtained in the following manner:

[0030] Compare the sub-region feature point deviation value with the sub-region deviation allowable value. The comparison process is as follows:

[0031] If the deviation value of the feature point in the sub-region is greater than or equal to the sub-region deviation allowable value, the sub-region is marked as a deviation abnormal sub-region;

[0032] If the deviation value of the feature point in the sub-region is less than the sub-region deviation allowable value, the sub-region is marked as a normal deviation sub-region;

[0033] Compare the sub-region image clarity value with the sub-region clarity threshold. The comparison process is as follows:

[0034] If the sub-region image clarity value is less than the sub-region clarity threshold, the sub-region is marked as a clarity abnormal sub-region;

[0035] If the sub-region image clarity value is greater than or equal to the sub-region clarity threshold, the sub-region is marked as a sub-region with normal clarity;

[0036] Obtaining an offset abnormal subregion, performing spatial overlap analysis on the offset abnormal subregion and the clarity abnormal subregion to obtain overlapping abnormal subregions, obtaining the number of overlapping abnormal subregions, and accumulating and summing the number of overlapping abnormal subregions to obtain the total number of abnormal regions;

[0037] The longitudinal offset component is obtained by calculating the ratio of the total number of abnormal areas to the total number of positioning sub-areas.

[0038] Preferably, the sub-region feature point deviation value is obtained in the following manner:

[0039] The visual image is divided into regions to obtain positioning sub-region images. Based on the positioning sub-region images, the actual coordinates of the feature points in each positioning sub-region are extracted and marked as the sub-region measured coordinate values. The sub-region measured coordinate values ​​are calculated with the sub-region standard coordinate values ​​to obtain the sub-region feature point deviation value;

[0040] The method for obtaining the sub-region image clarity value is:

[0041] Perform clarity detection on the visual image, divide the visual image into several positioning sub-areas, obtain the image grayscale gradient value in each positioning sub-area, mark it as the sub-area measured gradient value, calculate the difference between the sub-area measured gradient value and the sub-area standard gradient value, and obtain the sub-area image clarity value.

[0042] Preferably, the angle compensation value is obtained as follows:

[0043] The pixel displacement values ​​of the feature points and the image frame interval values ​​within the effective tracking period are obtained, and the ratio of the pixel displacement values ​​of the feature points and the image frame interval values ​​is calculated to obtain the displacement rate value. The displacement rate values ​​within each effective tracking period are accumulated and averaged to obtain the angle compensation value.

[0044] Preferably, the coordinate conversion coefficient is obtained as follows:

[0045] Based on the displacement synchronization coefficient corresponding to the above-mentioned effective tracking period, it is marked as the effective synchronization coefficient. The difference between all effective synchronization coefficients in the tracking period and the synchronization coefficient standard value is calculated to obtain the synchronization coefficient deviation value. The synchronization coefficient deviation value is calculated for variance to obtain the coordinate conversion coefficient.

[0046] Preferably, the verification method of the comprehensive positioning correction value is: obtain the angle compensation value and the coordinate conversion coefficient within the effective tracking period, calculate the ratio of the angle compensation value and the coordinate conversion coefficient, and obtain the correction verification coefficient. If the correction verification coefficient is within the preset verification range, the comprehensive positioning correction value is confirmed to be valid.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This AGV docking and positioning method, based on visual positioning technology, effectively overcomes the limitations of traditional positioning methods through multi-step collaborative design. During the initial positioning phase, initial positioning parameters are obtained by identifying feature point distribution information and comparing them with standard ranges in real time. This method can quickly detect positioning deviations in the early stages of docking, preventing them from accumulating and expanding in subsequent processes. This early intervention approach significantly reduces docking errors caused by initial positioning errors, compared to the delayed deviation detection in traditional methods.

