An intelligent material table robot system based on AI vision technology

Abstract

The present application relates to the technical field of material table robot, especially to an intelligent material table robot system based on AI vision technology, the present application is provided with material table control module, feature acquisition module, material analysis module, material identification module and grabbing control module, the material analysis module determines the gray level fluctuation of the to-be-sorted material according to the gray level characteristic parameters in the grabbing and identification area under different illumination angles, marks the grabbing and identification area with features, the material identification module determines the occlusion tendency sub-area, constructs the occlusion tendency vector, and the grabbing control module determines the grabbing compensation mode of the to-be-sorted material according to the comparison between the transmission direction vector and the occlusion tendency vector. The present application realizes the marking of the grabbing and identification area with features and occlusion interference according to the gray level of the material, adaptively adjusts the grabbing compensation mode based on the change trend of dynamic occlusion, and improves the efficiency and reliability of the intelligent material table robot system.

Description

Technical Field

[0001] This invention relates to the field of material handling robot technology, and in particular to an intelligent material handling robot system based on AI vision technology. Background Technology

[0002] In industrial automation and intelligent manufacturing scenarios, vision-based material handling robot systems serve as a core link connecting material storage and production execution. They are widely used in electronic manufacturing, automotive parts assembly, and logistics sorting, undertaking automated tasks such as material identification, positioning, grasping, and conveying. By acquiring image information of the material handling area through AI vision systems and combining it with deep learning algorithms, they achieve accurate material identification and pose estimation. Then, robotic arms and other actuators complete the grasping and sorting, replacing manual operations and improving production efficiency and stability. However, the lighting environment in industrial workshops is complex and variable. The superposition of natural and artificial light, the dynamic changes in equipment shadows, and the reflective properties of the materials themselves can affect the accuracy of visual recognition. Furthermore, the stacking of materials on the material handling platform, the movement of the robotic arm itself causing occlusion, and the temporary intervention of surrounding equipment can lead to material obstruction, making it difficult to adaptively adjust the grasping method according to the actual situation in the dynamic scene of the material handling platform. This affects the recognition accuracy and response efficiency of the AI ​​vision-based intelligent material handling robot system. Therefore, improving the efficiency and reliability of intelligent material handling robot systems is an urgent technical problem to be solved.

[0003] For example, Chinese Patent Publication No. CN115366152B discloses a robot vision automated grasping system, including a transport mechanism and a material sorting mechanism. The transport mechanism includes a workbench, a transport device, and a grasping component for grasping materials on the transport device. Both the transport device and the grasping component are mounted on the workbench, and the grasping component is equipped with a capture camera. The material sorting mechanism includes a material sorting chamber, a magnetic suction device, and a discharge conveyor belt and a recycling conveyor belt located on both sides of the transport device. The material sorting chamber is covered above the discharge end of the transport device. The magnetic suction device is installed inside the material sorting chamber and is used to transfer materials from the transport device to the discharge conveyor belt or the recycling conveyor belt. The material sorting chamber is equipped with a discharge camera and a recycling camera.

[0004] The following problems still exist in the existing technology:

[0005] Existing technologies do not consider the dynamic interference of lighting and occlusion in industrial environments, which affects the material recognition and gripping of the material-table robot. Existing technologies cannot mark the gripping and recognition areas with occlusion interference based on the grayscale characteristics of the material, nor can they adaptively adjust the gripping compensation method according to the changing trend of dynamic occlusion, thus affecting the efficiency and reliability of the intelligent material-table robot system. Summary of the Invention

[0006] To address this, the present invention provides an intelligent material handling robot system based on AI vision technology, which overcomes the problems of existing technologies that cannot mark the grasping and recognition area with occlusion interference based on the grayscale characteristics of the material, and cannot adaptively adjust the grasping compensation method according to the changing trend of dynamic occlusion, thus affecting the efficiency and reliability of the intelligent material handling robot system.

[0007] To achieve the above objectives, the present invention provides an intelligent material handling robot system based on AI vision technology, comprising:

[0008] The material platform control module includes a gripping unit for gripping materials to be sorted on the material platform and a conveying unit for conveying the materials to be sorted.

[0009] The feature acquisition module is connected to the material platform control module to acquire the grayscale characterization parameters and conveying direction vector of the material to be sorted.

[0010] The material analysis module, which is connected to the feature acquisition module, is used to divide the material platform into several gripping and recognition areas. Based on the grayscale characterization parameters in the gripping and recognition areas under different lighting angles, the grayscale fluctuation of the material to be sorted is determined to mark the feature gripping and recognition areas.

[0011] The material identification module is connected to the feature acquisition module and the material analysis module respectively. It determines the occlusion tendency sub-region based on the grayscale characterization parameters of the feature capture and identification area under the preset illumination angle, and constructs the occlusion tendency sub-vector based on the acquired occlusion tendency sub-regions.

[0012] The grasping control module is connected to the material platform control module, the feature acquisition module, and the material recognition module, respectively. It is used to determine the grasping compensation method for the material to be sorted based on the comparison between the conveying direction vector and the occlusion tendency vector. The grasping time is adjusted to grasp the material to be sorted within the feature grasping recognition area.

[0013] Alternatively, based on the occlusion tendency vector, determine whether there is a risk of abnormal grasping in the feature grasping and recognition region, and adjust the grasping compensation parameters of the feature grasping and recognition region accordingly;

[0014] The occlusion tendency vector is determined based on several occlusion tendency sub-vectors.

[0015] Furthermore, the material analysis module is used to determine the grayscale fluctuation representation quantity and the occlusion tendency coefficient based on the grayscale representation parameters within the grasping and recognition area, wherein,

[0016] The grayscale fluctuation characterization quantity is the difference between the maximum value and the minimum value of the grayscale characterization parameter within the capture and recognition area under the same illumination angle;

[0017] The occlusion tendency coefficient is the variance of the grayscale characterization parameter under different illumination angles.

[0018] Furthermore, the material analysis module is used to mark the grasping and recognition area as a feature grasping and recognition area based on the determination results of whether the grayscale fluctuation characterization quantity and the occlusion tendency coefficient within the grasping and recognition area meet the feature grasping and recognition area conditions under different illumination angles.