[0049] During deviation processing, feature point displacement data is tracked and a displacement change sequence is generated. This eliminates the need to calculate feature point offsets based on sampled data from a single time point, instead basing the calculation on trends over continuous periods. This allows offset assessment to better reflect actual dynamic processes. Furthermore, an offset impact level signal is generated based on the comparison of the offset with a threshold. This signal can distinguish the impact of varying degrees of offset on docking, avoiding the crude treatment of all offsets in traditional methods. This allows system resources to be focused on processing high-level offsets, improving the specificity of the deviation response.

[0050] For high-level offset signals, the coordinate conversion coefficient and the angle compensation value are superimposed to calculate a comprehensive positioning correction value. This takes into account both the deviation of the plane coordinates and the influence of angular deviation. Compared with the traditional method of only performing single-dimensional correction, the correction dimension is more comprehensive and can effectively compensate for the positioning error caused by the superposition of multiple angle offsets. At the same time, the dynamic adjustment coefficient is obtained by taking the weighted sum of the offsets of all feature points and taking the average value. This allows the adjustment of driving parameters to integrate the offsets of all feature points, avoiding the adjustment imbalance caused by the offset of a single feature point, and making the driving state adjustment of the AGV more in line with the actual docking needs.

[0051] When dealing with feature point offsets, this method focuses on both static positioning parameter correction and dynamic driving state adjustment. The synergistic effect of the two enables the AGV to quickly correct position deviations when faced with feature point offsets in complex environments, and to maintain stable operation by adjusting driving parameters, thereby reducing oscillations or stagnation caused by drastic adjustments. While improving docking smoothness, it also reduces energy consumption and loss of equipment, making it more adaptable to scenarios such as electronic assembly and precision warehousing that have high requirements for docking accuracy and stability.

[0052] This design concept breaks away from the reliance on a single technical means in traditional positioning methods. Through dynamic collaboration of multiple steps, it makes the application of visual positioning technology in AGV docking scenarios more adaptable. Whether facing fluctuations in feature point recognition caused by changes in lighting or local offsets in multi-obstacle environments, precise docking can be achieved through a systematic processing flow, providing a new technical path for the efficient application of AGV carts in various industrial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flowchart of the steps of the AGV trolley docking and positioning method based on visual positioning technology described in the present invention;

[0054] Figure 2 Flowchart of the method for obtaining initial positioning parameters;

[0055] Figure 3 is a flow chart of a method for obtaining a lateral offset component;

[0056] Figure 4 Flowchart of the method for obtaining coordinate conversion coefficients. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] See also Figure 1 The present invention provides an AGV trolley docking and positioning method based on visual positioning technology, and the specific implementation steps of the method are as follows:

[0059] Step 1: Capture an image of the target docking area. The AGV's onboard visual sensor (such as an industrial camera) captures a visual image containing docking marks. The image contains information about the distribution of feature points. An image recognition algorithm is used to extract the location, number, and distribution of these feature points. Initial positioning parameters are calculated based on this information.

[0060] Step 2: Compare the initial positioning parameters obtained in Step 1 with the preset standard positioning parameter range. If the initial positioning parameters are within the standard range, it indicates that the AGV's current positioning status meets the docking requirements and no adjustment is required. If the initial positioning parameters are outside the standard range, a positioning deviation signal is triggered, and the subsequent adjustment process begins.

[0061] Step 3: After the positioning deviation signal is triggered, the feature point displacement tracking mechanism is activated. Within the set tracking period, the displacement changes of the feature points are recorded by continuously capturing image frames, generating a displacement change sequence. Based on this sequence, the feature point offset is calculated and compared with a preset offset threshold. If the offset is less than the threshold, a low-level offset signal is generated, maintaining the current driving state. If the offset is greater than or equal to the threshold, a high-level offset signal is generated, and subsequent correction steps are executed.

[0062] Step 4: For high-level offset signals, the coordinate conversion coefficient is calculated using the coordinate conversion model. This is then combined with the angle sensor data or image analysis results to obtain the angle compensation value. The two are then superimposed to obtain a comprehensive positioning correction value, which is used to correct the AGV's spatial position deviation.

[0063] Step 5: Based on the high-level offset signal, the weighted sum of all feature point offsets is averaged to obtain the dynamic adjustment coefficient. The AGV's current driving speed parameters (including linear velocity and angular velocity) are combined with the dynamic adjustment coefficient to output the adjusted driving parameters, enabling real-time control of the vehicle's speed and direction, completing state adjustments during the docking and positioning process.

[0064] Example 1: See Figure 2 During the acquisition of initial positioning parameters, the docking area in the visual image must be properly divided. The docking area is then divided into several positioning sub-areas according to a preset grid specification. The grid specification can be set based on the actual requirements of the docking scenario, such as the size of the docking area and the required positioning accuracy. Once the division is complete, image recognition technology is used to extract the feature points contained in each positioning sub-area, and the specific number of feature points in each sub-area is counted.

[0065] A weighted sum is performed on the number of feature points within each sub-region. The weights are determined based on the importance of each sub-region in the overall docking process. Different sub-regions receive different weights due to their roles in docking. After the weighted sum is completed, the resulting sum is divided by the total number of sub-regions and the average is taken to obtain the feature point density value.

[0066] The obtained feature point density value is then compared to a pre-set standard density value. This standard density value is a benchmark for feature point density determined under ideal docking conditions through multiple experiments and data accumulation. By calculating the ratio of the two, the initial positioning parameters are ultimately obtained, which can, to a certain extent, reflect the AGV's initial positioning relative to the target docking area.

[0067] To obtain the feature point offset, we first need to obtain the horizontal offset component and the vertical offset component. The horizontal offset component mainly reflects the offset of the feature point in the horizontal direction, while the vertical offset component mainly reflects the offset of the feature point in the vertical direction.

[0068] Determining the lateral offset component requires combining relevant information from the image frames with the pixel displacement of the feature points. Within the same tracking period, the image frame interval and the pixel displacement of the feature points are analyzed. A specific calculation method is used to derive a parameter reflecting the relationship between the two. This allows valid tracking periods to be selected, and the lateral offset component is calculated based on the data within these valid tracking periods.

[0069] Obtaining the longitudinal offset component involves analyzing the deviation of feature points and image clarity in the positioning subregions. First, the feature point deviation value of each positioning subregion is calculated and compared with the corresponding allowable value, marking subregions with abnormal offset. Simultaneously, the image clarity value of each positioning subregion is calculated and compared with the corresponding threshold, marking subregions with abnormal clarity. Then, a spatial overlap analysis is performed on the abnormal offset and abnormal clarity subregions to determine the overlapping abnormal subregions. The number of these overlapping abnormal subregions is counted and the ratio is calculated with the total number of positioning subregions to obtain the longitudinal offset component.

[0070] After obtaining the horizontal and vertical offset components, they are combined according to the principle of vector synthesis. This synthesis method comprehensively considers the offset in both the horizontal and vertical directions, ultimately obtaining the feature point offset, which fully reflects the degree and direction of the feature point's deviation from the standard position.

[0071] Example 2: See Figure 3 To obtain the lateral offset component, the displacement synchronization coefficient must first be determined, followed by screening for valid tracking periods. Finally, the coefficient is calculated by comparing the duration of the valid tracking period to the total duration of the tracking cycle. The displacement synchronization coefficient is obtained by calculating the ratio of the image frame interval value and the pixel displacement value of the feature point within the same tracking period. This process requires precise extraction of these two values ​​within each tracking period to ensure calculation accuracy. After calculating the displacement synchronization coefficient, a preset condition is set based on the actual needs of AGV docking and positioning and past experience to screen out periods that meet the requirements. These screened periods are then marked as valid tracking periods. After determining the valid tracking periods, the duration of each valid tracking period is calculated, and the duration of all valid tracking periods is added together to obtain the total duration of the valid tracking periods. The total duration of the valid tracking periods is then calculated by comparing it to the total duration of the tracking cycle. The resulting lateral offset component is obtained.