[0019] The feature capture and recognition region condition is that the grayscale fluctuation representation quantity exceeds the preset grayscale fluctuation representation quantity threshold, and the occlusion tendency coefficient does not exceed the preset occlusion tendency coefficient threshold.

[0020] Furthermore, the material recognition module determines the feature sub-region as the occlusion tendency sub-region based on the determination result that the grayscale representation parameters of the feature sub-region of the feature grasping and recognition area meet the occlusion tendency sub-region condition, wherein,

[0021] The material recognition module divides the feature grasping and recognition area into several sub-regions and obtains grayscale characterization parameters at several location points of the material to be sorted within the sub-regions.

[0022] The occlusion tendency sub-region condition is that the gray-scale representation parameter of the feature sub-region does not exceed the preset gray-scale representation parameter threshold, and the feature sub-region is the sub-region where the minimum gray-scale representation parameter value is located within the feature grasping and recognition area.

[0023] Furthermore, the material identification module is used to construct an occlusion trend sub-vector for the next acquisition time within the adjacent acquisition times based on the occlusion trend sub-regions at adjacent acquisition times within a preset monitoring period, wherein,

[0024] The material recognition module acquires several occlusion-oriented sub-regions at several collection times within the feature capture and recognition area;

[0025] The occlusion directional sub-vector is constructed with the center point of the occlusion directional sub-region of the previous acquisition time in the adjacent acquisition time as the starting point of the occlusion directional sub-vector, and the center point of the occlusion directional sub-region of the next acquisition time in the adjacent acquisition time as the ending point of the occlusion directional sub-vector.

[0026] Furthermore, the grasping control module is used to determine the grasping compensation method for the materials to be sorted within the feature grasping recognition area, wherein,

[0027] If the conveying direction vector and the occlusion tendency vector of the material to be sorted in the feature grasping recognition area meet the first grasping compensation condition, then the grasping control module determines the grasping compensation method as adjusting the grasping time of grasping the material to be sorted in the feature grasping recognition area.

[0028] If the conveying direction vector and the occlusion tendency vector of the material to be sorted within the feature grasping recognition area do not meet the first grasping compensation condition, the grasping control module determines the grasping compensation method as determining whether there is a grasping abnormality risk in the feature grasping recognition area and adjusting the grasping compensation parameters.

[0029] Furthermore, the first capture compensation condition is that the relative offset tendency parameter exceeds a preset relative offset tendency parameter threshold. The relative offset tendency parameter is the vector angle between the transmission direction vector and the occlusion tendency vector. The occlusion tendency vector is the vector obtained by adding the occlusion tendency sub-vectors at several monitoring times.

[0030] Furthermore, the grasping control module is used to adjust the grasping time, wherein,

[0031] The delay duration of the capture moment is negatively correlated with the relative offset tendency parameter.

[0032] Furthermore, the grasping control module is used to determine whether there is a risk of grasping anomalies in the feature grasping and recognition region, wherein,

[0033] The grasping control module determines that the feature grasping recognition area has a grasping anomaly risk based on the judgment result that the occlusion tendency vector of the feature grasping recognition area meets the grasping anomaly conditions;

[0034] Based on the determination result that the occlusion tendency vector of the feature grasping recognition area does not meet the grasping anomaly conditions, it is determined that there is no grasping anomaly risk in the feature grasping recognition area, and the materials to be sorted in the feature grasping recognition area are grasped based on the grasping compensation parameters.

[0035] The capture anomaly condition is that the risk tendency parameter exceeds the preset risk tendency parameter threshold. The risk tendency parameter is the absolute value of the difference in the magnitude of the occlusion tendency vector at adjacent acquisition times.

[0036] Furthermore, the grasping control module is used to determine grasping compensation parameters, which include grasping direction and grasping speed, wherein,

[0037] The grasping direction is the direction of the material tendency vector, which is the vector obtained by adding the occlusion tendency vector and the conveying direction vector.

[0038] The reduction in gripping speed is positively correlated with the magnitude of the occlusion tendency vector. Compared with the prior art, the beneficial effects of the present invention are that it sets up a material platform control module, a feature acquisition module, a material analysis module, a material identification module, and a gripping control module. The feature acquisition module acquires the grayscale characterization parameters and conveying direction vector of the material to be sorted. The material analysis module determines the grayscale fluctuation of the material to be sorted based on the grayscale characterization parameters in the gripping identification area under different illumination angles to mark the feature gripping identification area. The material identification module determines the occlusion tendency sub-regions based on the grayscale characterization parameters of the feature gripping identification area. Based on the acquired occlusion tendency sub-regions, an occlusion tendency sub-vector is constructed. The gripping control module determines the gripping compensation method for the material to be sorted based on the comparison between the conveying direction vector and the occlusion tendency vector. Thus, it realizes the marking of gripping identification areas with occlusion interference based on the grayscale characteristics of the material, and adaptively adjusts the gripping compensation method based on the dynamic occlusion change trend, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0039] In particular, this invention uses a material analysis module to determine the grayscale fluctuation of the material to be sorted based on the grayscale characterization parameters within the grasping and recognition area under different lighting angles. This is used to mark the grasping and recognition areas based on their characteristics. It can be understood that by using grayscale fluctuation characterization parameters, areas with obvious characteristics are selected, avoiding invalid analysis of indistinguishable areas. Simultaneously, areas affected by lighting interference are excluded using an occlusion tendency coefficient, reducing misjudgments caused by lighting changes, such as shadows being misjudged as material occlusion. Even in scenarios with unstable lighting angles, such as changes in natural light in the workshop or vibrations in equipment light sources, the system can still stably identify material characteristic areas, enhancing its resistance to environmental interference and providing reliable analysis objects for subsequent grasping and control. By selecting key characteristic grasping and recognition areas, the full analysis of the entire material platform area is avoided, reducing invalid data processing and improving system operating efficiency. Therefore, this invention achieves the marking of grasping and recognition areas with occlusion interference based on the grayscale characteristics of the material, improving the efficiency and reliability of the intelligent material platform robot system.

[0040] In particular, this invention determines occlusion trend sub-regions based on the grayscale representation parameters of the feature grasping and recognition area through a material recognition module. Based on the acquired occlusion trend sub-regions, an occlusion trend sub-vector is constructed. Essentially, by determining the occlusion trend sub-regions through the grayscale values ​​of sub-regions within the feature grasping and recognition area, the invention locks in areas with occlusion shadows within the feature grasping and recognition area. By quantifying the movement direction and amplitude of the occlusion shadows, its dynamic changes are reflected in real time, providing predictive information for the intelligent material platform robot, capturing dynamic occlusion trends, and improving the system's adaptability to dynamic environments. This invention, through the material recognition module determining occlusion trend sub-regions based on the grayscale representation parameters of the feature grasping and recognition area, and constructing an occlusion trend sub-vector based on the acquired occlusion trend sub-regions, further filters occlusion trend sub-regions and constructs occlusion trend sub-vectors, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0041] In particular, this invention determines the gripping compensation method for the materials to be sorted by the gripping control module based on the comparison between the conveying direction vector and the occlusion tendency vector. This means that by observing the relative movement trend between the occluder and the material, the duration and risk level of the occlusion interference can be predicted, thereby selecting a targeted response strategy. The conveying direction vector represents the movement direction of the material to be sorted, such as the direction of movement with the conveyor belt or the direction of material pushing, reflecting the spatial positional change of the material itself. The occlusion tendency vector is determined by the occlusion tendency sub-vectors at several monitoring moments, representing the occluder, such as other stacked materials or the entirety of the moving robotic arm. The motion trend reflects the spatial positional change of the occlusion source. The angle between two vectors can characterize the relative motion relationship between the occluder and the material. The larger the angle, the more significant the difference in the motion direction between the two. The smaller the angle, the closer the motion direction between the two. Different grasping compensation methods are adaptively adjusted for different comparison situations to improve the accuracy of grasping timing, reduce the collision risk under continuous occlusion, balance efficiency and safety, and enhance the system's adaptability to dynamic scenes. Thus, the grasping compensation method is adaptively adjusted based on the changing trend of dynamic occlusion, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0042] In particular, this invention, through a gripping control module, adjusts the gripping time of the material to be sorted within the feature gripping recognition area when the feature gripping recognition area meets the first gripping compensation condition. It is understood that in actual material sorting, the speed and direction of the obstructing object may dynamically change. By adjusting the delay in real time through the vector angle, the system can adapt to the obstruction rhythm under different working conditions. The feature gripping recognition area meets the first gripping compensation condition, meaning the relative motion trends of the obstructing object and the material show an inconsistent trend. The larger the relative offset tendency parameter, the faster the obstruction is released, and shortening the delay time can reduce... The dwell time of materials on the conveyor belt is controlled to avoid subsequent material accumulation due to excessive waiting, thereby increasing the sorting volume per unit time, avoiding ineffective waiting, and improving gripping efficiency. The smaller the relative offset tendency parameter, the slower the speed of obstruction removal. Extending the delay time can prevent the robotic arm from accidentally touching obstructions due to premature gripping, reducing the risk of material damage and equipment failure, ensuring obstruction removal, improving gripping accuracy, reducing energy consumption and mechanical wear, and dynamically adapting to the obstruction rhythm. In this way, the gripping compensation method can be adaptively adjusted based on the changing trend of dynamic obstruction, thereby improving the efficiency and reliability of the intelligent material table robot system.

[0043] In particular, this invention, through its grasping control module, determines whether there is a grasping anomaly risk in the feature grasping recognition area when it does not meet the first grasping compensation condition, and adjusts the grasping compensation parameters accordingly. It can be understood that when the feature grasping recognition area does not meet the first grasping compensation condition, the relative movement trend of the obstruction and the material shows a relatively consistent trend, and the shadow cast by the obstruction on the material will not change due to material transport. By monitoring the rate of change of the magnitude of the obstruction tendency vector, i.e., the risk tendency parameter, the invention can quickly identify changes in the obstruction state, such as sudden large-area obstruction or material position shift caused by rapid movement of the obstruction. For situations with abnormal risks, the invention can pause or issue a warning in advance to avoid performing grasping when the obstruction is unstable. The grasping direction is determined based on the material tendency vector, which is equivalent to incorporating the indirect impact of obstruction on the material into the trajectory calculation, thus improving the grasping... The system tracks the actual position of materials more accurately, and the gripping speed adjustment is determined based on the magnitude of the occlusion trend vector. This achieves a dynamic balance where accuracy is prioritized when the occlusion effect is strong, and efficiency is prioritized when the effect is weak. The larger the magnitude of the occlusion trend vector, the greater the speed adjustment is required, such as decelerating to improve positioning accuracy or accelerating to escape the occlusion area, ensuring that the gripping action can still be executed stably under interference. The smaller the magnitude of the occlusion trend vector, the smaller the speed adjustment, avoiding the decrease in sorting efficiency caused by excessive deceleration. Through a combination of risk assessment and dynamic parameter adjustment, the intelligent material table robot system can maintain efficient gripping under dynamic occlusion conditions, balancing gripping accuracy and efficiency while effectively avoiding potential abnormal risks. Furthermore, it achieves adaptive adjustment of the gripping compensation method based on the changing trend of dynamic occlusion, improving the efficiency and reliability of the intelligent material table robot system. Attached Figure Description

[0044] Figure 1 This is a functional block diagram of the intelligent material platform robot system based on AI vision technology according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart illustrating the logic of the material analysis module's feature capture and recognition area in an embodiment of the present invention.

[0046] Figure 3 This is a flowchart illustrating the logic of the material identification module in this embodiment of the invention for determining the occlusion-oriented sub-region.

[0047] Figure 4 This is a flowchart illustrating the logic of how the grasping control module determines the grasping compensation method for materials to be sorted within the feature grasping recognition area in this embodiment of the invention. Detailed Implementation

[0048] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0049] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0050] It should be noted that in the description of this invention, the terms "upper," "lower," "inner," "outer," etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0051] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation" and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0052] Please see Figure 1 The diagram shown is a functional block diagram of an intelligent material platform robot system based on AI vision technology according to an embodiment of the present invention. The intelligent material platform robot system based on AI vision technology of the present invention includes:

[0053] The material platform control module includes a gripping unit for gripping materials to be sorted on the material platform and a conveying unit for conveying the materials to be sorted.