[0072] The image frame interval value is obtained as follows: the entire tracking cycle is divided into several time segments at regular intervals. The intervals can be determined based on the image acquisition frequency and the required docking positioning accuracy. A time-frame number coordinate system is established, with the horizontal dimension representing the divided time segments and the vertical dimension representing the image acquisition frame number corresponding to each time segment. The image acquisition frame number within each time segment is obtained using the image acquisition device. These frame number data are substituted into the established time-frame number coordinate system, and a frame number variation curve is plotted based on this data. After obtaining the frame number variation curve, the curve is analyzed to identify points where the frame number increases and decreases. Increasing points are points in the curve where the frame number shows an upward trend, while decreasing points are points where the frame number shows a downward trend. Adjacent increasing and decreasing points are selected, and the coordinate difference between these two points in the coordinate system is calculated. This difference is the image frame interval value.

[0073] Obtaining pixel displacement values ​​for feature points is similar to obtaining image frame interval values, but with some specific differences. First, the tracking period is divided into several time segments, using the same division method as used to obtain image frame interval values ​​to ensure data consistency and comparability. Next, a time-displacement coordinate system is established, with the horizontal dimension representing the time segments and the vertical dimension representing the pixel positions of the feature points corresponding to each time segment. Image recognition technology is used to obtain the pixel positions of the feature points within each time segment. This position data is then substituted into the time-displacement coordinate system to plot a pixel displacement curve. Once plotted, the pixel displacement curve is analyzed to identify peak and valley points. A peak point is the point in the curve where the pixel displacement of the feature point reaches its maximum value, while a valley point is the point where the displacement reaches its minimum value. Next, adjacent peak and valley points are selected and their coordinate differences in the coordinate system are calculated. This difference represents the pixel displacement value of the feature point.

[0074] Throughout the implementation process, it is necessary to ensure the continuity and stability of image acquisition to ensure the accuracy of data such as the acquired image frame number and feature point pixel position. At the same time, in the process of dividing time segments, establishing coordinate systems, and analyzing curves to obtain relevant points, unified standards and methods must be adopted to avoid data deviations due to inconsistent operations. The drawn frame number change curve and pixel displacement change curve require careful observation and analysis to ensure accurate identification of increasing points, decreasing points, peak points, and valley points, because the accuracy of these points directly affects the calculation results of the image frame interval value and the feature point pixel displacement value, which in turn affects the accuracy of the lateral offset component, and ultimately affects the accuracy of the AGV docking and positioning.

[0075] Example 3: This example mainly focuses on the acquisition of the longitudinal offset component, which involves sub-region feature point deviation analysis, sub-region image clarity analysis, overlapping abnormal sub-region identification, and final longitudinal offset component calculation.

[0076] First, the deviation values ​​of the feature points in each subregion are compared with the permissible deviation value. In the visual image, the docking area is divided into several positioning subregions. For each positioning subregion, the actual coordinate information of the feature points within it must be extracted. This coordinate information is obtained from the visual image using image recognition technology and reflects the specific location of the feature points within the current subregion. Furthermore, each positioning subregion has preset standard coordinate values, which represent the ideal location of the feature points under ideal docking conditions. These standard coordinate values ​​are pre-set and stored through methods such as offline calibration. The actual coordinate values ​​of the feature points in each positioning subregion are calculated against the corresponding standard coordinate values, and the resulting difference is the feature point deviation value for that subregion. Next, the feature point deviation value for each subregion is compared with the permissible deviation value for the corresponding subregion. The permissible deviation value for the subregion is a threshold value set based on docking accuracy requirements and the actual application scenario. If the feature point deviation value for a subregion reaches or exceeds the permissible deviation value, the subregion is marked as having an abnormal deviation. If the feature point deviation value does not reach the permissible deviation value, the subregion is marked as having a normal deviation.