[0054] Specifically, the embodiments of the present invention do not limit the specific structure of the gripping unit and the conveying unit. Preferably, the gripping unit can be a multi-degree-of-freedom robotic arm with an end effector, such as a pneumatic gripper, a vacuum suction cup, or a magnetic gripper, to grip the materials to be sorted on the material platform. The conveying unit can be a belt conveyor to convey the materials to be sorted in one direction. This will not be described in detail.

[0055] The feature acquisition module is connected to the material platform control module to acquire the grayscale characterization parameters and conveying direction vector of the material to be sorted.

[0056] Specifically, the embodiments of the present invention do not limit the specific structure of the feature acquisition module. Preferably, it can be an industrial camera with a microprocessor to acquire the grayscale characterization parameters and the conveying direction vector of the material to be sorted. The direction of the conveying direction vector is the conveying direction of the conveying mechanism, such as the belt conveyor, where the material to be sorted is located. This will not be elaborated further.

[0057] The material analysis module, which is connected to the feature acquisition module, is used to divide the material platform into several gripping and recognition areas. Based on the grayscale characterization parameters in the gripping and recognition areas under different lighting angles, the grayscale fluctuation of the material to be sorted is determined to mark the feature gripping and recognition areas.

[0058] Specifically, the area of ​​the grasping and recognition region is the product of the material platform area and the first region division factor. The first region division factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the smaller the first region division factor is set. The value range of the first region division factor can be [0.1, 0.3], preferably 0.2.

[0059] Specifically, the single change amount and the number of changes of the illumination angle can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the smaller the single change amount and the more changes. The value range of the single change amount can be [5, 10], with the interval unit being °. The value range of the number of changes can be [3, 8], with the interval unit being times. Preferably, the single change amount can be 6° and the number of changes can be 5 times.

[0060] Specifically, the embodiments of the present invention do not limit the specific structure of the material analysis module. Preferably, it can be a processor used in a computer to determine the grayscale fluctuation of the material to be sorted and mark the feature grasping and recognition area, which will not be elaborated further.

[0061] The material identification module is connected to the feature acquisition module and the material analysis module respectively. It determines the occlusion tendency sub-region based on the grayscale characterization parameters of the feature capture and identification area under the preset illumination angle, and constructs the occlusion tendency sub-vector based on the acquired occlusion tendency sub-regions.

[0062] Specifically, the embodiments of the present invention do not limit the specific structure of the material identification module. Preferably, it can be a microprocessor used to determine the occlusion directional sub-region and construct the occlusion directional sub-vector, which will not be elaborated further.

[0063] Specifically, the preset illumination angle can be determined by those skilled in the art based on the visual recognition accuracy of materials of the same material to be grasped in historical data. The value range of the preset illumination angle can be [30, 60], with the interval unit being °. Preferably, it can be 45°.

[0064] The grasping control module is connected to the material platform control module, the feature acquisition module, and the material recognition module, respectively. It is used to determine the grasping compensation method for the material to be sorted based on the comparison between the conveying direction vector and the occlusion tendency vector. The grasping time is adjusted to grasp the material to be sorted within the feature grasping recognition area.

[0065] Alternatively, based on the occlusion tendency vector, determine whether there is a risk of abnormal grasping in the feature grasping and recognition region, and adjust the grasping compensation parameters of the feature grasping and recognition region accordingly;

[0066] The occlusion tendency vector is determined based on several occlusion tendency sub-vectors.

[0067] Specifically, the embodiments of the present invention do not limit the specific structure of the gripping control module. Preferably, it can be a microprocessor used to determine the gripping compensation method for the materials to be sorted, adjust the gripping time, and gripping compensation parameters, which will not be elaborated further.

[0068] Specifically, the material analysis module is used to determine the grayscale fluctuation representation quantity and the occlusion tendency coefficient based on the grayscale representation parameters within the grasping and recognition area, wherein,

[0069] The grayscale fluctuation characterization quantity is the difference between the maximum value and the minimum value of the grayscale characterization parameter within the capture and recognition area under the same illumination angle;

[0070] The occlusion tendency coefficient is the variance of the grayscale characterization parameter under different illumination angles.

[0071] Please see Figure 2As shown, this is a logical flowchart of the material analysis module marking the feature grasping and recognition region in an embodiment of the present invention. The material analysis module marks the grasping and recognition region as a feature grasping and recognition region based on the determination results of the grayscale fluctuation characterization quantity and the occlusion tendency coefficient within the grasping and recognition region under different illumination angles meeting the conditions for feature grasping and recognition region.

[0072] If the grayscale fluctuation representation and the occlusion tendency coefficient within the grasping and recognition area under different lighting angles do not meet the feature grasping and recognition area conditions, then the grasping and recognition area will not be marked.

[0073] The feature capture and recognition region condition is that the grayscale fluctuation representation quantity exceeds the preset grayscale fluctuation representation quantity threshold, and the occlusion tendency coefficient does not exceed the preset occlusion tendency coefficient threshold.

[0074] Specifically, the preset threshold for grayscale fluctuation is the product of the grayscale fluctuation reference value and the grayscale fluctuation coefficient. The preset threshold for occlusion tendency coefficient is the product of the occlusion tendency coefficient reference value and the occlusion tendency factor. The grayscale fluctuation reference value is the average value of the grayscale fluctuation under the same working conditions in historical data. The occlusion tendency coefficient reference value is the average value of the occlusion tendency coefficient under the same working conditions in historical data. The grayscale fluctuation coefficient and the occlusion tendency factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the larger the grayscale fluctuation coefficient is set, and the smaller the occlusion tendency factor is set. The range of grayscale fluctuation coefficient can be [1.2, 1.3], and the range of occlusion tendency factor can be [1.1, 1.25]. Preferably, the grayscale fluctuation coefficient can be 1.25, and the occlusion tendency factor can be 1.15.

[0075] Specifically, in this embodiment of the invention, the material analysis module determines the grayscale fluctuation of the material to be sorted based on the grayscale characterization parameters within the grasping and recognition area under different lighting angles. This is used to mark the grasping and recognition area based on its features. It can be understood that by using the grayscale fluctuation characterization parameters, areas with obvious features are selected, avoiding invalid analysis of indistinguishable areas. Simultaneously, areas affected by lighting interference are excluded using the occlusion tendency coefficient, reducing misjudgments caused by lighting changes, such as shadows being misjudged as material occlusion. Even in scenarios with unstable lighting angles, such as changes in natural light in the workshop or vibrations in the equipment light source, the system can still stably identify material feature areas, enhancing its resistance to environmental interference and providing reliable analysis objects for subsequent grasping and control. By selecting key feature grasping and recognition areas, full analysis of the entire material platform area is avoided, reducing invalid data processing and improving system operating efficiency. Therefore, the system achieves the marking of grasping and recognition areas with occlusion interference based on the grayscale characteristics of the material, improving the efficiency and reliability of the intelligent material platform robot system.

[0076] Specifically, it can be understood that, based on the differentiated performance of the material's own characteristics and environmental interference (lighting, occlusion), key areas are screened through quantitative analysis, providing reliable analytical objects for subsequent accurate identification and grasping. The grayscale fluctuation characterization quantity, which is the difference between the maximum and minimum values ​​of the grayscale characterization parameters within the grasping and identification area under the same lighting angle, can characterize whether there is a shadow phenomenon in that area. The larger the grayscale fluctuation characterization quantity, the more likely there is a shadow phenomenon in the area. Areas with shadow phenomena are screened out. The occlusion tendency coefficient, which is the variance of the grayscale characterization parameters under different lighting angles, can characterize the sensitivity of the grayscale features of the area to changes in the lighting angle. If the shadow is caused by occlusion interference, its position and grayscale features are relatively stable and less affected by changes in the lighting angle, resulting in a small occlusion tendency coefficient. If the shadow is caused by changes in the lighting angle, the grayscale features will fluctuate with the lighting angle, resulting in a larger occlusion tendency coefficient. Thus, the grasping and identification areas with occlusion interference are marked according to the grayscale features of the material, improving the efficiency and reliability of the intelligent material platform robot system.

[0077] Please see Figure 3 As shown, this is a flowchart illustrating the logic of the material identification module in this embodiment of the invention for determining the occlusion-oriented sub-region. The material identification module determines the feature sub-region as the occlusion-oriented sub-region based on the determination result that the grayscale representation parameters of the feature sub-region of the feature capture and identification area meet the conditions for an occlusion-oriented sub-region.

[0078] The material recognition module divides the feature grasping and recognition area into several sub-regions and obtains grayscale characterization parameters at several location points of the material to be sorted within the sub-regions.

[0079] If the grayscale representation parameters of the feature sub-region of the feature capture and recognition area do not meet the occlusion tendency sub-region conditions, then the feature sub-region will not be filtered.

[0080] The occlusion tendency sub-region condition is that the gray-scale representation parameter of the feature sub-region does not exceed the preset gray-scale representation parameter threshold, and the feature sub-region is the sub-region where the minimum gray-scale representation parameter value is located within the feature grasping and recognition area.

[0081] Specifically, the preset grayscale representation parameter threshold is the product of the grayscale representation parameter reference value and the grayscale representation factor. The grayscale representation parameter reference value is the average value of the grayscale representation parameters under the same working conditions in historical data. The grayscale representation factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the smaller the grayscale representation factor is set. The value range of the grayscale representation factor can be [1.12, 1.25], preferably 1.15.

[0082] Specifically, the area of ​​the sub-region is the product of the area of ​​the feature grasping and recognition area and the second region division factor. The second region division factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the smaller the second region division factor is set. The value range of the second region division factor can be [0.1, 0.3], preferably 0.2.

[0083] Specifically, the material identification module is used to construct an occlusion trend sub-vector for the next acquisition time within adjacent acquisition times based on the occlusion trend sub-regions at adjacent acquisition times within a preset monitoring period, wherein,

[0084] The material recognition module acquires several occlusion-oriented sub-regions at several collection times within the feature capture and recognition area;

[0085] The occlusion directional sub-vector is constructed with the center point of the occlusion directional sub-region of the previous acquisition time in the adjacent acquisition time as the starting point of the occlusion directional sub-vector, and the center point of the occlusion directional sub-region of the next acquisition time in the adjacent acquisition time as the ending point of the occlusion directional sub-vector.

[0086] Specifically, the preset monitoring period and the interval between adjacent acquisition times of the preset monitoring period can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the shorter the preset monitoring period and the shorter the interval. The value range of the preset monitoring period can be [20, 40], with the interval unit being min, and the value range of the interval can be [30, 50], with the interval unit being s. Preferably, the preset monitoring period can be 30 min and the interval can be 45 s.

[0087] Specifically, in this embodiment of the invention, the material recognition module determines the occlusion trend sub-regions based on the grayscale representation parameters of the feature grasping and recognition area. Based on the acquired occlusion trend sub-regions, an occlusion trend sub-vector is constructed. In essence, the occlusion trend sub-regions are determined by the grayscale values ​​of the sub-regions within the feature grasping and recognition area, locking in areas with occlusion shadows within the feature grasping and recognition area. By quantifying the movement direction and amplitude of the occlusion shadows, their dynamic changes are reflected in real time, providing predictive information for the intelligent material platform robot, capturing dynamic occlusion trends, and improving the system's adaptability to dynamic environments. This embodiment of the invention, through the material recognition module determining the occlusion trend sub-regions based on the grayscale representation parameters of the feature grasping and recognition area, and constructing an occlusion trend sub-vector based on the acquired occlusion trend sub-regions, further filters the occlusion trend sub-regions and constructs the occlusion trend sub-vector, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0088] Specifically, it can be understood that the occlusion tendency sub-region is the shadow area projected onto the surface of the material to be grasped by the occlusion source. The shadow area is the area of ​​light deficiency formed on the surface of the material to be grasped after the occlusion object blocks the light source. When the surface of the material is blocked by an occlusion object, such as other materials or robotic arm components, the light source is blocked, and this area can only receive weak scattered light from the environment, and the intensity of reflected light is significantly reduced. In the material table environment, the shadow area projected onto the surface of the material to be grasped by the occlusion source may be affected by slight shaking of other materials or external force, or changes in the position of robotic arm components. The occlusion trend sub-region center point at adjacent acquisition times represents the shadow position of the occlusion source projected onto the surface of the material to be grasped at the previous moment and the shadow position of the occlusion source projected onto the surface of the material to be grasped at the current moment, respectively. The vector constructed with the two points as the starting point and the ending point is the occlusion trend sub-vector, which can intuitively quantify the movement direction and movement range of the occlusion source relative to the material to be grasped, thereby capturing the dynamic trend of the occlusion source relative to the material to be grasped. In this way, the occlusion trend sub-region is screened and the occlusion trend sub-vector is constructed, which improves the efficiency and reliability of the intelligent material platform robot system.