[0077] Then, the sub-region image clarity value is compared with the sub-region clarity threshold. For each positioning sub-region, its image clarity value needs to be obtained. The image clarity value is obtained by analyzing the grayscale gradient of the sub-region image. Specifically, the grayscale change of the image in the sub-region is calculated. The more obvious the grayscale change, the clearer the image and the higher the corresponding clarity value. Each positioning sub-region is also provided with a sub-region clarity threshold. This threshold is the standard for judging whether the image is clear and is set according to factors such as the performance of the image acquisition device and the lighting conditions of the docking environment. The image clarity value of each sub-region is compared with the corresponding sub-region clarity threshold. If the sub-region image clarity value is lower than the sub-region clarity threshold, it indicates that the sub-region image is blurred and it is marked as a sub-region with abnormal clarity. If the sub-region image clarity value reaches or exceeds the sub-region clarity threshold, the sub-region image is clear and marked as a sub-region with normal clarity.

[0078] After marking the two types of abnormal sub-regions, a spatial overlap analysis is performed on the offset and clarity abnormal sub-regions. Spatial overlap analysis compares the spatial location information of the two abnormal sub-regions to identify sub-regions marked with both offset and clarity abnormalities. These sub-regions are referred to as overlapping abnormal sub-regions. The number of overlapping abnormal sub-regions is counted and summed to obtain the total number of abnormal regions.

[0079] Finally, the longitudinal offset component is calculated using the following formula:

[0080] Longitudinal offset component = total number of abnormal areas ÷ total number of positioning sub-areas

[0081] In the formula, the total number of anomaly regions refers to the total number of overlapping anomaly subregions determined after spatial overlap analysis; the total number of positioning subregions refers to the total number of positioning subregions divided from the docking area. The result calculated using this formula is the longitudinal offset component, which reflects the overall offset of the feature points in the longitudinal direction.

[0082] Throughout the entire process, the analysis of each localized subregion must be consistent, ensuring uniform standards for marking anomalous subregions and avoiding errors caused by inconsistent standards. Furthermore, spatial overlap analysis requires precise matching of the subregions' spatial positions to ensure accurate identification of overlapping anomalous subregions, thereby ensuring that the calculation of the longitudinal offset component truly reflects the actual offset condition.

[0083] Example 4: It mainly involves the process of obtaining the sub-region feature point deviation value, sub-region image clarity value and angle compensation value. These steps are interrelated and together provide data support for the docking positioning of the AGV.

[0084] Obtaining sub-region feature point deviation values ​​begins with region segmentation of the visual image. The visual image containing the docking mark is segmented into multiple positioning sub-regions according to pre-set rules. The segmentation method can be flexibly configured based on the shape and size of the docking region, ensuring that each sub-region clearly reflects the local features of the docking region. After segmentation, the pixel coordinates of all feature points within each positioning sub-region are extracted using image recognition technology. These pixel coordinates represent the specific positions of the feature points in the image. Using parameters such as the camera's intrinsic parameter matrix, these pixel coordinates are converted to actual spatial coordinates. The actual spatial coordinates of all feature points within each sub-region are then averaged to obtain the measured coordinate value for that sub-region. Each positioning sub-region corresponds to a pre-set standard coordinate value. This standard coordinate value is determined through multiple measurements and calibrations under ideal docking conditions and represents the coordinates of the sub-region at its correct position. The measured coordinate value of each sub-region is then compared with the corresponding standard coordinate value to obtain the feature point deviation value for that sub-region. This value reflects the degree of deviation of the feature points within that sub-region from the standard position.

[0085] Obtaining the sub-region image clarity value requires grayscale processing of each located sub-region image, converting the color image into a grayscale image to more clearly analyze changes in image details. A specific algorithm is then used to process the grayscaled sub-region image, calculating its horizontal and vertical grayscale gradients. The grayscale gradient reflects the rate of change of pixel grayscale values ​​within the image. A greater rate of change indicates clearer edges and higher image clarity. The horizontal and vertical grayscale gradients are combined to obtain the measured gradient value for the sub-region. Each located sub-region also has a standard gradient value. This standard gradient value is the gradient value corresponding to a clear image under ideal lighting conditions and equipment status, determined through preliminary experiments and data accumulation. The difference between the measured gradient value and the corresponding standard gradient value for each sub-region is the image clarity value for that sub-region, which can be used to determine the image clarity of the sub-region.

[0086] The angle compensation value is obtained based on the valid tracking period. During this period, the pixel displacement value and image frame interval value of the feature point are obtained for each period. The pixel displacement value is obtained by analyzing the position change of the feature point in consecutive image frames and reflects the distance the feature point has moved at the pixel level. The image frame interval value is the time difference between two consecutive image frames and reflects the time interval between image acquisitions. The pixel displacement value of the feature point in each valid tracking period is calculated together with the image frame interval value to obtain the displacement rate value for that period. This value represents the pixel-level movement speed of the feature point per unit time. The displacement rate values ​​for all valid tracking periods are aggregated, summed, and then divided by the number of valid tracking periods to obtain the average displacement rate value. Since the displacement rate value is based on the pixel level, it is necessary to consider parameters such as the camera's focal length and pixel size to convert it into an angular rate of change in real space. This angular rate of change is the angle compensation value used to compensate for any angular deviation that may occur during the AGV docking process.

[0087] The sub-region feature point deviation value, sub-region image clarity value and angle compensation value obtained through these steps can reflect the status of the AGV during the docking and positioning process from different angles.

[0088] Example 5: See Figure 4The acquisition of coordinate conversion coefficients begins with the extraction of effective synchronization coefficients. During the effective tracking period, displacement synchronization coefficients are selected as effective synchronization coefficients. These effective synchronization coefficients are pre-verified parameters that meet preset conditions and can reflect the coordination relationship between feature point displacement and image frame interval. Each effective synchronization coefficient corresponds to a preset standard value. This standard value is determined under ideal operating conditions through multiple simulated docking scenarios and combined with device performance parameters. It represents the ideal matching state between feature point displacement and image frame interval. Each effective synchronization coefficient is compared with the corresponding standard value and the difference between the two is calculated. These differences are the synchronization coefficient deviations. The synchronization coefficient deviations reflect the degree of deviation between the actual synchronization state and the ideal state. The smaller the deviation, the closer the current synchronization state is to the ideal state. After obtaining all synchronization coefficient deviations, these deviations are statistically analyzed. The variance is calculated by calculating their dispersion, and this variance is the coordinate conversion coefficient. The coordinate conversion coefficient is used to adjust the conversion relationship between different coordinate systems to eliminate coordinate deviations caused by factors such as device installation errors and environmental interference.

[0089] The verification process of the comprehensive positioning correction value is based on the angle compensation value and coordinate conversion coefficient within the effective tracking period. The angle compensation value is obtained by analyzing the displacement rate of the feature point during the effective tracking period, and is used to correct the angle deviation caused by the steering error during the driving process of the AGV; the coordinate conversion coefficient is used to correct the error in the coordinate system conversion as mentioned above. The angle compensation value and the coordinate conversion coefficient are correlated and calculated to obtain a ratio for verification, which is called the correction verification coefficient. The correction verification coefficient can reflect the degree of matching between the angle compensation and the coordinate conversion. If the degree of matching between the two is high, it means that the comprehensive positioning correction value can accurately reflect the actual deviation of the AGV. The preset verification range is a reasonable interval set according to the docking accuracy requirements and the equipment response characteristics. This range is determined by a large amount of actual test data to ensure that the correction verification coefficient within this range can guarantee the validity of the comprehensive positioning correction value. When the correction verification coefficient is within the preset verification range, it indicates that the matching status of the angle compensation value and the coordinate conversion coefficient meets the requirements, and the comprehensive positioning correction value can be used to adjust the positioning parameters of the AGV vehicle; if the correction verification coefficient exceeds the preset verification range, it is necessary to recheck the selection of the effective tracking period, the calculation of the angle compensation value, and the acquisition process of the coordinate conversion coefficient, eliminate abnormal factors in the data collection or calculation process, and recalculate the relevant parameters until the correction verification coefficient falls within the preset verification range to ensure that the comprehensive positioning correction value can accurately guide the docking positioning adjustment of the AGV vehicle.