[0089] Please see Figure 4 The diagram shown is a logic flowchart illustrating how the gripping control module determines the gripping compensation method for the materials to be sorted within the feature gripping recognition area, according to an embodiment of the present invention. The gripping control module is used to determine the gripping compensation method for the materials to be sorted within the feature gripping recognition area.

[0090] If the conveying direction vector and the occlusion tendency vector of the material to be sorted in the feature grasping recognition area meet the first grasping compensation condition, then the grasping control module determines the grasping compensation method as adjusting the grasping time of grasping the material to be sorted in the feature grasping recognition area.

[0091] If the conveying direction vector and the occlusion tendency vector of the material to be sorted within the feature grasping recognition area do not meet the first grasping compensation condition, the grasping control module determines the grasping compensation method as determining whether there is a grasping abnormality risk in the feature grasping recognition area and adjusting the grasping compensation parameters.

[0092] Specifically, in this embodiment of the invention, the gripping control module determines the gripping compensation method for the materials to be sorted based on the comparison between the conveying direction vector and the occlusion tendency vector. This means that by analyzing the relative movement trend between the occluder and the material, the duration and risk level of the occlusion interference are predicted, thereby selecting a targeted response strategy. The conveying direction vector represents the movement direction of the material to be sorted, such as the direction of movement with the conveyor belt or the direction of material pushing, reflecting the spatial positional change of the material itself. The occlusion tendency vector is determined by several occlusion tendency sub-vectors at several monitoring moments, representing the occluder, such as other stacked materials or a moving robotic arm. The overall motion trend reflects the spatial positional change of the occlusion source. The angle between the two vectors can characterize the relative motion relationship between the occlusion object and the material. The larger the angle, the more significant the difference in the motion direction between the two. The smaller the angle, the closer the motion direction between the two. Different grasping compensation methods are adaptively adjusted for different comparison situations to improve the accuracy of grasping timing, reduce the collision risk under continuous occlusion, balance efficiency and safety, and enhance the system's adaptability to dynamic scenes. Thus, the grasping compensation method is adaptively adjusted based on the changing trend of dynamic occlusion, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0093] Specifically, the first capture compensation condition is that the relative offset tendency parameter exceeds a preset relative offset tendency parameter threshold. The relative offset tendency parameter is the angle between the transmission direction vector and the occlusion tendency vector. The occlusion tendency vector is the vector obtained by adding the occlusion tendency sub-vectors at several monitoring times.

[0094] Specifically, the preset threshold for the relative offset tendency parameter is the product of the relative offset tendency parameter reference value and the relative offset coefficient. The relative offset tendency parameter reference value is the average value of the relative offset tendency parameter under the same working conditions in historical data. The relative offset coefficient can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the larger the relative offset coefficient is set. The value range of the relative offset coefficient can be [1.15, 1.3], preferably 1.2.

[0095] Specifically, the capture control module is used to adjust the capture time, wherein,

[0096] The delay duration of the capture moment is negatively correlated with the relative offset tendency parameter.

[0097] Specifically, the delay duration is the relative offset tendency parameter reference value / relative offset tendency parameter × delay factor. The relative offset tendency parameter reference value is the average value of the relative offset tendency parameter under the same working conditions in historical data. The delay factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material table robot system. The value range of the delay factor can be [0.3, 0.7] to avoid the delay duration being set too large or too small. Preferably, the delay factor can be 0.4.

[0098] Specifically, in this embodiment of the invention, the grasping control module adjusts the grasping time of the material to be sorted within the feature grasping recognition area when the feature grasping recognition area meets the first grasping compensation condition. It is understood that in actual material sorting, the speed and direction of the obstructing object may dynamically change. By adjusting the delay in real time through the vector angle, the system can adapt to the obstruction rhythm under different working conditions. The feature grasping recognition area meets the first grasping compensation condition, meaning the relative motion trends of the obstructing object and the material show an inconsistent trend. The larger the relative offset tendency parameter, the faster the obstruction is released, and the shorter the delay time can be reduced. By reducing the dwell time of materials on the conveyor belt and avoiding subsequent material accumulation due to excessive waiting, the sorting volume per unit time can be increased. This avoids ineffective waiting and improves gripping efficiency. The smaller the relative offset tendency parameter, the slower the speed of obstruction removal. Extending the delay time can prevent the robotic arm from accidentally touching obstructions due to premature gripping, reducing the risk of material damage and equipment failure. Ensuring obstruction removal improves gripping accuracy, reduces energy consumption and mechanical wear, and dynamically adapts to the obstruction rhythm. In this way, the gripping compensation method can be adaptively adjusted based on the changing trend of dynamic obstruction, thereby improving the efficiency and reliability of the intelligent material table robot system.

[0099] Specifically, it can be understood that the angle between the conveying direction vector and the occlusion tendency vector can characterize the separation efficiency of the relative motion between the material and the occluder. The larger the relative offset tendency parameter, i.e., the larger the angle, the more significant the difference in the motion direction between the occluder and the material to be grasped, the faster the relative separation speed, and the shorter the duration of the occlusion state, such as shadow coverage. The smaller the relative offset tendency parameter, i.e., the smaller the angle, the slower the separation speed and the longer the occlusion duration. Based on the relative motion speed between the occluder and the material, the time window for occlusion removal is dynamically matched to achieve precise quantification of the delay time. Thus, the grasping compensation method is adaptively adjusted based on the changing trend of dynamic occlusion, thereby improving the efficiency and reliability of the intelligent material platform robot system.

[0100] Specifically, the grasping control module is used to determine whether there is a risk of grasping anomalies in the feature grasping and recognition area, wherein,

[0101] The grasping control module determines that there is a risk of grasping anomalies in the feature grasping recognition area based on the judgment result that the occlusion tendency vector of the feature grasping recognition area meets the grasping anomaly conditions, and issues an anomaly warning signal.

[0102] Based on the determination result that the occlusion tendency vector of the feature grasping recognition area does not meet the grasping anomaly conditions, it is determined that there is no grasping anomaly risk in the feature grasping recognition area, and the materials to be sorted in the feature grasping recognition area are grasped based on the grasping compensation parameters.

[0103] The capture anomaly condition is that the risk tendency parameter exceeds a preset risk tendency parameter threshold. The risk tendency parameter is the absolute value of the difference in the magnitude of the occlusion tendency vector at adjacent acquisition times.

[0104] Specifically, the preset risk tendency parameter threshold is the product of the risk tendency parameter reference value and the risk tendency factor. The risk tendency parameter reference value is the average value of the risk tendency parameter under the same working conditions in historical data. The risk tendency factor can be set by those skilled in the art according to the accuracy requirements of the intelligent material platform robot system. The higher the accuracy requirement, the smaller the risk tendency factor is set. The value range of the risk tendency factor can be [1.2, 1.35], preferably 1.25.

[0105] Specifically, the grasping control module is used to determine grasping compensation parameters, which include grasping direction and grasping speed, wherein,

[0106] The grasping direction is the direction of the material tendency vector, which is the vector obtained by adding the occlusion tendency vector and the conveying direction vector.

[0107] The decrease in the grasping speed is positively correlated with the magnitude of the occlusion tendency vector.

[0108] Specifically, the reduction in grasping speed is calculated as the current grasping speed × the size of the occlusion tendency vector × the speed control factor. The speed control factor can be set by those skilled in the art based on the accuracy requirements of the intelligent material table robot system. The value range of the speed control factor can be [0.2, 0.4] to avoid the speed reduction being too large or too small. Preferably, the speed control factor can be 0.3.

[0109] Specifically, in this embodiment of the invention, the grasping control module determines whether there is a grasping anomaly risk in the feature grasping recognition area when it does not meet the first grasping compensation condition. The grasping compensation parameters are then adjusted. It can be understood that the feature grasping recognition area not meeting the first grasping compensation condition means that the relative movement trend of the occluder and the material is relatively consistent, and the shadow cast by the occluder on the material will not change due to material transfer. By monitoring the rate of change of the occlusion tendency vector magnitude, i.e., the risk tendency parameter, the risk of changes in the occlusion state can be quickly identified, such as sudden large-area occlusion or material position shift caused by rapid movement of the occluder. For situations with abnormal risks, the process can be paused or warned in advance to avoid performing grasping when the occlusion is unstable. The grasping direction is determined based on the material tendency vector, which is equivalent to incorporating the indirect impact of occlusion on the material into the trajectory calculation. The gripping action tracks the actual position of the material more accurately. The gripping speed adjustment is determined based on the magnitude of the occlusion trend vector, achieving a dynamic balance where accuracy is prioritized when the occlusion influence is strong, and efficiency is prioritized when the influence is weak. The larger the magnitude of the occlusion trend vector, the greater the speed adjustment is required, such as decelerating to improve positioning accuracy or accelerating to escape the occlusion area, ensuring that the gripping action can still be executed stably under interference. The smaller the magnitude of the occlusion trend vector, the smaller the speed adjustment, avoiding the decrease in sorting efficiency caused by excessive deceleration. Through a combination of risk assessment and dynamic parameter adjustment strategies, the intelligent material table robot system can maintain efficient gripping under dynamic occlusion conditions, balancing gripping accuracy and efficiency while effectively avoiding potential abnormal risks. Furthermore, it achieves adaptive adjustment of the gripping compensation method based on the changing trend of dynamic occlusion, improving the efficiency and reliability of the intelligent material table robot system.

[0110] Specifically, it can be understood that the risk propensity parameter characterizes the rate of change of the occlusion tendency, and the magnitude of the occlusion tendency vector characterizes the intensity of the occlusion's influence on the material. The larger the magnitude, the more significant the change in occlusion shadow density. The absolute value of the difference in magnitude can characterize the rate of change of this influence intensity. The larger the risk propensity parameter, the more drastic the occlusion state changes in a short period of time, which may lead to rapid changes in material surface characteristics such as gripping points. At this time, gripping is prone to abnormal risks such as positioning deviation and collision with occluders. By adjusting the gripping direction and speed, the implicit interference of stable occlusion on the material's movement trajectory can be offset. The actual movement trend of the material is the indirect influence of the transmission force (transmission direction vector) and the occlusion, such as the slight pushing or frictional force that the occluder may exert on the material. This is manifested as the superposition of occlusion tendency vectors. The material tendency vector can characterize the actual movement direction of the material under occlusion interference. The larger the magnitude of the occlusion tendency vector, the stronger the impact of occlusion on the material. For example, if the occlusion range is large, the visual positioning error caused by shadow is large, or the interaction between the occluder and the material is more significant, the speed adjustment increases with the increase of the magnitude. The interference is offset by dynamically adapting the speed. The larger the magnitude of the occlusion tendency vector, the more necessary it is to slow down the speed to improve the grasping and positioning accuracy. The smaller the magnitude of the occlusion tendency vector, the smaller the speed reduction, avoiding over-adjustment that leads to efficiency loss. Thus, the grasping compensation method is adaptively adjusted based on the changing trend of dynamic occlusion, improving the efficiency and reliability of the intelligent material table robot system.