[0090] Throughout the implementation process, the accuracy of the effective tracking period must be ensured, as its quality directly impacts the reliability of the angle compensation value, coordinate conversion coefficient, and correction verification coefficient. Furthermore, the standard value of the synchronization coefficient and the preset verification range must be set based on the specific docking scenario and device parameters to avoid deviations in verification results due to improper parameter settings.

[0091] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0092] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The AGV trolley docking positioning method based on visual positioning technology is characterized by: The following steps are involved: Step 1: Capture an image of the target docking area to obtain a visual image containing docking marks, wherein the visual image includes feature point distribution information, and obtain initial positioning parameters by identifying the feature point distribution information; Step 2: Based on the initial positioning parameters, the initial positioning parameters are compared with the standard positioning parameter range. If the initial positioning parameters exceed the standard positioning parameter range, a positioning deviation signal is generated; Step 3: Based on the positioning deviation signal, track the displacement data of the feature points in the visual image. During the tracking period, obtain the displacement change sequence, obtain the feature point offset, and compare it with the offset threshold to generate the offset impact level signal; If the feature point offset is greater than or equal to the offset threshold, a high-level offset signal is generated; Step 4: Based on the high-level offset signal, obtain the coordinate conversion coefficient and angle compensation value, and superimpose the coordinate conversion coefficient and the angle compensation value to obtain the comprehensive positioning correction value; Step 5: Based on the high-level offset signal, the dynamic adjustment coefficient is obtained, and the current driving speed parameter is integrated with the dynamic adjustment coefficient to obtain the required adjustment driving parameter, thus completing the adjustment of the AGV driving state; The dynamic adjustment coefficient is obtained as follows: The weighted sum of all feature point offsets is taken to get the average value to obtain the dynamic adjustment coefficient; The angle compensation value is obtained as follows: Obtain the pixel displacement value of the feature point and the image frame interval value within the effective tracking period, calculate the ratio of the pixel displacement value of the feature point and the image frame interval value to obtain the displacement rate value, accumulate and average the displacement rate values ​​within each effective tracking period to obtain the angle compensation value; The coordinate conversion coefficient is obtained as follows: Based on the displacement synchronization coefficient corresponding to the above effective tracking period, marked as the effective synchronization coefficient, all the effective synchronization coefficients in the tracking period are calculated with the synchronization coefficient standard value to obtain the synchronization coefficient deviation value, and the synchronization coefficient deviation value is calculated with the variance to obtain the coordinate conversion coefficient; The verification method of the comprehensive positioning correction value is: obtain the angle compensation value and the coordinate conversion coefficient within the effective tracking period, calculate the ratio of the angle compensation value and the coordinate conversion coefficient, and obtain the correction verification coefficient. If the correction verification coefficient is within the preset verification range, the comprehensive positioning correction value is confirmed to be valid.

2. The AGV docking and positioning method based on visual positioning technology according to claim 1 is characterized in that: The initial positioning parameters are obtained as follows: The docking area in the visual image is divided into several positioning sub-areas, the number of feature points in each positioning sub-area is obtained, and the number of feature points in each positioning sub-area is weighted and averaged to obtain the feature point density value; The ratio of the feature point density value to the standard density value is calculated to obtain the initial positioning parameters.

3. The AGV docking and positioning method based on visual positioning technology according to claim 2 is characterized in that: The method for obtaining the feature point offset is: The horizontal offset component and the vertical offset component are obtained, and vector synthesis of the horizontal offset component and the vertical offset component is performed to obtain the feature point offset.