[0111] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An AI vision technology-based intelligent magazine robot system, characterized in that, The method comprises the following steps: a material table control module comprising a grabbing unit for grabbing the to-be-sorted material on the material table and a conveying unit for conveying the to-be-sorted material; a feature acquisition module connected to the material table control module for acquiring the gray scale characteristic parameters and the conveying direction vector of the to-be-sorted material; a material analysis module connected to the feature acquisition module for dividing the material table into a plurality of grabbing recognition areas, determining the gray scale fluctuation of the to-be-sorted material according to the gray scale characteristic parameters in the grabbing recognition areas under different illumination angles, and marking the feature grabbing recognition area; the material analysis module is used to determine the gray scale fluctuation characteristic quantity and the shielding tendency coefficient based on the gray scale characteristic parameters in the grabbing recognition area, wherein the gray scale fluctuation characteristic quantity is the difference between the maximum value and the minimum value of the gray scale characteristic parameters in the grabbing recognition area under the same illumination angle, and the shielding tendency coefficient is the variance of the gray scale characteristic parameters under different illumination angles; the material analysis module is used to mark the grabbing recognition area as a feature grabbing recognition area based on the determination result that the gray scale fluctuation characteristic quantity and the shielding tendency coefficient in the grabbing recognition area under different illumination angles meet the feature grabbing recognition area condition, wherein the feature grabbing recognition area condition is that the gray scale fluctuation characteristic quantity exceeds the preset gray scale fluctuation characteristic quantity threshold value, and the shielding tendency coefficient does not exceed the preset shielding tendency coefficient threshold value; a material recognition module connected to the feature acquisition module and the material analysis module respectively, for determining a shielding tendency sub-area according to the gray scale characteristic parameters of the feature grabbing recognition area under a preset illumination angle, and constructing a shielding tendency sub-vector based on a plurality of acquired shielding tendency sub-areas; a grabbing control module connected to the material table control module, the feature acquisition module, and the material recognition module, for determining the grabbing compensation mode for the to-be-sorted material according to the comparison between the conveying direction vector and the shielding tendency vector, adjusting the grabbing time for the to-be-sorted material in the feature grabbing recognition area; or determining whether the feature grabbing recognition area has a grabbing abnormal risk according to the shielding tendency vector, and adjusting the grabbing compensation parameters of the feature grabbing recognition area; wherein the shielding tendency vector is determined according to a plurality of shielding tendency sub-vectors. 2.The AI vision technology-based intelligent magazine robot system according to claim 1, wherein, the material recognition module is used to determine the feature sub-area of the feature grabbing recognition area as the shielding tendency sub-area based on the determination result that the gray scale characteristic parameters of the feature sub-area meet the shielding tendency sub-area condition, wherein the material recognition module divides the feature grabbing recognition area into a plurality of sub-areas and acquires the gray scale characteristic parameters of the to-be-sorted material at a plurality of position points in the sub-areas; the shielding tendency sub-area condition is that the gray scale characteristic parameters of the feature sub-area do not exceed the preset gray scale characteristic parameter threshold value, and the feature sub-area is the sub-area where the minimum value of the gray scale characteristic parameters is located. 3.The AI vision technology-based intelligent magazine robot system according to claim 2, wherein, the material recognition module is used to construct the shielding tendency sub-vector of the latter acquisition time among adjacent acquisition times within a preset monitoring period based on the shielding tendency sub-areas under the adjacent acquisition times, wherein The material recognition module acquires the shielding tendency sub-regions at several collection time points in the feature grabbing recognition region; The shielding tendency sub-vector is constructed with the region center point of the shielding tendency sub-region at the previous collection time point in the adjacent collection time points as the vector starting point of the shielding tendency sub-vector, and with the region center point of the shielding tendency sub-region at the next collection time point in the adjacent collection time points as the vector terminal point of the shielding tendency sub-vector. 4.The AI vision technology-based intelligent magazine robot system according to claim 3, wherein, The grabbing control module is used to determine the grabbing compensation mode of the material to be sorted in the feature grabbing recognition region, wherein, if the transmission direction vector and the shielding tendency vector of the material to be sorted in the feature grabbing recognition region meet the first grabbing compensation condition, the grabbing control module determines that the grabbing compensation mode is to adjust the grabbing time of the material to be sorted in the feature grabbing recognition region; if the transmission direction vector and the shielding tendency vector of the material to be sorted in the feature grabbing recognition region do not meet the first grabbing compensation condition, the grabbing control module determines that the grabbing compensation mode is to determine whether there is a grabbing abnormal risk in the feature grabbing recognition region, and to adjust the grabbing compensation parameter. 5.The AI vision technology-based intelligent magazine robot system according to claim 4, wherein, The first grabbing compensation condition is that the relative offset tendency parameter exceeds a preset relative offset tendency parameter threshold, the relative offset tendency parameter is the vector angle between the transmission direction vector and the shielding tendency vector, and the shielding tendency vector is the vector obtained by adding the shielding tendency sub-vectors at the several monitoring time points. 6.The AI vision technology-based intelligent magazine robot system according to claim 5, wherein, The grabbing control module is used to adjust the grabbing time, wherein, the delay length of the grabbing time is in a negative correlation with the relative offset tendency parameter. 7.The AI vision technology-based intelligent magazine robot system according to claim 6, wherein, The grabbing control module is used to determine whether there is a grabbing abnormal risk in the feature grabbing recognition region, wherein, the grabbing control module determines that there is a grabbing abnormal risk in the feature grabbing recognition region based on the determination result that the shielding tendency vector of the feature grabbing recognition region meets the grabbing abnormal condition; based on the determination result that the shielding tendency vector of the feature grabbing recognition region does not meet the grabbing abnormal condition, it is determined that there is no grabbing abnormal risk in the feature grabbing recognition region, and the material to be sorted in the feature grabbing recognition region is grabbed based on the grabbing compensation parameter; The grabbing abnormal condition is that the risk tendency parameter exceeds a preset risk tendency parameter threshold, and the risk tendency parameter is the absolute value of the difference between the lengths of the shielding tendency vectors at the adjacent collection time points. 8.The AI vision technology-based intelligent magazine robot system according to claim 7, wherein, The grabbing control module is used to determine the grabbing compensation parameter, and the grabbing compensation parameter includes the grabbing direction and the grabbing speed, wherein, the grabbing direction is the vector direction of the material tendency vector, and the material tendency vector is the vector obtained by adding the shielding tendency vector and the transmission direction vector; the reduction of the grabbing speed is in a positive correlation with the vector size of the shielding tendency vector.

Images