4. The AGV docking and positioning method based on visual positioning technology according to claim 3 is characterized in that: The lateral offset component is obtained as follows: Calculate the ratio of the image frame interval value and the feature point pixel displacement value within the same tracking period to obtain the displacement synchronization coefficient, select the period where the displacement synchronization coefficient meets the preset conditions, and mark it as the valid tracking period; The duration of the effective tracking period is obtained, and the ratio of the duration of the effective tracking period to the total duration of the tracking cycle is calculated to obtain the lateral offset component.

5. The AGV docking and positioning method based on visual positioning technology according to claim 4 is characterized in that: The image frame interval value and the feature point pixel displacement value are obtained as follows: During the tracking period, the tracking period is divided into several time segments, and a time-displacement coordinate system is established. The horizontal dimension represents the time segment, and the vertical dimension represents the pixel position of the feature point corresponding to each time segment. The obtained pixel position of the feature point is substituted into the time-displacement coordinate system to draw the pixel displacement change curve; Based on the pixel displacement change curve, the pixel displacement peak point and the pixel displacement valley point are obtained, and the coordinates of the adjacent pixel displacement peak point and the pixel displacement valley point are calculated to obtain the pixel displacement value of the feature point; During the tracking period, the tracking period is divided into several time segments, and a time-frame number coordinate system is established. The horizontal dimension represents the time segment, and the vertical dimension represents the image acquisition frame number corresponding to each time segment. The obtained image acquisition frame number is substituted into the time-frame number coordinate system to draw a frame number change curve. Based on the frame number change curve, the frame number increasing points and the frame number decreasing points are obtained, and the coordinates of adjacent frame number increasing points and frame number decreasing points are calculated to obtain the image frame interval value.

6. The AGV docking and positioning method based on visual positioning technology according to claim 5 is characterized in that: The longitudinal offset component is obtained as follows: Compare the sub-region feature point deviation value with the sub-region deviation allowable value. The comparison process is as follows: If the deviation value of the feature point in the sub-region is greater than or equal to the sub-region deviation allowable value, the sub-region is marked as a deviation abnormal sub-region; If the deviation value of the feature point in the sub-region is less than the sub-region deviation allowable value, the sub-region is marked as a normal deviation sub-region; Compare the sub-region image clarity value with the sub-region clarity threshold. The comparison process is as follows: If the sub-region image clarity value is less than the sub-region clarity threshold, the sub-region is marked as a clarity abnormal sub-region; If the sub-region image clarity value is greater than or equal to the sub-region clarity threshold, the sub-region is marked as a sub-region with normal clarity; Obtaining an offset abnormal subregion, performing spatial overlap analysis on the offset abnormal subregion and the clarity abnormal subregion to obtain overlapping abnormal subregions, obtaining the number of overlapping abnormal subregions, and accumulating and summing the number of overlapping abnormal subregions to obtain the total number of abnormal regions; The longitudinal offset component is obtained by calculating the ratio of the total number of abnormal areas to the total number of positioning sub-areas.

7. The AGV docking and positioning method based on visual positioning technology according to claim 6 is characterized in that: The method for obtaining the deviation value of the sub-region feature point is: The visual image is divided into regions to obtain positioning sub-region images. Based on the positioning sub-region images, the actual coordinates of the feature points in each positioning sub-region are extracted and marked as the sub-region measured coordinate values. The sub-region measured coordinate values ​​are calculated with the sub-region standard coordinate values ​​to obtain the sub-region feature point deviation value; The method for obtaining the sub-region image clarity value is: Perform clarity detection on the visual image, divide the visual image into several positioning sub-areas, obtain the image grayscale gradient value in each positioning sub-area, mark it as the sub-area measured gradient value, calculate the difference between the sub-area measured gradient value and the sub-area standard gradient value, and obtain the sub-area image clarity value.

Images